Handbook in Operations Research

and Management Science

Volume on \Networks"
Chapter on \Design of Survivable Networks"

by
M. Grotschel*
C.L. Monma**
M. Stoer*

March 1993

*Konrad-Zuse-Zentrum fur
Informationstechnik Berlin
Heilbronner Str. 10
D-1000 Berlin 31, Germany

**Bellcore
445 South Street
Morristown, New Jersey 07960

1
Contents
1 Overview 3

2 Motivation 3

3 Integer Programming Models of Survivability 6

3.1 The General Model for Undirected Networks . . . . . . . . . . . . . . . . . . 7
3.2 A Brief Discussion of the Model . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.3 A Model Used in Practice . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

4 Structural Properties and Heuristics 15

4.1 Lifting and the Structure of Optimum Solutions . . . . . . . . . . . . . . . . 15
4.2 Construction and Improvement Heuristics . . . . . . . . . . . . . . . . . . . 17
4.3 Heuristics with Performance Guarantees . . . . . . . . . . . . . . . . . . . . 20

5 Polynomially Solvable Special Cases 23

5.1 Node Type Restrictions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
5.2 Cost Restrictions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
5.3 Special Classes of Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

6 Polyhedral Results 27
6.1 Classes of Valid Inequalities . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
6.2 Facet Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
6.3 Separation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
6.4 Complete Descriptions of Small Cases . . . . . . . . . . . . . . . . . . . . . . 37
6.4.1 The 2ECON Polytope for K5 . . . . . . . . . . . . . . . . . . . . . . 38
6.4.2 The 2NCON Polytope for K5 . . . . . . . . . . . . . . . . . . . . . . 39

2
7 Computational Results 44
7.1 Outline of the Cutting Plane Algorithm . . . . . . . . . . . . . . . . . . . . . 44
7.2 Implementation Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
7.3 Computational Results for Low-Connectivity Problems . . . . . . . . . . . . 48
7.4 Computational Results for High-Connectivity Problems . . . . . . . . . . . . 56
7.4.1 Random Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
7.4.2 Ship Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

8 Directed Variants of the General Model 65

8.1 Survivability Models for Directed Networks . . . . . . . . . . . . . . . . . . . 65
8.2 Projection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

3
1 Overview
This chapter focuses on the important practical and theoretical problem of designing sur-
vivable communication networks, i.e., communication networks that are still functional after
the failure of certain network components. We motivate this topic in Section 2 by using
the example of ber optic communication network design for telephone companies. A very
general model (for undirected networks) is presented in Section 3 which includes practical,
as well as theoretical, problems, including the well-studied minimum spanning tree, Steiner
tree, and minimum cost k-connected network design problems.
The development of this area starts with outlining structural properties in Section 4 which
are useful for the design and analysis of algorithms for designing survivable networks. These
lead to worst-case upper and lower bounds. Heuristics that work well in practice are also
described in Section 4. Polynomially-solvable special cases of the general survivable network
design problem are summarized in Section 5.
Section 6 contains polyhedral results from the study of these problems as integer program-
ming models. We present large classes of valid inequalities and indicate when they are facet-
dening. We also summarize the complexity of the separation problem for these inequalities.
Finally we provide complete and nonredundant descriptions of a number of polytopes related
to network survivability problems of small dimensions.
Section 7 contains computational results using cutting plane approaches based on the poly-
hedral results of Section 6 and the heuristics described in Section 4. The results show that
these methods are ecient and eective in producing optimal or near-optimal solutions to
real-world problems.
A brief review of the work on survivability models of directed networks is given in Section 8.
We also show here how directed models can help to solve undirected cases.

2 Motivation
In this section we set the stage for the topic of this chapter by considering an application to
designing communication networks for telephone companies based on ber optic technology.
We will use this to introduce the concept of survivability in network design and to motivate
the general optimization models described in the next section. It will become clear later that
our models capture many other situations that arise in practice and in theory as well.

4
Fiber optic technology is rapidly being deployed in communication networks throughout the
world because of its nearly-unlimited capacity, its reliability and cost-eectiveness. The high
capacity of new technology ber transmission systems has resulted in the capability to carry
many thousands of telephone conversations and high-speed data on a few strands of ber.
These advantages oer the prospect of ushering in many new information networking services
which were previously either technically impossible or economically infeasible.
The economics of ber systems dier signicantly from the economics of traditional copper-
based technologies. Copper-based technologies tend to be bandwidth limited. This results
in a mesh-like network topology, which necessarily has many diverse paths between any
two locations with each link carrying only a very small amount of trac. In contrast, the
high-capacity ber technologies tend to suggest the design of sparse tree-like network
topologies. These topologies have only a few diverse paths between locations (often just
a single path) and each link has a very high trac volume. This raises the possibility of
signicant service disruptions due to the failure of a single link or single node in the network.
The special report Keeping the phone lines open by Zorpette (1989) describes the many
man-made and natural causes that can disrupt communication networks, including res,
tornados, oods, earthquakes, construction activity and terrorist activities. Such failures
occur surprisingly frequently and with devastating results as described in this report and
in the popular press (e.g., see Newark Star Ledger (1987,1988a,1988b), New York Times
(1988,1989), Wall Street Journal (1988)).
Hence, it is vital to take into account such failure scenarios and their potential negative
consequence when designing ber communication networks. Recall that one of the major
functions of a communication network is to provide connectivity between users in order to
provide a desired service. We use the term survivability to mean the ability to restore net-
work service in the event of a catastrophic failure, such as the complete loss of a transmission
link or a facility switching node. Service could be restored by means of routing trac around
the damage through other existing facilities and switches, if this contingency is provided for
in the network architecture. This requires additional connectivity in the network topology
and a means to automatically reroute trac after the detection of a failure.
A network topology could provide protection against a single link failure if it remains con-
nected after the failure of any single link. Such a network is called two-edge connected
since at least two edges have to be removed in order to disconnect the network. However, if
there is a node in the network whose removal does disconnect the network, such a network
would not protect against a single node failure. Protection against a single node failure can
be provided in an analogous manner by two-node connected networks.
In the case of ber communication networks for telephone companies, two-connected topolo-
gies provide an adequate level of survivability since most failures usually can be repaired

5
relatively quickly and, as statistical studies have revealed, it is unlikely that a second failure
will occur in the duration. However, for other applications it may be necessary to provide
higher levels of connectivity.
One simple and cost eective means of using a two-connected topology to achieve an auto-
matic recovery from a single failure is called diverse protection routing. Most ber transmis-
sion systems employ a protection system to back up the working ber systems. An automatic
protection switch detects failure of a working system and switches the working service auto-
matically to the protection system. By routing the protection system on a physically diverse
route from the working system, one provides automatic recovery from a single failure at the
small cost of a slightly longer path for the protection system than if it were routed on the
same physical path as the working system. One would suspect, and studies have proven, that
the additional cost of materials and installation of diverse ber routes would be acceptable
for the benets gained (see Wu, Kolar and Cardwell (1988a, 1988b), Kolar and Wu (1988),
and Cardwell, Wu and Woodall (1988)).
We consider the problem of designing minimum cost networks satisfying certain connectivity
requirements that model network survivability. A formal model will be described in the next
section. This model allows for dierent levels of survivability that arise in practical and
theoretical models including the minimum spanning tree, Steiner tree, and minimum cost
k-connected network design problems.
The application described here requires three distinct types of nodes: special oces, which
must be protected from single edge or node failures, ordinary oces, which need only be
connected by a single path, and optional oces, which may be included or excluded from
the network design depending only upon cost considerations. The designation of oce type
is performed by a planner based on a cost/benet analysis. Normally, the special oces are
highly important and/or high-revenue-producing oces, with perhaps a high proportion of
priority services. It may not be economically possible to ensure the service of ordinary or
optional oces in the face of potential failures. In fact, some oces only have one outlet to
the network and so it would be technologically impossible to provide recovery if this path
were blocked by a network failure.
We note that two-connected network topologies are cost eective. For example, a typical
real-world problem which we solved has a survivable network with cost of only 6% above the
minimum spanning tree; however, a single failure in the tree network could result in the loss
of 33% of the total trac while the survivable network could lose at most 3% from any single
failure. We also note that the heuristic methods described in Section 4 and the polyhedral
cutting plane methods described in Section 6 are ecient and very eective in generating
optimal and near-optimal network topologies as we describe in Section 7.

6
We conclude this section by pointing out that the topology design problem considered here
is just the rst step in the overall design of ber communications networks. There are other
issues that need to be addressed once the topology is in place. For instance, demands for
services are usually in units of circuits, called DS0 rate, or bundles of 24 circuits, called DS1
rate. Fiber optic transmission rates come in units of 28 DS1s or 672 DS0s, called DS3 rate.
Hence, it is necessary to place multiplexing equipment to convert between the three rates,
and to route these demands through the network. Another issue is that the ber signals
need to be amplied using repeater equipment if the signals travel beyond a given threshold
distance.
Furthermore, the network is generally organized into a facility hierarchy. That is, oces
are grouped together into clusters, with each cluster having one hub oce to handle trac
between clusters. This grouping considers such factors as community-of-interest and geo-
graphic area. Groups of clusters can be grouped into sectors, with each sector having one
gateway oce, which is a hub building designated to handle inter-sector trac. This hier-
archy allows trac to be concentrated into high capacity routes to central locations where
the demands are sorted according to destination. This concept of hub routing has proven
to give near-optimal results, see Wu and Cardwell (1988). A two-connected network allows
for dual homing, i.e., for the possibility of splitting the demand at an oce between a home
hub and foreign hub to protect against a hub failure.
Given the complexity of the overall design problem, these issues are generally handled in
a sequential fashion with the topology question being decided rst. This process is often
iterated until an acceptable network design is obtained. In this chapter, we only deal with
the network topology design aspect and not the multiplexing or bundling aspects. For an
overview of this combined process and a description of a computer-based planning tool
(Bellcore (1988)) incorporating all of these features, see Cardwell, Monma and Wu (1989).

3 Integer Programming Models of Survivability

In this section, we formalize the undirected survivable network design problems that are
being considered in this chapter. Variants that are based on directed networks will be briey
treated in Section 8.

7
3.1 The General Model for Undirected Networks
A set V of nodes is given representing the locations of the oces that must be interconnected
into a network in order to provide the desired services. A collection E of edges is also
specied that represent the possible pairs of nodes between which a direct transmission
link can be placed. We let G = (V, E) be the (undirected) graph of possible direct link
connections. Each edge e E has a nonnegative fixed cost ce of establishing the direct
link connection. For our range of applications, loops are irrelevant. Thus we assume in the
following that graphs have no loops. Parallel transmission links occur in practice; therefore,
our graphs may have parallel edges. For technical reasons to be explained later, we will
however restrict ourselves in this paper to simple graphs (i.e., loopless graphs without parallel
edges) when we consider node survivability problems and node connectivity.
The cost of establishing a network consisting of a subset F E of edges is the sum of the
costs of the individual links contained in F . The goal is to build a minimum-cost network so
that the required survivability conditions, which we describe below, are satised. We note
that the cost here represents setting up the topology for the communication network and
includes placing conduits in which to lay the ber cable, placing the cables into service, and
other related costs. We do not consider costs that depend on how the network is implemented
such as routing, multiplexing, and repeater costs. Although these costs are also important,
it is (as mentioned in Section 2) usually the case that a topology is designed rst and then
these other costs are considered in a second stage of optimization.
If G = (V, E) is a graph, W V and F E, then we denote by G W and G F the graph
that is obtained from G by deleting the node set W and the edge set F , respectively. For
notational convenience we write Gv and Ge instead of G{v} and G{e}, respectively.
The dierence of two sets M and N is denoted by M\N.
For any pair of distinct nodes s, t V , an [s, t]-path P is a sequence of nodes and edges
(v0, e1, v1 , e2, . . . , vl1 , el, vl ), where each edge ei is incident with the nodes vi1 and vi (i =
1, . . . , l), where v0 = s and vl = t, and where no node or edge appears more than once
in P . A collection P1 , P2 , . . . , Pk of [s, t]-paths is called edge-disjoint if no edge appears in
more than one path, and is called node-disjoint if no node (except for s and t) appears in
more than one path. In standard graph theory two parallel edges are not considered as node
disjoint paths. For our applications it is sensible to do so. However, this modication would
lead to considering nonstandard variations of node connectivity, reformulations of Mengers
theorem etc. In order not to trouble the reader with these technicalities we have decided
to restrict ourselves to simple graphs when node disjoint paths are treated. The results
presented in this paper carry, appropriately stated, over to the case where parallel edges are
considered as node disjoint paths. This theory is developed in Stoer (1992).

8
Two dierent nodes s, t of a graph are called k-edge (resp. k-node) connected if there are
k edge-disjoint (resp. node-disjoint) paths between s and t. A graph with at least two nodes
is called k-edge or k-node connected if all pairs of distinct nodes of G are k-edge or k-node
connected, respectively. For our purposes, the graph K1 consisting of just one node is k-edge
and k-node connected for every natural number k. For a graph G = K1 , the largest integer k
such that G is k-edge connected (resp. k-node connected) is denoted by (G) (resp. (G))
and is called the edge connectivity (resp. node connectivity) of G.
The survivability conditions require that the network satisfy certain edge and node connec-
tivity requirements. To specify these, three nonegative integers rst , kst and dst are given for
each pair of distinct nodes s, t V . The numbers rst represent the edge survivability
requirements, and the numbers k st and dst represent the node survivability require-
ments; this means that the network N = (V, F ) to be designed has to have the property
that, for each pair s, t V of distinct nodes, N must contain at least rst edge disjoint
[s, t]-paths, and that the removal of at most k st nodes (dierent from s and t) from N must
leave at least dst edge disjoint [s, t]-paths. (Clearly, we may assume that k st |V | 2 for all
s, t V , and we will do this throughout this chapter). These conditions ensure that some
communication path between s and t will survive a prespecied level of combined failures of
both nodes and links. The levels of survivability specied depend on the relative importance
placed on maintaining connectivity between dierent pairs of oces.
Let us now introduce a variable xe for each edge e E, and consider the vector space RE .
Every subset F E induces an incidence vector F = (Fe )eE RE by setting Fe := 1
if e F , Fe := 0 otherwise; and vice versa, each 0/1-vector x RE induces a subset
F x := {e E | xe = 1} of the edge set E of G. If we speak of the incidence vector of a path
in the sequel we mean the incidence vector of the edges of the path. We can now formulate
the network design problem introduced above as the following integer linear program.


(3.1) min ijE cij xij

subject to
 
(i) iW jV \W xij rst for all pairs s, t V, s = t and
for all W V with s W, t  W ;
 
(ii) iW jV \(ZW ) xij dst for all pairs s, t V, s = t; and
for all Z V \ {s, t} with |Z| = kst and
for all W V \ Z; with s W, t  W ;

(iii) 0 xij 1 for all ij E;

9
 (iv) xij integral for all ij E.

Note that if, for each pair s, t of distinct nodes in V and for each set Z V \ {s, t} with
|Z| = kst , N Z contains at least dst edge disjoint [s, t]-paths, then all node survivability
requirements are satised, i.e., inequalities of type (ii) need not be considered for node sets
Z V \ {s, t} with |Z| < kst . It follows from Mengers theorem (see Chapter ??) that,
for every feasible solution x of (3.1), the subgraph N = (V, F x) of G denes a network that
satises the given edge and node survivability requirements.
This model, introduced in Grotschel & Monma (1990), generalizes and unies a number of
problems that have been investigated in the literature either from a practical or theoretical
point of view. We mention here some of these cases.
The classical network synthesis problem for multiterminal ows (see Chapter 13) is obtained
from (3.1) by dropping the constraints (ii) and (iv). In the standard formulation of the
network synthesis problem, the upper bounds xe 1 are not present. But our model
allows parallel edges in the underlying direct-link graph, or equivalently, allows the upper
bound in constraints (iii) to take on any nonnegative values for each edge. This linear
programming problem has a number of constraints that is exponential in the number of
nodes of G. For the case cij = c for all ij E, where c is a constant, Gomory & Hu (1961)
found a simple algorithm for its solution. Bland, Goldfarb & Todd (1981) pointed out that
the separation problem for the class of inequalities (i) can be solved in polynomial time by
computing a minimum capacity cut; thus, it follows by the ellipsoid method that the classical
network synthesis problem can be solved in polynomial time. (See Grotschel, Lovasz and
Schrijver (1988) for details on the ellipsoid algorithm and its applications to combinatorial
optimization.) The integer network synthesis problem, i.e., the problem obtained from (3.1)
by dropping constraints (ii) and the upper bounds xe 1, was solved by Chou & H. Frank
(1970), for the case cij = c for all ij V V , and rij 2 for all ij.
The minimum spanning tree problem can be phrased as the task to nd a minimum-cost
connected subset F E of edges spanning V (see Chapter 12). This problem can be viewed
as a special case of (3.1) as follows. We drop the constraints (ii) and set rst := 1 for all
distinct s, t V in constraints (i).
Similarly, the closely-related Steiner tree problem is to nd a minimum-cost connected subset
F E of edges which span a specied subset S V of nodes. This problem is a special
case of (3.1) where we drop constraints (ii) and set rst := 1 in constraints (i), for all s, t S,
and rst := 0 otherwise. (See also Chapter 12.) Let us remark at this point that the Steiner
tree problem is well-known to belong to the class of NP-hard problems. As it is a special

10
case of (3.1), our general problem of designing survivable networks is NP-hard as well. (See
Garey & Johnson (1979) for details on the theory of NP-completeness.)
The problem of nding a minimum-cost k-edge connected network in a given graph is a
special case of (3.1) where all inequalities (ii) are dropped and where rst = k for all distinct
s, t V . The problem of nding an optimal k-node connected network, for k |V | 1, is a
special case of (3.1) where we drop the constraints (i) and set kst := k 1 and dst := 1 for
all distinct s, t V .

3.2 A Brief Discussion of the Model

The rest of this chapter is mainly devoted to studying various aspects of model (3.1) and
some of its special cases. Among other subjects, we describe heuristics to compute upper
and lower bounds for the optimum value of (3.1). To compute a lower bound one is naturally
led to dropping the integrality constraints (iv) and solving the LP-relaxation (3.1) (i), (ii),
(iii) of the survivable network design problem.
Two questions immediately arise. Can one solve this LP despite the fact that it has ex-
ponentially many constraints? Can one nd a better (or a series of increasingly better)
LP-relaxations that are solvable in polynomial (or practically reasonable) time?
We will address the rst question in Section 7 and the second in Section 6. But here we would
like to give a glimpse at the approach that leads to answering these questions. The method
involved is known as polyhedral combinatorics. It is a vehicle to provide (in some sense) the
best possible LP-relaxation. We want to demonstrate now how the second question leads
to the investigation of certain integral polyhedra in a very natural way. See Grotschel &
Padberg (1985) and Pulleyblank (1990) for a survey on polyhedral combinatorics.
To obtain a better LP-relaxation of (3.1) than the one arising from dropping the integrality
constraints (iv), we dene the following polytope. Let G = (V, E) be a graph, let EV :=
{st | s, t V, s = t}, and let r, k, d Z E V
+ be given. Then

(3.2) CON(G; r, k, d) := conv{x RE | x satises (i), . . . , (iv) of (3.1)}

is the polytope associated with the network design problem given by the graph G and the
edge and node survivability requirements r, k, and d. (Above conv denotes the convex
hull operator.) In the sequel, we will study CON(G; r, k, d) for various special choices of
r, k and d. Let us mention here a few general properties of CON(G; r, k, d) that are easy to
derive.

11
Let G = (V, E) be a graph and r, k, d ZE + be given as above. We say that e E is
V

essential with respect to (G; r, k, d) (short: (G; r, k, d)-essential) if CON(G e; r, k, d) =

. In other words, e is essential with respect to (G; r, k, d) if its deletion from G results in a
graph such that at least one of the survivability requirements cannot be satised. We denote
the set of edges in E that are essential with respect to (G; r, k, d) by ES(G; r, k, d). Clearly,
for all subsets F E \ ES(G; r, k, d), ES(G; r, k, d) ES(G F ; r, k, d) holds. Let dim(S)
denote the dimension of a set S Rn , i.e., the maximum number of anely independent
elements in S minus 1. Then one can easily prove the following two results (see Grotschel
and Monma (1990)).

+ such that CON(G; r, k, d) = .

(3.3) Theorem. Let G = (V, E) be a graph and r, k, d Z E V

Then
CON(G; r, k, d) {x RE | xe = 1 for all e ES(G; r, k, d)},
dim(CON(G; r, k, d)) = |E| |ES(G; r, k, d)|.2

An inequality aT x is valid with respect to a polyhedron P , if P {x | aT x }; the set

Fa := {x P | aT x = } is called the face of P dened by aT x . If dim(Fa) = dim(P )1
and Fa = then Fa is a facet of P , and aT x is called facet-defining or facet-inducing.
(3.4) Theorem. Let G = (V, E) be a graph and r, k, d ZE
+ such that CON(G; r, k, d) = .
V

Then

(a) xe 1 denes a facet of CON(G; r, k, d) if and only if e E \ ES(G; r, k, d);

(b) xe 0 denes a facet of CON(G; r, k, d) if and only if e E \ ES(G; r, k, d) and

ES(G; r, k, d) = ES(G e; r, k, d). 2

Theorems (3.3) and (3.4) solve the dimension problem and characterize the trivial facets.
But these characterizations are (in a certain sense) algorithmically intractible as the next
observation shows which follows from results of Ling and Kameda (1987).
(3.5) Remark. The following three problems are NP-hard.

Instance: A graph G = (V, E) and vectors r, k, d ZE

+.

Question 1: Is CON(G; r, k, d) empty?

Question 2: Is e E (G; r, k, d)-essential?

Question 3: What is the dimension of CON(G; r, k, d)? 2

12
However, for most cases of practical interest in the design of survivable networks, the sets
ES(G; r, k, d) of essential edges can be determined easily, and thus the trivial LP-relaxation
of (3.1) following from (3.3) and (3.4) can be set up without diculties. (We will comment
on this in the sequel.)
Before continuing, let us remark that there is an easy way to improve upon the formulation
of (3.1) by excluding a number of inequalities that are obviously redundant.
Given G = (V, E) and r, k, d ZE V
+ , extend the functions r and d to functions operating on
sets by setting

(3.6) con(W ) := max{rst | s W, t V \W }

and

(3.7) d(Z, W ) := max{dst | s W \Z, t V \(ZW ), kst |Z|} for Z, W V.

We call a pair (Z, W ) with Z, W V eligible (with respect to k) if Z W = and |Z| = kst
for at least one pair of nodes with s W and t V \ (Z W ). Then CON(G; r, k, d) is
clearly contained in the solution set of the following system of equations and inequalities.
 
(3.8) (i) iW jV \W xij con(W ) for all W V, = W = V ;
 
(ii) iW jV \(ZW ) xij d(Z, W ) for all eligible pairs (Z, W )
of subsets of V ;
(iii) xij 1 for all ij E\ES(G; r, k, d);

(iv) xij = 1 for all ij ES (G; r, k, d);

(v) xij 0 for all E \ ES(G; r, k, d) with

ES(G; r, k, d) = ES(G e; r, k, d).

Of course, CON(G; r, k, d) is the convex hull of the integral solution of (3.8).

3.3 A Model Used in Practice

The model discussed so far mixes node and edge connectivity, and provides, for each pair
of nodes, the possibility to specify particular connectivity requirements. It is, as mentioned
before, a quite general framework that models many practical situations simultaneously.
This generality demands a considerable amount of data. In our real-world application, it

13
turned out that the network designers were either interested in node connectivity or in edge
connectivity requirements but not in both simultaneously. Also, the data for implementing
the general model were not available in practice. A specialized version, to be described below,
proved to be acceptable from the point of view of data acquisition and was still considered
a reasonable model of reality by practitioners.
To model these (slightly more restrictive) survivability conditions, we introduce the concept
of node types. For each node s V a nonnegative integer rs , called the type of s, is specied.
For any W V , the integer r(W ) := max{rv | V W } is called the type of W . We say
that the network N = (V, F ) to be designed satises the node survivability conditions
if, for each pair s, t V of distinct nodes, N contains at least

(3.9) rst := min{rs , rt }

node disjoint [s, t]-paths. Similarly, we say that N = (V, F ) satises the edge survivability
conditions if, for each pair s, t V of distinct nodes, N contains rst edge disjoint [s, t]-
paths. These conditions ensure that some communication path between s and t will survive
a prespecied level of node or link failures. We will discuss these special cases in more
detail later. To make the (somewhat clumsy) general notation easier, we introduce further
symbols and conventions to denote these node- or edge-survivability models. Given a graph
G = (V, E) and a vector of node types r = (rs )sV we assume without loss of generality
that there are at least two nodes of the largest type. If we say that we consider the kNCON
problem (for G and r) then we mean that we are looking for a minimum-cost network that
satises the node survivability conditions and where k = max{rs |s V }. Similarly, we speak
of the kECON problem (for G and r), when we consider the edge survivability conditions.
Let G = (V, E) be a graph. For Z V , let G (Z) denote the set of edges with one
endnode in Z and the other in V \ Z. It is customary to call G (Z) a cut. If it is clear
with respect to which graph a cut G (Z) is considered, we simply drop the index and write
(Z). We also write (v) for ({v}). If X, Y are subsets of V with X Y = , we set
[X : Y ] := {ij E | i X, j Y }, thus (X) = [X : V \X]. For any subset of edges F E,

we let x(F ) stand for the sum eF xe . Consider the following integer linear program for a
graph G = (V, E) with edge costs ce for all e E and node types rs for all s V (using
(3.9) in the denition of con(W ) in (3.6)):

14
 (3.10) min cT x

subject to

(i) x((W )) con(W ) for all W V , = W = V ;

(ii) x(GZ (W )) con(W ) |Z| for all pairs s, t V, s = t, and

for all = Z V \ {s, t} with |Z| rst 1, and
for all W V \ Z with s W, t  W ;

(iii) 0 xij 1 for all ij E;

(iv) xij integral for all ij E.

It follows from Mengers theorem that the feasible solutions of (3.10) are the incidence vectors
of edge sets F such that N = (V, F ) satises all node survivability conditions; i.e., (3.10) is
an integer programming formulation of the kNCON problem. Deleting inequalities (ii) from
(3.10) we obtain, again from Mengers theorem, an integer programming formulation for the
kECON problem. The inequalities of type (i) will be called cut inequalities and those of
type (ii) will be called node cut inequalities.
The polyhedral approach to the solution of the kNCON (and similarly the kECON) prob-
lem consists of studying the polyhedron obtained by taking the convex hull of the feasible
solutions of (3.9). We set

NCON(G; r) := conv{x RE |x satises (3.10)(i), . . . , (iv)},

ECON(G; r) := conv{x RE |x satises. (3.10)(i), (iii), (iv)}.

It will sometimes be convenient to denote these polyhedra by kNCON(G; r) and kECON
(G; r), where k = max{rs |s V }, since this implicitly provides a notation for the maximum
node type. To tie this notation with the previously introduced more general concept, note
that

kECON(G; r) = CON(G; r , 0, 0),

where r RV V with rst


:= min{rs , rt } for all s, t V . Also, if there are no parallel edges

kNCON(G; r) = CON(G; r , k  , d ),
 
where kst := max{0, rst 1} for all s, t V and d := r k  .

15
This survivability model, the polyhedra, and the integer and linear programming problems
associated with the kECON and kNCON problems, will be studied in more detail in the
sequel.

4 Structural Properties and Heuristics

In this section we describe some heuristic approaches for the solution of ECON and NCON
problems. There are standard methods like greedy and interchange heuristics and heuristics
that are motivated by techniques for the approximate solution of other NP-hard problems,
like the (much investigated) traveling salesman problem. Some heuristics are more special-
purpose since they make use of structural properties of k-connected graphs. A few structural
results and their uses are reviewed in Sections 4.1 and 4.2. Section 4.2 concentrates on the
design of practically-eective heuristics for ECON and NCON problems, while Section 4.3
discusses lower bounds and heuristics with worst-case performance guarantees.

4.1 Lifting and the Structure of Optimum Solutions

Connectivity is a very rich and active topic of graph theory and it is conceivable that the
knowledge that has been accumulated in this area can be exploited further for the design of
eective approximation algorithms. We do not attempt to cover the work on connectivity
here in detail and refer to A. Frank (1992) for a comprehensive survey of connectivity results.
We will just mention a few structural properties of connected graphs that have been employed
for the design of practically-useful heuristics. We begin by describing a local construction
technique. Lovasz (1976) and Mader (1978) have proved so-called lifting theorems that
show that simple manipulations of a graph can be made without destroying certain edge
connectivity properties. These manipulations are useful for construction heuristics for the
ECON problem (where parallel edges are allowed) for general connectivity requirements r vw .
In order to state these results we have to introduce a few denitions. Let G = (V, E) be a
graph, and let x be a node of G that we call special. We assume that x is adjacent to distinct
nodes y and z. The graph G obtained from G by deleting the edges xy and xz and adding
the edge yz is called lifting of G at x.
If the edge-connectivity of G between any two nodes of V  x is not smaller than that of
G, then the lifting is called admissible.

16
(4.1) Theorem. [Mader 1978]. Let G = (V, E) be a graph with special node x.

(a) If the degree of x is at least 4 and x is not a cut node then G has an admissible lifting
at x.

(b) If x is a cut node and no single edge incident to x is a cut then G has an admissible
lifting at x. 2

(4.2) Theorem. [Lovasz 1976]. Let G = (V, E) be an Eulerian graph, let x be a (special)
node of even degree, and let y be any node adjacent to x in G. Then there is another neighbor
node z of x, so that the lifting at x involving y and z is admissible. 2
Consider the kECON and kNCON problems for the complete graph K n where rx = k for all
nodes x, and where the nonnegative costs satisfy the triangle inequality, i.e., cxz cxy + cyz
for all nodes x, y and z. Using the Lifting Theorems (4.1) and (4.2), Monma, Munson and
Pulleyblank (1990) showed the following result for kECON for Kn with k = 2.
(4.3) Theorem. Given costs satisfying the triangle inequality, there is an optimal k-edge
connected spanning network satisfying the following conditions:

(a) all nodes are of degree k or k + 1; and

(b) removing any set of at most k edges does not leave all of the resulting connected com-
ponents k-edge connected. 2

It is easy to see that if the cost function satises the triangle inequalities, there is an optimal
2-node connected solution with cost equal to an optimal 2-edge connected solution. In
fact, Monma, Munson and Pulleyblank (1990) show that the term k-edge can be replaced
by k-node throughout (4.3) to obtain a similar result for kNCON for Kn with k = 2.
Furthermore, they show that these characterize the optimal 2-connected networks in the
following sense: given any 2-connected graph G = (V, E) satisfying conditions (4.3) (a) and
(b) for k = 2, then there exist costs satisfying the triangle inequality such that G is the
unique optimal solution.
Bienstock, Brickell and Monma (1990) showed that (4.3) holds for arbitrary k. They also
showed that k-edge can be replaced by k-node throughout (4.3) with the technical
restriction that |V | 2k in order for (a) to hold. This technical restriction is necessary since
without it there is an innite family of examples where condition (a) fails.
We note that the cost of an optimal k-edge connected solution may be strictly less than the
cost of an optimal k-node connected solution for any k 3, and that the conditions in (4.3)

17
do not characterize the optimal solutions as they do in the case k = 2. The proof of (4.3) for
the k-edge connected case uses the Lifting Theorem (4.1). The proof of (4.3) in the k-node
connected case is much more dicult and requires the use of pairs of liftings as well as a
number of further technical results.

4.2 Construction and Improvement Heuristics

We now describe some heuristics for constructing feasible networks and heuristics for improv-
ing the cost of a feasible solution for the kECON and kNCON problems. This is, to a large
extent, a summary of the work of Monma and Shallcross (1989) for the low-connected
survivable network design problem, that is, problems with node types in {0, 1, 2}. The per-
formance on real-world problems is described in Section 7. These heuristics were inspired by
the wide variety of heuristics for the traveling salesman problem and other combinatorial op-
timization problems and were designed to work on sparse underlying graphs. The heuristics
are used in a local search approach to obtain low-cost network designs; see Papadimitriou
and Steiglitz (1982) for a general discussion of local search procedures for combinatorial
optimization problems. It is obvious how these heuristics can be generalized and applied
to the general survivable network design problem, and so we just briey mention one such
extension.
One useful structural fact is that any edge-minimal two-connected graph G = (V, E) can be
constructed by an ear decomposition procedure; see Lovasz and Plummer (1986). That
is, rst nd a cycle C in G. Then repeatedly nd a path P , called an ear, that starts at a
node v in the solution, passes through nodes not yet in the solution, and ends up at a node
w in the solution. All edge-minimal two-edge connected graphs can be constructed in this
manner. If the nodes v and w are required to be distinct, then every edge-minimal two-node
connected graph can be constructed in this manner.
The ear decomposition approach can be employed to construct a feasible 2-connected sub-
graph of a graph G = (V, E) using costs to add ears in a greedy fashion. We call this the
greedy ears construction heuristic. The rst step is to construct a partial solution
consisting of a cycle C spanning the set of nodes of type 2, called the special nodes. This
is done by randomly selecting a special node v, and then selecting a special node w whose
shortest path P to v is the longest among all special nodes. Let node u be the node next to
w on the path P . We will construct a short cycle through the edge uw by nding a shortest
path from u to w not using the edge uw. (There must be such a path; if not, there would
not two node-disjoint paths between the special nodes v and w and so the problem would be
infeasible). The next step is to repeatedly add short ears to the current partial solution
until all special nodes are on this two-connected network. This is done by rst selecting

18
a special node z, not yet in the solution, whose shortest path P to the partial solution is
longest among all special nodes not yet included. We will nd another shortest path Q from
z to the partial solution that does not use any edges of P and that terminates on the partial
solution at a node w other than v. (Again, such a path must exist for the problem to be
feasible). The combination of paths P and Q must contain an ear which is added to the
partial solution.
A second construction heuristic uses the ear decomposition approach to construct a feasible
2-connected subgraph of a graph G = (V, E) in a random fashion. We call this the random
sparse construction heuristic. The rst step is to construct an initial random cycle C
spanning a subset of the special nodes. This is done by randomly choosing a special node v,
and constructing a depth-rst-search tree T rooted at v. Form a cycle by randomly choosing
an edge of the form vz that is not in T . (There must be such an edge or else v is not on
any cycle and so the problem is infeasible.) Next, random ears are repeatedly added until
all special nodes are on the two-connected network. This is done by rst constructing a
depth-rst-search forest F rooted at the nodes that are in the partial solution. A node v is
said to be allowed if v is not yet in the solution, but it has an edge vw in E but not F , where
w is in the solution. Randomly choose a node v from among the allowed nodes. (Again,
there must be such a node or the problem is infeasible.) Let T be the tree in the forest F
containing v, and let z be the root of T . The random ear chosen is the path from v to z in T ,
together with the edge vw. Since this method does not use cost information, generally it
does not produce a low cost solution. However, this method is useful for generating starting
random initial solutions on which to apply the improvement methods.
Improvement heuristics apply local transformations to any feasible network in order to reduce
the cost while preserving feasibility. These transformations are applied until a locally optimal
network is obtained; that is, no further improvements of these types are possible. Six local
transformations are described in the sequel. These transformations are general enough to
cover a wide range of feasible topologies, yet fast enough to be performed in real time, even
on a personal computer.
Every two-connected network contains at least one cycle, and often is made up of many
interconnected cycles. Furthermore, replacing the edges in a cycle C by edges forming a
cycle C  on the same nodes preserves the feasibility of a solution. So it is natural to draw
upon the extensive research on the traveling salesman problem (see Lawler et al. (1985)) for
nding a near-optimal cycle. This is the basis of the two-optimal cycle and three-optimal
cycle improvement heuristics. The pretzel, quetzel and degree improvement heuristics alter
the structure of the two-connected part of the solution in less obvious ways. The one-
optimal improvement heuristic alters the structure of the entire solution. These heuristics
are described below.

19
The two-optimal interchange heuristic attempts to replace two edges of a cycle C by
two edges not in the cycle to form a new cycle C  of lower cost. Similarly, the three-optimal
interchange heuristic attempts to replace three edges of a cycle C by three edges not in
the cycle to form a new cycle C  of lower cost.
These improvement heuristics replace one cycle C by another cycle C  on the same nodes,
and so do not change the fundamental structure underlying the solution. The pretzel
transformation replaces an edge uv of a cycle C by two crossing edges ux and vy to form
a pretzel P , where the nodes u, v, x and y appear in order on the cycle.
The quetzel transformation is the reverse of the pretzel transformation; that is, a pretzel P
is replaced by a cycle C by removing two crossing edges ux and vy and adding an edge uv.
As mentioned before, if the costs satisfy the triangle inequality, then there is an optimal
two-connected solution where all nodes are of degree two or three. The proof of this result
employs the Lifting Theorems (4.1) and (4.2). We will describe now the algorithmic use of
these theorems in the form of degree improvement transformations.
Let node u be of degree four or more, and let nodes a, b, c, and d be four of its neighbors.
There are three cases to consider. In Case a, there are node-disjoint paths Pab and Pbc from
a to b, and from b to c, respectively, which miss node u. In this case, edge ub is a chord and
can be deleted. So we may assume that no three of the nodes a, b, c and d have such paths.
Therefore, the paths Pab and Pbc must intersect in node v and the paths Pbc and Pcd must
intersect in node w. If nodes v and w are dierent, we are in Case b, and edges bu and uc
are removed, and edge bc is added; this preserves the connectivity requirements and does
not increase the cost if the triangle inequality holds. If nodes v and w are the same node,
we are in Case c. Let a, b, c and d be the neighbors of v(= w). Edges au, ub, bv and vc
are removed, and edges ab and bc are added; this preserves the connectivity requirements
and does not increase the cost if the triangle inequality holds. In all cases, the degree of
node u is decreased, and no other degrees are increased; so repeated application of these
transformations guarantees that an optimal solution where all degrees are two or three will
be obtained so long as the triangle inequality holds.
All of the previous improvement heuristics operated only on the two-connected part of the
solution. The one-optimal interchange heuristic considers the entire solution. This
heuristic attempts to remove an edge uv from the current feasible solution and replace it
with another edge of the from ux not currently in the solution. Such an interchange is made
only if the resultant network is feasible and of lower cost.
These heuristics were tested on randomly generated problems and on real-world problems of
telephone network design; see Monma and Shallcross (1989) for details. They could decrease
the cost of manually constructed solutions by about 10 percent in a test of these methods on

20
a real-world application. Comparison with the optimal solutions computed by the cutting
plane algorithm (to be described in Section 7) on the same examples show that the gap was
usually very small, about 1 percent of the optimal value. The highest running time for a
sparse 116-node problem was 156 seconds on an IBM-PC/AT.
Ko and Monma (1989) modied the low-connectivity heuristics of Monma and Shallcross to
the design of k-edge or k-node connected networks. A rst feasible solution is constructed
either by deleting edges successively from the whole graph while maintaining feasibility, or
by adding successively k edge-disjoint [1, j]-paths of minimum overall length, for all nodes
j = 1. The output is necessarily k-edge connected. The local transformations for the low-
connectivity case carry over to the high-connectivity case, except that the feasibility checks
have to be implemented dierently.
These heuristics could only be tested on random examples, as real-world examples were not
yet available. When the cost of the best heuristic solution was compared with the optimal
solution produced by our cutting plane algorithm, the gap was approximately 6 percent.
Running times on a VAX 8650 ranged between 13 seconds for dense graphs of 20 nodes and
k = 3 and 120 seconds for dense graphs of 40 nodes and k = 5.
Let us remark that to our knowledge the rst heuristics for the design of minimum-cost
survivable networks under general connectivity requirements date back to Steiglitz, Weiner
and Kleitman (1969). Their heuristic consists of a randomized starting routine and an
optimization routine where local transformations are applied to a feasible solution as follows.
Given a random ordering of the nodes, the starting routine adds edges between the rst
node with the highest connectivity requirement and the rst node with the next highest
connectivity requirement. In each step the connectivity requirements are updated. If the
solution is feasible, the optimizing routine tries to improve this solution by successively
replacing one pair of edges with another pair of edges to obtain another feasible solution of
lower cost until no more improvements can be made this way.
They applied their heuristics to two real-world problems with 10 nodes and 58 nodes, and
connectivity requirements in {3, 4, 5, 6} and {6}, respectively. (Unfortunately, the data are
not available any more.) The 58-node problem took about 12 minutes (on a UNIVAC 1108)
per local optimum. Since no lower bounds on the optimal value for these problems are given,
we cannot say how well these heuristics work.

4.3 Heuristics with Performance Guarantees

The last remark leads us to an important issue: quality guarantees, i.e., worst-case perfor-
mance and lower bounds. The polyhedral approach, to be described later, can be viewed

21
as a technique to obtain very good lower bounds for the optimum value of a kECON or
kNCON problem. But sometimes nice performance guarantees for heuristics or estimates
for optimum values can be given with less elaborate techniques.
Let us rst relate the 2ECON and 2NCON problems to the traveling salesman problem.
Since every Hamiltonian cycle is 2-node (and thus 2-edge) connected, the optimum TSP-
value, CTSP say, is not smaller than the optimum value COPT of the 2ECON problem. On
the other hand, using (4.3), Monma, Munson and Pulleyblank (1990) were able to show that
if the triangle inequality holds, COPT can be bounded by CTSP from below by a constant
factor, more precisely

3
CTSP COPT CTSP.
4

To solve the 2ECON problem approximately, Frederickson and JaJa (1982) modied the
Christodes heuristic for the traveling salesman problem and proved that, when the triangle
inequality holds, the solution attains a cost CCHR with the same worst-case bound as in
the traveling salesman problem, namely
3
CCHR COPT.
2

Another type of lower bound for the k-edge connected network design problem can be derived
from the subtour elimination polytope, which is a natural linear programming relaxation of
the traveling salesman polytope. Let CSUB denote the value of an optimal solution to the
subtour elimination linear program; see Chapter 3. Goemans and Bertsimas (1990) showed
that
k/2 CSUB COPT.

For k = 2, this was previously shown by Cunningham, see Monma, Munson and Pulleyblank
(1990). These results make use of the Lifting Theorem (4.2).
We conclude this section by describing heuristics by Goemans and Bertsimas (1990) and
Agrawal et al. (1991) with worst-case guarantees for the kECON problem. Goemans and
Bertsimas developed a tree heuristic with worst-case guarantee for a version of kECON
problem with general node types r ZV+ , where edges may be used more than once. This
means that a feasible solution to this problem is a vector x Z E + of nonnegative integers
that satises all cut inequalities of x((W )) r(W ), but not necessarily the upper bounds
xe 1. A component xe 2 may be interpreted as edge e used xe times.
Let 1 < 2 < . . . < p be the ordering of the distinct node types in r, let 0 := 0, and let
Vk be the set of nodes of type at least k , k = 1, . . . , p.

22
(4.4) Tree Heuristic

1. Compute, for all pairs of nodes u, v, the shortest path length cuv with respect to the
given costs c.
2. Set xe := 0 for all e V V .
3. For k = 1 to p do
Compute Tk = (Vk , Ek ) as the minimum spanning tree of the complete graph
induced by Vk with respect to costs ce .
Set xe := xe + (k k1 ) for all e Ek .
4. Replace each uv V V with xuv > 0 by xuv times the shortest [u, v]-path with value
cuv .

Note that the output of this heuristic is a vector x ZE + that satises all cut inequalities
x((W )) r(W ). Let y be the solution to the LP-relaxation of the ECON problem consisting
of cut inequalities (2.4)(i) and nonnegativity constraints xe 0. Goemans and Bertsimas
show that
  p 
xT c 2  k k1
2 .
T
y c |V1 | k=1 k

This bound is tight, if we consider for instance a 2ECON problem on a cycle of costs 1.
Goemans and Bertsimas also describe a more rened tree heuristic with a better worst-case
guarantee.
Agrawal, Klein, and Ravi (1991) state a heuristic for an ECON problem, where edge-
connectivity requirements are given by r ZV+ V and the use of multiple parallel edges
is allowed in the solution. They prove that their algorithm outputs a solution whose cost is
approximately within 2 log R of the optimal, where R is the highest requirement value.
The worst-case guarantees for the heuristics of Goemans and Bertsimas and of Agrawal et al.
were found by reduction to an ECON problem with costs satisfying the triangle inequality
(see Step 1 and 4 of the tree heuristic). This reduction does not work, if the use of parallel
edges is forbidden in the solution.
Recently, Khuller and Vishkin (1992) gave a heuristic for nding a minimum-cost k-edge
connected subgraph of a graph (where parallel edges do not appear in the solution). This
heuristic has a worst-case guarantee of 2, even when the costs do not satisfy the triangle
inequality. Worst-case guarantees for heuristics for the NCON problem are not known.

23
5 Polynomially Solvable Special Cases
We have already remarked that the general problem of designing survivable networks is NP-
hard; in fact, quite a number of special cases are known to be NP-hard. The intention of
this section is to give an overview of those special cases where polynomial time solution
procedures are known.
There are basically three ways to restrict the general model. One either considers special
choices of node types, special cost functions or special classes of graphs. It turns out that
some combinations of these restrictions lead to easy (but still interesting) problems.

5.1 Node Type Restrictions

Let us start by considering restrictions on the node types. G = (V, E) may be any graph
with costs ce R for all e E.
If rv = 1 for all v V and ce > 0 for all e E, the 1NCON and 1ECON problems for G
and r are equivalent to nding a minimum cost spanning tree. This problem is well known
to be solvable in polynomial time; see Chapter 12. If the costs are not necessarily positive,
we look for a minimum cost connected subgraph. This can be found by rst choosing all
edges of nonpositive cost, shrinking all resulting components to single nodes, and computing
a minimum spanning tree in the resulting graph.
If two nodes have type 1, say nodes u and v, all other nodes have type 0, and if all costs
are positive, then the ECON (and NCON) problem for G and r is nothing but the problem
of computing a shortest [u, v]-path; see Chapter 1 for polynomial time methods for that
problem. If costs are general, we seek an edge set of minimum cost that contains a path
from u to v. This can be solved in polynomial time by shrinking away the components
induced by the nonpositive edges and then computing a shortest [u, v]-path in the resulting
graph.
The slightly more general case, when rv {0, 1} for all v V , is the Steiner tree problem,
which is NP-hard even if ce = 1 for all e E. If there is a xed number of nodes of type
0 or a xed number of type 1, then the Steiner tree problem is polynomially solvable, see
Lawler (1976).
The shortest path problem has an extension that is solvable in polynomial time. Namely, if
two nodes have type k, say ru = rv = k, all others type 0, and if all costs are positive, then
the kNCON problem asks for a collection of k node disjoint [u, v]-paths that are of minimum
total cost and, similarly, the kECON problem requires nding a minimum collection of

24
k edge disjoint [u, v]-paths. Both problems can be solved with min-cost ow algorithms in
polynomial time; see Chapter 1. As above, it is trivial to extend this to the case where costs
are arbitrary.
We do not know of any other (nontrivial) case where special choices of node types lead to
polynomial time solvability.

5.2 Cost Restrictions

The cases we want to mention here are two well-known problems in graph theory with
uniform edge costs. First, given some nodes with node types, design a graph that satises
all connectivity requirements and has a minimum number of edges; and second, given a graph
and node types, add a minimum number of edges to the graph such that the resulting graph
meets the connectivity requirements. Node and edge connectivity have to be distinguished
here and whether or not parallel edges may be added.
Consider the following problem:
(5.1) Problem. Given a set V of nodes, and a requirement rst 0, for each pair s, t V ,
nd a graph G = (V, E) such that each pair s, t is at least rst -edge connected in G and such
that |E| is as small as possible.
Chou and H. Frank (1970) gave a polynomial-time algorithm to solve this problem if G
may contain parallel edges. H. Frank and Chou (1970) also solved this problem when no
parallel edges are allowed, but additional nodes are allowed in the construction. If neither
parallel edges nor further nodes are allowed, it is not known whether one can solve (5.1) in
polynomial time.
The algorithm of Chou and Frank (1970) solves the ECON problem for K n (with parallel
edges), arbitrary edge connectivity requirements rst 2, and uniform edge costs.
The node connectivity problem analogous to (5.1) is open. We are only aware of a result
of Harary (1962) who proved that, given n and k, the minimum number of edges in a
k-connected graph on n nodes (without parallel edges) is
kn/2. Such a graph can be
constructed easily.
To our knowledge these are the only solved cases with uniform edge costs.
The following problem often runs under the name augmentation problem in the graph
theory literature.

25
(5.2) Problem. Given a graph, augment it by a minimum number of edges so that the new
graph meets certain connectivity requirements.
In our terminology, (5.2) can be viewed as an ECON or NCON problem where all edges in
the given graph have cost 0, and all edges that are allowed to be added have cost 1.
This type of problem was solved by Eswaran and Tarjan (1976) for 2-edge connected graphs,
and by Rosenthal and Goldner (1977) for 2-node connected graphs. The k-edge connected
graph augmentation problem was studied by Watanabe and Nakamura (1987), Ueno et al.
(1988), Cai and Sun (1989), and Naor et al. (1990). Recently, A. Frank (1990) solved the
general edge connectivity case. All solutions are algorithmic and require polynomial time.
A. Frank (1990) proved a nice min-max result for the minimum number of edges needed to
augment a given graph G to satisfy given edge connectivity requirements r ij . Let us dene
the deficit def(A) of a node set A as

def(A) := min ruv |G (A)|.

uA,vA

The decit of A is the smallest number of edges that have to be added to G (A) in order to
connect A suciently to all other nodes. Clearly, if several disjoint sets Ai (i = 1, . . . , t) have
decit def(Ai ), then a lower bound on the number of edges to be added is

1 t
def(Ai ).
2 i=1

Frank (1990) shows that, under certain assumptions, the best such lower bound is exactly
the minimum number of edges needed in an augmentation.
(5.3) Theorem. Given a graph G = (V, E) and edge connectivity requirements rst for all
pairs of nodes s, t V . Then the following holds:
If all sets A V have decit larger than 1, then the minimum number of edges to be added
to G to satisfy the edge connectivity requirements, is

1 t
max def(Ai).
A1 ,...,At disjoint 2 i=1

If some set A has decit at most one and all proper subsets of A have nonpositive decit,
then the minimum number of edges to be added to G is def(A) plus the minimum number of
edges to be added to G A. 2

26
Franks proof is constructive and results in polynomial-time algorithm to create a minimum
cost augmentation. In the same paper, Frank proved similar results for directed graphs.
But he did not consider augmentations to node-connected graphs, nor does he consider
augmentations where the use of parallel edges is forbidden.

5.3 Special Classes of Graphs

It is often the case that NP-hard problems become easy when restricted to graphs with
special structural properties. In the case of the ECON and NCON problems, there are only
very few (and relatively simple) classes of graphs known where some versions can be solved
in polynomial time. Most notable is the class of series-parallel graphs.
Series parallel graphs are created from a single edge by two operations:

addition of parallel edges, and

subdivision of edges.

(For our purposes we note that all series-parallel graphs, except K2 , are 2-connected.) The
Steiner tree problem can be solved in linear time on this class of graphs by a recursive
algorithm. This was mentioned by Takamizawa, Nishizeki and Saito (1982), and stated
explicitly by Wald and Colbourn (1983). By a modication of this recursive algorithm, the
2NCON problem can also be solved, where node types 0, 1, and 2 are allowed.
Winter has developed linear-time algorithms for 2ECON and 2NCON problems with node
types 0 and 2 on outerplanar, series-parallel, and Halin graphs, see Winter (1985a,b) and
Winter (1986). In his survey article, Winter (1987) mentioned that he also found linear-time
algorithms for Halin graphs that solve the 3ECON and 3NCON problems with node types
0 and 3.
Mahjoub (1988) has found a complete characterization of the polytope of 2-edge connected
subgraphs of series-parallel graphs.
Complete characterizations of Steiner tree and related polytopes in extended spaces can be
found in Prodon, Liebling, and Groin (1985) and Goemans (1991a, 1991b). Cornuejols,
Fonlupt, and Naddef (1985) found a complete description of the dominant of the 2-edge
connected subgraph polytope of series parallel graphs, and Chopra (1991) investigated, for
odd k, the k-edge connected subgraphs of a given outerplanar graph, and found a complete
description of the dominant of the associated polyhedron by the so-called lifted outerplanar
partition inequalities. (Outerplanar graphs are those graphs that can be drawn in the

27
plane as a cycle with non-crossing chords. They form a subclass of the class of series-parallel
graphs.) Chopras result is a follows:
(5.4) Theorem. For outerplanar graphs G = (V, E) and uniform node types r {k}V ,
k odd, the dominant of the kECON(G; r) polytope is completely characterized by the in-
equalities:
p

1
2
x((Wi)) p
k/2 1 for all partitions {W1 , . . . , Wp } of Vj
i=1
xe 0 for all e in E. 2
Note, however, that these inequalities are not generally valid for kECON(G; r) if the un-
derlying graph is, say, complete!

6 Polyhedral Results
Except for the results of Grotschel & Monma (1990) mentioned in Section 3, there is not
much known about the polytope CON(G; r, k, d) for general edge and node survivability
requirements r, k and d. We will thus concentrate on the kNCON and kECON problems
that have been investigated in more depth and survey some of the known results. Par-
ticular attention has been paid to the low-connectivity case, that is, where r {0, 1, 2}E .
See Grotschel & Padberg (1985) and Pulleyblank (1990) for a general survey of polyhedral
combinatorics and the basics of polyhedral theory.
Let us mention again the idea behind this approach and its goal. We consider an integer
programming problem like (3.1) or (3.10). We want to turn such an integer program into a
linear program and solve it using the (quite advanced) techniques of this area. To do this,
we dene a polytope associated with the problem by taking the convex hull of the feasible
(integral) solutions of a program like (3.1) or (3.10). Let P be such a convex hull. We
know from linear programming theory that, for any objective function c, the linear program
min{cT x | x P } has an optimum vertex solution (if it has a solution). This vertex solution
is, by denition, a feasible solution of the initial integer program and thus, by construction,
an optimum solution of this program.
The diculty with this approach is that max cT x, x P is a linear program only in prin-
ciple. To provide an instance to an LP-solver, we have to nd a dierent description of P .
The polytope P is dened as the convex hull of (usually many) points in RE , but we need a
complete (linear) descriptions of P by means of linear equations or inequalities. The Weyl-
Minkowski theorem tells us that both descriptions are in a sense equivalent, in fact, there are
constructive procedures that compute one description of P from the other. However, these

28
procedures are inherently exponential and nobody knows how to make eective use of them,
in particular, for NP-hard problem classes. Moreover, there are results in complexity theory,
see Papadimitriou and Yannakakis (1982), that indicate that it might be much harder to
nd a complete linear description of such a polytope P than to solve min c T x, x P .
At present, no eective general techniques are known for nding complete or good partial
descriptions of such a polytope or large classes of facets. There are a few basic techniques
like the derivation of so-called Chvatal cuts (see Chvatal (1973)). But most of the work is
a kind of art. Valid inequalities are derived from structural insights and the proofs that
many of these inequalities dene facets use technically complicated, ad-hoc arguments.
If large classes of facet-dening inequalities are found, one has to think about their algorith-
mic use. The standard technique is to employ such inequalities in the framework of a cutting
plane algorithm. We will explain this in Section 7. It has turned out in the recent years
that such eorts seem worthwhile. If one wants to nd true optimum solutions or extremely
good lower bounds, the methods of polyhedral combinatorics are the route to take.

6.1 Classes of Valid Inequalities

We will now give an overview of some of the results of Grotschel, Monma and Stoer (1992a,
1992b, 1992c) and Stoer (1992) concerning classes of valid inequalities for the kECON and
kNCON problems. We will motivate these inequalities and mention how they arise.
As before, we consider a loopless graph G = (V, E), and in the kECON case possibly with
multiple edges. We assume that for each node v V a nonnegative integer r v , its node type,
is given, that k = max{rv | v V } and that at least two nodes are of type k. Recall that
r(W ) = max{rv | v W } is called the node type of W .
We start out by repeating those classes we have already introduced in Section 3. Clearly,
the trivial inequalities

(6.1) 0 xe 1 for all e E

are valid for kECON(G; r) and kNCON(G; r) since problem (3.10) is a 0/1-optimization
problem. The cut inequalities

(6.2) x((W )) con(W ) for all W V, = W = V,

where con(W ) = min{r(W ), r(V \W )}, are valid for kECON(G; r) and kNCON(G; r) since
the survivable network to be designed has to contain at least con(W ) edge disjoint paths

29
that connect nodes in W to nodes in V \ W . (Recall that rst = min{rs , rt }, s, t V .) In
the node connectivity case the requirement that if, for a given pair s, t of nodes, at most
rst 1 nodes dierent from s and t are deleted from G then there has to be at least one
more path connecting s and t in the remaining graph. This requirement leads to the node
cut inequalities.

(6.3) x(GZ (W )) con(W ) |Z| for all pairs s, t V, s = t and

for all = Z V \ {s, t} with |Z| rst 1 and
for all W V \ Z with s W, t  W.

These inequalities are valid for kNCON(G; r) but of course not for kECON(G; r).
How does one nd further classes of valid inequalities? One approach is to infer inequal-
ities from structural investigations. For instance, the cut inequalities ensure that every
cut separating two nodes contains at least rst edges. These correspond to partitioning the
node set into two parts and guaranteeing that there are enough edges linking them. We
can generalize this idea as follows. Let us call a system W1 , . . . , Wp of subsets of V with
Wi Wj = for 1 i < j p, Wi = for i = 1, . . . , p and W1 . . . Wp = V a partition
of V and let us call

(W1, . . . , Wp) := {uv E | i, j, 1 i, j p, i = j with u Wi , v Wj }

a multicut or p-cut (if we want to specify the number p of shores W 1 , . . . , Wp of the

multicut). Depending on the types r(W1 ), . . . , r(Wp ), any survivable network (V, F ) will
have to contain at least a certain number of edges of the multicut (W 1 , . . . , Wp ). For every
partition it is possible to compute a lower bound of this number, and thus to derive a valid
inequality for every node partition (resp. multicut). This goes as follows.
Suppose W1, . . . , Wp is a partition of V such that r(Wi ) 1 for i = 1, . . . , p. Let I1 := {i
{1, . . . p} | r(Wi ) = 1}, I2 := {i {1, . . . , p} | r(Wi ) 2}. Then the partition inequality
(or multicut inequality) induced by W1, . . . Wp is dened as

1 p
12 r(Wi ) + |I1| if I2 =

(6.4) x((W1, . . . , Wp)) = x((Wi)) iI2
2 i=1
p1 if I2 =

Every partition inequality is valid for kECON(G; r) and thus for kNCON(G; r).
Structural insight of this type is often hard to get without any hint. We mention a
second approach that frequently helps to get a fresh view. One solves a number of linear
programs using various objective functions and all inequalities and equations which are
currently known. In some cases, fractional optimum solution will result. The question is

30
now to invent new inequalities that cut these fractional solutions o but do not cut o any
integral solution. If one can nd such inequalities then the research challenge is to produce
a large class of more general valid inequalities that contain the initial ones as special cases.
The hint towards establishing new types of valid inequalities is given by the structure of
the fractional optimum solutions. Again there is no general technique known for eectively
using this information. Some creativity is necessary.
For example, consider a complete graph K 3 = (V, E) on the three nodes V = {a, b, c}, the
node types are rv = 1 for all v V . The polyhedron dened by the trivial inequalities (6.1)
and the cut inequalities (6.2) has the fractional vertex xe = 12 for all e E. This fractional
solution does not satisfy the partition inequality x(({a}, {b}, {c})) = xab + xac + xbc 2.
Just as the cut inequalities (6.2) can be generalized as outlined above to partition inequalities
(6.4), the node cut inequalities (6.3) can be generalized to a class of inequalities that we will
call node partition inequalities, as follows.
Let Z V be some node set with |Z| 1. If we delete Z from G then the resulting graph
must contain an [s, t]-path for every pair of nodes s, t of type larger than |Z|. In other words,
if W1 , . . . , Wp is a partition of V \ Z into node sets with r(Wi ) |Z| + 1 then the graph
G := (GZ)/W1 / . . . /Wp obtained by deleting Z and contracting W1 , W2, . . . , and Wp must
be connected. This observation gives the following class of node partition inequalities
valid for kNCON(G; r), but not for kECON(G; r):
1 p
(6.5) 2 i=1 x(GZ (Wi )) p 1 for every node set Z V, |Z| 1 and
every partition W1 , . . . , Wp of V \ Z
such that r(Wi ) |Z| + 1, i = 1, . . . , p.

If r(Wi ) |Z| + 2 for at least two node sets in the partition, then the right-hand side of
the node partition inequality can be increased. This leads to further generalizations of the
classes (6.4) and (6.5), but their description is quite technical and complicated, see Stoer
(1992). So we do not discuss them here.
We now mention another approach (that can be viewed as a special case of our rst one) to
nding new classes of valid inequalities.
The idea here is to relax the problem in question combinatorially by nding a (hopefully
easier) further combinatorial optimization problem such that every solution of the given
problem is feasible for the new problem and by studying the polytope associated with the
new combinatorial optimization problem. If the relaxation is carefully chosen (and one is
lucky) some valid inequalities for the relaxed polytope turn out to be facet-dening for the
polytope one wants to consider. (These inequalities are trivially valid.)
In our case, a relaxation that is self-suggesting is to consider the r-cover problem. This ties

31
the survivability problem to matching theory and, in fact, one can make good use of the
results of this theory for the survivability problem.
The survivability requirements imply that if v V is a node of type r v , then v has degree
at least rv for any feasible solution of the kECON problem. Thus, if we can nd an edge
set such that each node has degree at least rv (we call such a set an r-cover) and that
has minimum cost we obtain a lower bound for the optimum value of the kECON problem.
Clearly, such an edge set can be found by solving the integer linear program
(6.6) min cT x
(i) x((v)) rv for all v V,
(ii) 0 xe 1 for all e E, and
(iii) xe integer for all e E,
which is obtained from (3.1) (i), (iii), (iv) by considering only sets of cardinality one in
(3.1) (i). The inequalities (i) are called degree constraints. This integer program can be
turned into a linear program, i.e., the integrality constraints (iii) are replaced by a system of
linear inequalities, using Edmonds polyhedral results on b-matching, see Edmonds (1965).
Edmonds proved that, for any vector b Z+V , the vertices of the polyhedron dened by

(6.7) (i) y(E(H)) + y(T ) (bv + |T |)/2 for all W V and all T (H), and
vH
(ii) 0 ye 1 for all e E
are precisely the incidence vectors of all (1-capacitated) b-matchings of G, i.e., of edge
sets M such that no node v V is contained in more than bv edges of M. For the case
bv := deg(v) rv , where deg(v) denotes the degree of v in G, the b-matchings M are nothing
but the complements M = E \ F of r-covers F of G. Using the transformation x := 1 y
and T := (H) \ T we obtain the system

(6.8) (i) x(E(H)) + x((H) \ T ) (rv |T |)/2 for all H V and all T (H), and
vH
(ii) 0 xe 1 for all e E.

(6.8) gives a complete description of the convex hull of the incidence vectors of all r-covers
of G. We call the inequalities (6.8)(i) r-cover inequalities. Since every solution of the
kECON problem for G and r is an r-cover, all inequalities (6.8)(i) are valid for kECON(G; r).

It is a trivial matter to observe that those inequalities (6.8)(i) where vH rv |T | is even
are redundant. For the case rv = 2 for all v V , Mahjoub (1988) described the class of
r-cover inequalities, which he calls odd wheel inequalities.
Based on these observations one can extend inequalities (6.8)(i) to more general classes of
inequalities valid for kECON(G; r) (but possibly not valid for the r-cover polytope). We
present here one such generalization.

32
Let H be a subset of V called the handle, and T (H) with |T | odd and |T | 3.
For each e T , let Te denote the set of the two endnodes of e. The sets Te , e T , are
called teeth. Let H1 , . . . , Hp be a partition of H into nonempty disjoint subsets such that
r(Hi ) 1 for i = 1, . . . , p and |Hi Te | r(Hi ) 1 for all i {1, . . . , p} and all e T . Let
I1 := {i {1, . . . , p} | r(Hi ) = 1} and I2 = {i {1, . . . , p} | r(Hi ) 2}. We call
p

(6.9) x(E(H)) x(E(Hi )) + x((H) \ T )
i=1


(r(Hi ) |T |)/2 + |I1|

iI2

the lifted r-cover inequality (induced by H1 , . . . , Hp , T ). All inequalities of type (6.9) are
valid for ECON(G; r).
The names handle and teeth used above derive from the observation that there is some
relationship of these types of inequalities with the 2-matching, comb and clique tree inequal-
ities for the symmetric traveling salesman polytope; see Chapter 3. In fact, there is a further
generalization of the r-cover inequalities that makes this resemblance (in form and name)
more apparent. This generalization, however, applies only to the case where rv {0, 1, 2}
for all v V .
Let H, T1 , . . . , Tt be subsets of V (H is the handle, the sets T1 , . . . , Tt are the teeth) that
satisfy the following conditions. The number t of teeth is at least 3 and odd. Two teeth have
at most one node in common. Each tooth T i intersects the handle H in exactly one node; we
denote this node by ti for i = 1, . . . , t. In each tooth Ti we choose a node zi Ti \H and call zi
a special node. The choice is restricted if Ti Tj = ; in this case Ti Tj = {zi } = {zj } (i.e.,
the nodes zi are not necessarily distinct). Moreover, we assume that rti = 2 for i = 1, . . . , t

and rv 1 for all v H ( ti=1 (Ti \ {zi })). Under these assumptions one can show that the
following comb inequality


t 
t
(6.10) x(E(H)) + x((H)) + x(E(Ti)) + x([Ti \ (H {zi }) : V \ Ti ])
i=1 i=1


t 
t 
t
t
x([{ti} : Ti ]) x([{zi} : Ti V2 ]) |H| + (|Ti | 2) 
i=1 i=1 i=1 2

is valid for 2NCON(G; r); it is not valid for 2ECON(G; r). Again there are further general-
izations that we do not want to discuss here.
Another technique for nding further classes of valid and facet-dening inequalities will be
mentioned in Section 6.4.

33
6.2 Facet Results
To develop successful cutting plane algorithms, it is not enough to know some inequalities
valid for the polytope over which one wants to optimize. The classes of inequalities should
contain large numbers of facets of the polytope. Ideally, one would like to use classes of
facet-dening inequalities only.
In our case, it turned out to be extremely complicated to give (checkable) necessary and
sucient conditions for an inequality in one of the classes described above to dene a facet of
NCON(G; r) or ECON(G; r). Lots of technicalities creep in. Nevertheless, it could be shown
that large subsets of these classes are facet-dening. These results provide a theoretical
justication for the use of these inequalities in a cutting plane algorithm.
We will not go deeply into the results that have been obtained so far; see Grotschel and
Monma (1990), Grotschel, Monma and Stoer (1992a,1992b,1992c) and Stoer (1992) for de-
tails. We just provide a glimpse at the subject here.
We restrict ourselves to the cut inequalities (6.2). We rst concentrate on the low-connectiv-
ity case, where rv {0, 1, 2} for all v V . To state the results we have to introduce further
terminology.
Given a graph G = (V, E), a node type vector r ZV+ , and a node set W V with |W | 2,
we set
Vi := {v V |rv i} for i = 0, 1, 2,
(G, W ) := minimum cardinality of a subset of E whose removal from G
disconnects two nodes of W , and
(G, W ) := minimum cardinality of a set S V E whose removal from
G disconnects two nodes of W in G (recall that we assume G
to be simple in this case).
If |W | < 2 then (G, W ) and (G, W ) are dened to be .
We assume throughout this section that G is two-node connected and satises (G, V2 )
3 when we deal with the 2ECON case, and that G is two-node connected and satises
(G, V2 ) 3 when we deal with the 2NCON case. If this is not so then the problem can
be decomposed (in polynomial time) into independent smaller problems that are trivially
solvable or satisfy these conditions. The exact decomposition procedure involves many tech-
nical details and is described in Grotschel, Monma and Stoer (1992b). A side benet of this
assumption is that the polyhedra 2ECON(G; r) and 2NCON(G; r) are full dimensional (see
Grotschel, Monma (1990)), i.e., they contain |E| + 1 anely independent vectors.
The next theorem describes necessary conditions for the cut inequalities (6.2) to dene facets.

34
(6.11) Theorem.Let G = (V, E) be a graph, r {0, 1, 2}V and W V with = W = V .

(i) If r(W ) = 2 then x((W )) 2 denes a facet of 2ECON(G; r) only if (G[W ], V1 ) 2,

(G[V \ W ], V1) 2, and G[W ] and G[V \ W ] are connected.
(ii) If r(W ) = 1 then x((W )) 1 denes a facet of 2ECON(G; r) if and only if G[W ] and
G[V \ W ] are connected, (G[W ], Vi ) i + 1 for i = 1, 2, and (G[V \ W ], V2) 3.
(iii) If r(W ) = 0 then x((W )) 0 does not dene a facet of 2ECON(G; r) or 2NCON
(G; r).
(iv) If r(W ) = 2 then x((W )) 2 denes a facet of 2NCON(G; r) only if all conditions
of (i) are satised and (G[W ], V2 ) 2 and (G[V \ W ], V2) 2.
(v) If r(W ) = 1 then x((W )) 1 denes a facet of 2NCON(G; r) only if all conditions of
(ii) are satised and (G[V \ W ] e, V2) 2 for all e E(V \ W ). 2

If we look at the uniform case where rv = 2 for all v V , then in a 3-edge connected graph a
cut inequality x((W )) 2 denes a facet of 2ECON(G; r) if and only if G[W ] and G[V \ W ]
are both 2-edge connected. This result was independently obtained in Mahjoub (1988).
Results (i), (iv) and (v) above can be turned into if and only if statements by adding
further (complicated) technical conditions, see Grotschel, Monma and Stoer (1992a). Things
get even more complicated if we want to characterize those node cut inequalities (6.3) that
are facet-dening for 2NCON(G; r); see Stoer (1992). For partition, node partition, r-cover,
lifted r-cover and comb inequalities to dene a facet only sucient conditions are known (in
the low- as well as in the high- connectivity cases).
If r is a vector of arbitrary nonnegative integers, even more complications arise due to the
fact that the structure of the graph and the distribution of the node types interact and
heavily restrict the structure of the feasible solutions. We can oer only one result that
provides necessary and sucient conditions. It is, again, for the structurally rather simple
cut inequalities and for the edge-connectivity case. Moreover, we have to assume that the
given graph is highly connected and all nodes have the same type. This result is due to Stoer
(1992).
(6.12) Theorem. Let G = (V, E) be a (k + 1)-edge connected graph, let rv = k for all nodes
v V , and let W = V be a nonempty node set. Dene for each Wi W with = Wi = W
the decit of Wi as
def G (Wi ) := max{0, k |G[W ] (Wi )|}.
Dene similarly for Ui V \ W with = Ui = V \ W
def G (Ui ) := max{0, k |G[V \W ] (Ui )|}.

35
The cut inequality
x((W )) k
denes a facet of the polytope kECON(G; r) of k-edge connected graphs if and only if

(a) G[W ] and G[V \ W ] are connected, and

(b) for all edges e E(W ) E(V \ W ),

for all pairwise disjoint node sets W1 , . . . , Wp of W , with Wi = W for all i = 1, . . . , p
(p 0), and
for all pairwise disjoint node sets U1 , . . . , Uq of V \ W , with Ui =  V \ W for all
i = 1, . . . , q (q 0),
the following inequality holds:
p
 q
 p
 q

def Ge (Wi ) + def Ge (Ui ) |[ Wi : Ui ]| k. 2
i=1 i=1 i=1 i=1

Summarizing, we can say that, in the general kECON or kNCON case, facet characteriza-
tions are technically rather dicult and that in many cases it seems quite hopeless to get
(reasonable) necessary and sucient conditions for an inequality (of one of the classes intro-
duced above) to dene a facet of kECON(G; r) or kNCON(G; r). Nevertheless, the results
established so far (i.e., the sucient conditions) show that these classes do contain many
facet-dening inequalities.

6.3 Separation
Note that except for the trivial inequalities all classes of valid inequalities for the kECON
and kNCON problem described in Section 6.1 contain a number of inequalities that is ex-
ponential in the number of nodes of the given graph. So it is impossible to input these
inequalities into an LP-solver. But there is an alternative approach. Instead of solving an
LP with all inequalities, we solve one with a few carefully selected inequalities and we gen-
erate new inequalities as we need them. This approach is called a cutting plane algorithm
and works as follows.
We start with an initial linear program. In our case, it consists of the linear program (6.6)
without the integrality constraints (iii). We solve this LP. If the optimum solution y is
feasible for the kECON or kNCON problem, then we are done. Otherwise we have to nd
some inequalities that are valid for kECON(G; r) or kNCON(G; r) but are violated by y.
We add these inequalities to the current LP and repeat.

36
The main diculty of this approach is in eciently generating violated inequalities. We
state this task formally.
(6.13) Separation Problem (for a class C of inequalities)
Given a vector y decide whether y satises all inequalities in C and, if not, output an
inequality violated by y.
A trivial way to solve (6.13) is to substitute y into each of the inequalities in C and check
whether one of the inequalities is violated. But in our case this is too time consuming since
C is of size exponential in |V |. Note that all the classes C described before have an implicit
description by means of a formula with which all inequalities can be generated. It thus may
happen that algorithms can be designed that check violation much more eciently than the
trivial substitution process. We call an algorithm that solves (6.13) an (exact) separation
algorithm for C, and we say that it runs in polynomial time if its running time is bounded
by a polynomial in |V | and the encoding length of y.
A deep result of the theory of linear programming (see Grotschel, Lovasz, Schrijver (1988))
states (roughly) that a linear program over a class C of inequalities can be solved in poly-
nomial time if and only if the separation problem for C can be solved in polynomial time.
Being able to solve the separation problem thus has considerable theoretical consequences.
This result makes use of the ellipsoid method and does not imply the existence of a prac-
tically ecient algorithm. However, by combining separation algorithms with other LP
solvers (like the simplex algorithms) can result in quite successful cutting plane algorithms;
see Section 7.
Our task now is to nd out whether reasonable separation algorithms can be designed for
any of the classes (6.2), (6.3), (6.4), (6.5), (6.8)(i), (6.9) or (6.10).
There is some good and some bad news. The good news is that for the cut inequalities
(6.2), the node cut inequalities (6.3) and the r-cover inequalities (6.8)(i), exact separation
algorithms are known that run in polynomial time; see Grotschel, Monma and Stoer (1992c).
Let us consider the class C of cut inequalities (6.2), let y be some vector in RE with 0 ye 1
for all e E and let r be in ZV+ . We view the components ye as capacities of the edges of the
given graph G = (V, E) and compute a Gomory-Hu tree (using the Gomory-Hu algorithm
that consists of |V | 1 calls of a max-ow algorithm and some bookkeeping; see Gomory-
Hu (1961)). The Gomory-Hu tree has the property that, for any two nodes u, v V , the
minimum capacity of a cut separating u and v is given by the smallest weight of an edge
that is contained in the unique path linking u and v in the Gomory-Hu tree. Having the
Gomory-Hu tree T with weights we for all e T we can determine whether y satises all
cut inequalities as follows. For each edge e, let Ve and Ve denote the node sets of the two

37
components of (V, T e). If there is an edge e E with

we < min{max{r v | v Ve }, max{rv |v Ve }}

then y violates the inequality x((Ve )) r(W ), otherwise y satises all cut inequalities
x((W )) r(W ). Hao & Orlin (1992) describe a faster algorithm that nds a minimum
cut in a graph without nding a Gomory-Hu tree, and whose running time is equal to the
running time of one maxow routine. The separation problem for the node cut constraints
(6.3) can be reduced to a sequence of minimum (s, t)-cut computations in a directed graph.
This polynomial time method is described in Grotschel, Monma & Stoer (1992c).
The polynomial time exact separation algorithm for the r-cover inequalities is based on the
Padberg-Rao procedure for solving the separation problem for the capacitated b-matching
inequalities; see Padberg and Rao (1982). The trick is to reverse the transformation from
the b-matching to the r-cover problem described in (6.7) and (6.8) and call the Padberg-Rao
algorithm. It is easy to see that y satises all r-cover inequalities (6.8)(i) if and only if its
transformation satises all b-matching inequalities (6.7)(i). The Padberg-Rao procedure is
quite complicated to describe, so we do not discuss it here.
The bad news is that it was shown in Grotschel, Monma and Stoer (1992b) that the sep-
aration problems for partition inequalities (6.4), node partition inequalities (6.5) and lifted
r-cover inequalities (6.9) are NP-hard. (See Section 8 for a way to cure this problem in some
cases by using directed models and projections.)
Thus, in these cases we have to revert to separation heuristics, i.e., fast procedures that
check whether they can nd an inequality in the given class that is violated by y, but which
are not guaranteed to nd one even if one exists. We discuss separation heuristics in more
detail in Section 7.

6.4 Complete Descriptions of Small Cases

For the investigation of combinatorially-dened polyhedra, it is often useful to study small
cases rst. A detailed examination of such examples provides insight into the relationship
between such polyhedra and gives rise to conjectures about general properties of these poly-
hedra. Quite frequently a certain inequality is found, by numerical computation, to dene a
facet of a kECON or kNCON polytope of small dimension. Afterwards it is often possible
to come up with a class (or several classes) of inequalities that generalize the given one and
to prove that many inequalities of these classes dene facets of combinatorial polytopes of
any dimension.

38
By means of a computer program, we have computed complete descriptions of all kECON
and kNCON polytopes of small dimensions. To give a glimpse of these numerically obtained
results, we report here the complete descriptions of all 2ECON and all 2NCON polytopes of
the complete graphs on ve vertices K5 . More information about small kECON polytopes
can be found in Stoer (1992).

6.4.1 The 2ECON Polytope for K5

Let us begin with the polytopes 2ECON (K5 , r) where r = (r1 , . . . , r5 ) is the vector of
node types. The node types ri have value 0, 1 or 2, and by assumption, at least two
nodes are of highest type 2. Clearly, we can suppose that ri ri+1 for i = 1, . . . , 4. These
assumptions result in ten node type vectors to be considered. It is obvious that, if a node type
vector r componentwise dominates a vector r  (i.e., ri ri for all i) then 2ECON(Kn , r)
2ECON(Kn , r ).
Figure 6.1 provides a comprehensive summary of our ndings. In this gure, a polytope
2ECON(K5 , r) is depicted by its node type vector r = (r1 , . . . , r5 ). A line linking two such
vectors indicates that the polytope at the lower end of the line directly contains the polytope
at the upper end of the line and that no other 2ECON polytope is in between. For exam-
ple, the polytope 2ECON(K5 , (2, 2, 2, 1, 0)) is directly contained in 2ECON(K5 , (2, 2, 1, 1, 0))
and 2ECON(K5 (2, 2, 2, 0, 0)), and it contains directly 2ECON(K5 , (2, 2, 2, 1, 1)) and 2ECON
(K5 (2, 2, 2, 2, 0)). Next to the right or left of a node type vector r, a box indicates which
inequalities are needed to dene the corresponding 2ECON(K5 , r) polytope completely and
nonredundantly. The notation is as follows:

The type of inequality appears in the rst column.

- p stands for partition inequality, see (6.4),

- cut stands for cut constraint, see (6.2),
- rc stands for lifted r-cover inequality, see (6.9),
- d stands for degree constraint, see (6.6)(i),
- rc+1, I1, and I2, stand for new types of inequalities explained later.

The next column lists, for each partition of the handle (rc) or the whole node set
(p), the node types in each node set. The dierent sets are separated by commas.

- For instance, p 200,2,1 stands for a partition inequality induced by a partition

of V , whose rst node set contains one node of type 2 and two nodes of type 0,

39
 whose second set contains exactly one node of type 2, and whose last set contains
exactly one node of type 1.
- rc 20,2,2 stands for a lifted r-cover inequality induced by a handle that is parti-
tioned into three node sets, the rst one containing two nodes, a node of type 2
and a node of type 0, the second node set containing exactly one node of type 2,
and the third node set containing exactly one node of type 2; the number of teeth
can be computed with the help of the right-hand side.

The next column gives the right-hand side.

The last column contains the number of inequalities of the given type.

We do not list the trivial inequalities 0 xe 1. We know that xe 1 always denes

a facet of the considered polytopes, and that xe 0 denes a facet if and only if e is not
contained in a cutset (W ) of size con(W ) + 1.
The inequalities denoted by rc+1, I1, and I2 in the polytopes 2ECON(K5 ; (2 2 2 0 0)),
and 2ECON(K5 ; (2 2 2 1 0)), are depicted in Figure 6.2. All except I2 have coecients in
{0, 1, 2}. The coecients of inequality I2 in Figure 6.2 take values in {0, 1, 2, 3}. Edges
of coecient 0 are not drawn, edges of coecient 1 are drawn by thin lines, and edges
of coecient 2 are drawn by bold lines. To make this distinction somewhat clearer, we
additionally display the coecients of all thin or of all bold lines.
This numerical study of 2ECON polytopes of K5 reveals that degree, cut, partition and lifted
r-cover inequalities play a major role in describing the polytopes completely and nonredun-
dantly. But at the same time we found three new classes of (complicated) inequalities that
are presently the subject of further studies.

6.4.2 The 2NCON Polytope for K5

We now turn our attention to the 2NCON polytopes of the complete graph K5 . It turned
out that only two further classes of inequalities are needed to describe all polytopes 2NCON
(K5 , r) completely and nonredundantly. These are the classes of node cut and node partition
inequalities (6.3) and (6.5). Figure 6.3 displays the complete descriptions of the 2NCON
polytopes for K5 in the same way as Figure 6.1 does for the 2ECON polytopes. The (new)
entries for the node cut and node partition inequalities read as follows.
ncut 20,20 denotes a node cut inequality x(Gz (W )), where W contains a node of type 2
and a node of type 0, and V \(W {z}) contains a node of type 2 and a node of type 0; the

40
. in ncut 2.,2. represents a node of any type; and np 20,2,2 denotes a node partition
inequality induced by a partition of V \{z}, where z is some node in V , and the rst shore
of the partition consists of a node of type 2 and a node of type 0, the second shore consists
of a node of type 2, and the third shore consists of a node of type 2.
This concludes our examination of polyhedra for small instances of 2NCON and 2ECON.
For further such results, see Stoer (1992).

41
 cut 200,20 2 6
22000 d 2 2

p 20,2,10 3 6
p 200,2,1 3 2
cut 210,20 2 4
22100 d 1 1
d 2 2
; @
; @
; @ I2 6 2
; @ I1 4 6
p 2,2,1,10 4 2 ; @ rc 20,2,2 2 6
p 20,2,1,1 4 2 ; @ rc 2,2,2 2 8
p 210,2,1 3 4 cut 220,20 2 6
cut 211,20 2 2 22110 22200 d 2 3
d 1 2 H
HH
d 2 2 H rc+1 2,2,2,1 8 1
HH rc 2,2,2,1 3 1
HH p 2,2,2,10 4 1
HH p 20,2,2,1 4 3
p 2,2,1,1,1 5 1 H I2 6 1
H
p 211,2,1 3 6 I1 4 3
d 1 3 22111 2 2 2 1 0 rc 20,2,2 2 3
d 2 2  rc 2,2,2 2 7

  p 220,2,1 3 3
 cut 221,20 2 3

 d 1 1
p 2,2,2,1,1 5 1  d 2 3
rc 2,2,2,1 3 2  rc 2,2,2,2 3 4
rc 2,2,2 2 6
p 221,2,1 3 6 22211 2 2 2 2 0 I2 6 4
d 1 2 I1 4 12
@ ; rc 20,2,2 2 12
d 2 3 @ ; rc 2,2,2 2 16
@ ; cut 222,20 2 4
@ ; d 2 4
@ ;
@ ; p 2,2,2,2,1 5 1
2 2 2 2 1 rc 2,2,2,2 3 4
rc 2,2,2,1 3 4
rc 2,2,2 2 12
p 222,2,1 3 4
d 1 1
d 2 4

rc 2,2,2,2 3 20
22222 d 2 5

Figure 6.1: 2ECON(K5 ; r) polyhedra

42
 0
rc + 1 for 2 2 2 1 0 I1 for 2 2 2 1 0
2
2 2
1 H 
2 H2
 C H C
H

1 
 C HH  C HH
0 C 2 0 C 2
Q1  C 
 Q C
A Q    A Q 2 
A Q 
C A QQ C
 Q C  2
AA   Q AA Q CC 
 QC  Q
1 2 2 2

8 4

0
I2 for 2 2 2 1 0 2
2 1
2 no line 0

C HH
2 3
 2 C
H
H
0 C 2
Q2 C
A Q 3 
A QQ C
3
AA Q CC 
Q
2 2

Figure 6.2: Inequalities for 2ECON(K5 ; r)

43
 ncut 20,20 1 6
2 2 0 0 0 cut 200,20 2 6
d 2 2

p 20,2,10 3 6
p 200,2,1 3 2
d 1 1
ncut 2.,2. 1 6
cut 210,20 2 4
22100 d 2 2
; @
; @
; @
p 2,2,1,10 4 2 ; @
p 20,2,1,1 4 2 ; @
p 210,2,1 3 4 ; @ np 20,2,2 2 6
ncut 20,20 1 6
d 1 2 22110 2 2 2 0 0 cut
ncut 2.,2. 1 6 220,20 2 6
H d 2 3
cut 211,20 2 2 HH
d 2 2 H
HH rc+1 2,2,2,1 8 1
HH np 2,2,2,1 3 1
HH p 2,2,2,10 4 1
p 2,2,1,1,1 5 1 H
p 211,2,1 3 6 H p 20,2,2,1 4 3
np 2.,2,2 2 6
d 1 3 22111 22210 p
ncut 21,21 1 6 220,2,1 3 3
 d 1 1
d 2 2 
  ncut 21,20 1 6
 cut 221,20 2 3

  d 2 3
p 2,2,2,1,1 5 1 
np 2,2,2,1 3 2 
np 21,2,2 2 6  np 2,2,2,2 3 1
np 20,2,2 2 12
p 221,2,1 3 6 22211 2 2 2 2 0 cut 222,20 2 4
d 1 2 d 2 4
ncut 21,21 1 6 @ ;
d 2 3 @ ;
@ ;
@ ;
@ ;
@ ; np 2,2,2,2 3 1
22221 p 2,2,2,2,1 5 1
np 2,2,2,1 3 4
np 21,2,2 2 12
p 222,2,1 3 4
d 1 1
d 2 4

np 2,2,2,2 3 5
22222 d 2 5

Figure 6.3: 2NCON(K5 ; r) polyhedra

44
7 Computational Results
For applied mathematicians, the ultimate test of the power of a theory is its success in helping
solve the practical problems for which it was developed. In our case, this means that we
have to determine whether the polyhedral theory for the survivable network design problem
can be used in the framework of a cutting plane algorithm to solve the design problems of
the sizes arising in practice. The results reported in Grotschel, Monma and Stoer (1992b)
show that the design problems for the LATA networks arising at Bellcore can be solved to
optimality quite easily. There is good reason to hope that other network design problems
of this type can also be attacked successfully with this approach. Moreover, the known
heuristics also seem to work quite well, at least for the low connectivity case.

7.1 Outline of the Cutting Plane Algorithm

We have already mentioned in Section 6.3 how a cutting plane approach for our problem
works. Let us repeat this process here a little more formally.
We assume that a graph G = (V, E) with edge cost ce R for all e E and node types
rv Z+ for all v V is given. Let k := max{rv | v V }. We want to solve either the
kNCON or the kECON problem for G and r and the given cost function, i.e., we want to
solve

min cT x or min cT x
x kNCON(G; r) x kECON(G; r).

We do this by solving a sequence of linear programming relaxations that are based on the
results we have described in Section 7. The initial LP (in both cases) consists of the degree
constraints and trivial inequalities, i.e.,

(7.1) min cT x
x((v)) rv for all v V ;
0 xe 1 for all e E.

If a current LP-relaxation has been solved and z is a basic optimum solution, we check
whether z is in kNCON(G; r) or kECON(G; r). If it is, the problem is already solved to
optimality. Otherwise we call our separation routines and try to nd inequalities in one of
the classes described in Section 6 that are violated by z. If one of the separation algorithms
is successful, we add the inequalities found to the current LP and repeat the procedure. If

45
no violated inequality can be identied, there are two options. We may simply stop and
report a lower bound or we may resort to an enumerative procedure like branch and bound.
The best option to select in the case that the cutting plane phase has not produced an
optimum solution of the kECON or kNCON problem depends on the demands of practice.
Before starting the cutting plane algorithm, one usually runs some heuristics to produce
(hopefully) good feasible solutions, see Section 4 for a description of such heuristics. By
comparing the lower bound L from the cutting plane algorithm with the upper bound U from
the heuristics, one can easily get an idea about the quality of the solutions. If the percentage
deviation of these values, usually taken as 100 (U L)/L, is less than some threshold,
say 5%, and if the cost data were somewhat fuzzy anyway, the best heuristic solution might
simply be considered as an appropriate good solution of the practical problem. If the data
are precise and an optimum solution is needed, branch and bound has a good chance to
terminate in a reasonable amount of time.
If, however, the deviation of U and L is large, no simple advice can be given. Either the
heuristic or the cutting plane algorithm or both may have produced poor bounds. Further
research is usually necessary to determine the reasons for failure, to detect special structures
of the problem in question that may have resulted in traps for the heuristics or poor perfor-
mance of the cutting plane algorithms. If such structural insight can be obtained, one has a
good chance to improve the result, at least for the current problem instance.
The last case is denitely unsatisfactory, but since we are dealing with hard problems, we have
to expect such behavior every now and then. For our real world applications we can report
that, for the LATA network design problems, the cutting plane algorithm always found an
optimum integral solution, except in three cases that were easily solved by branch and bound
or manual interaction. Random problems with low and high connectivity requirements and
random cost structure were solved to optimality extremely quickly. But there is one large
scale practical problem with 494 nodes, and 1096 edges and highly structured topology and
costs, where we ran into considerable diculties. Using special purpose separation heuristics
etc., we were nally able to solve two versions of the problem to optimality, with quite some
eort however.

7.2 Implementation Details

We will not discuss implementation details of the heuristics outlined in Section 4; see Monma
and Shallcross (1989) for details. This is quite straightforward, although of course, the
use of good data structures and search strategies may result in considerable running time
improvements.

46
We will focus here on implementation issues of the cutting plane algorithm, a number of
which are vital for obtaining satisfactory running time performances. Before starting the
cutting plane algorithm we try to reduce the size of the problem instance by decomposing it.
In fact, the practical problems we solved have rather sparse graphs of possible direct links,
and the survivability requirements often force certain edges to be present in every feasible
solution. Such edges can be xed and removed from the problem by appropriately changing
certain node types. This removal may break the original problem into several smaller ones
that can be solved independently.
There are further ways of decomposing a problem into independent subproblems like decom-
posing on articulation nodes, on cut sets of size two, and on articulation sets of size two. In
each of these cases one can perform the decomposition or determine that no such decom-
position is possible, using polynomial time methods like depth-rst search or connectivity
algorithms. All of this is quite easy, though a precise description of the necessary transfor-
mations would require considerable space. Details can be found in Grotschel, Monma and
Stoer (1992b), and Stoer (1992).
The main purpose of this decomposition step is to speed up the computation by getting rid of
some trivial special cases that the cutting plane algorithm does not need to check any more,
and by reducing the sizes of the problems to be solved. At the end of this preprocessing
phase we have decomposed the original problem into a list of subproblems for which we
call the cutting plane algorithm. The optimal solution of the original problem can then be
composed from the optimal solutions of the subproblems in a straightforward manner.
An issue of particular importance is the implementation of the separation algorithms. Good
heuristics have to be found that solve the separation problems for those classes that are
not known to be separable in polynomial time. And even if polynomial exact separation
routines are known, separation heuristics may help considerably to speed up the overall
program. Further problems are to determine the order in which the heuristic and exact
separation routines are to be called, when to stop the separation process in case many
violated inequalities have been found, which cutting planes to add and which to eliminate
from the current LP. These issues can not be decided by theoretical insight only. Practical
experience on many examples is necessary to come up with recipes that result in satisfactory
overall performance of such an algorithm. We outline some of the techniques used in the
sequel.
Let G be a graph with node types r ZV+ , and let y be some point in RE with 0 ye 1.
Our aim is to nd a partition (resp. node partition, lifted r-cover) inequality violated by this
point. By the NP-completeness results mentioned in Section 6.3, it seems hopeless to nd
an ecient exact algorithm for the separation of these inequalities; therefore we have to use
heuristics. Nevertheless, it is possible to solve the separation problem for a certain subclass

47
of these inequalities, namely cut constraints (resp., node cut and r-cover constraints) in
polynomial time. So in our heuristics we often use almost violated inequalities of these
subclasses and transform them into violated inequalities of the larger class. Here an almost
violated inequality is an inequality aT x b with aT y b + for some small parameter
(we used = 0.5). aT x b is a violated inequality, if aT y < b.
The heuristic that we applied has the following general form:
(7.2) Heuristic for Finding Violated Partition Inequalities

1. Shrink all or some edges e E with the property that any violated partition inequality
using this edge can be transformed into some at-least-as-violated partition inequality
not using this edge. (Using e means; e has coecient 1 in the partition inequality.)

2. Find some violated or almost violated cut constraints in the resulting graph.

3. Attempt to modify these cut constraints into violated partition inequalities.

Exactly the same approach is used for separating node partition (6.5) and lifted r-cover
inequalities (6.9), except that we have to use other shrinking criteria and, in Step 2, plug in
the appropriate subroutine for separating node cut (6.2), resp., r-cover constraints (6.8)(i).
Shrinking is important for reducing graph sizes before applying the Gomory-Hu algorithm
for nding cut constraints, which has the rather high complexity of O(|V |4 ). If we are looking
for related partition inequalities, we test whether edge e = uv satises one of the following
shrinking criteria.
(7.3) Shrinking Criteria

1. ye q := max{rw : w V }

2. ye rv and ye y((v)) ye

3. ye max{y((v)) ye , y((u)) ye } and there is a node w  {u, v} with rw

max{ru, rv }.

If these criteria are satised for edge e, we shrink it by identifying u and v, giving type
r({u, v}) to the new node, and identifying parallel edges by adding their y-values.
It can be shown, that if cases (7.3)1. or 2. apply, then any violated partition inequality using
e can be transformed into some at-least-as-violated partition inequality not using e. In case

48
(7.3)3., edge e has the same property with respect to cut inequalities. Similar shrinking
criteria can be found for node partition inequalities and lifted r-cover inequalities.
In the reduced graph G we now nd violated or almost violated cut constraints (resp., node
partition and r-cover constraints) using the Gomory-Hu algorithm (or, for r-cover constraints
the Padberg-Rao algorithm). These inequalities, dened for G , are transformed back into
the original graph G. For instance, a cut inequality in G , x(G (W )) r(W  ) is rst
transformed into a cut inequality x(G(W )) r(W ) in G by blowing up all shrunk nodes
in W  . This provides the enlarged node set W . Secondly, this cut inequality is transformed
into a (hopefully) violated partition inequality by splitting W or V \ W into smaller disjoint
node sets W1 , . . . , Wp .
We also check whether the given cut inequality satises some simple necessary criteria for
dening a facet of kECON(G; r) (or kNCON(G; r)). If this is not so, it can usually be trans-
formed into a partition inequality that denes a higher-dimensional face of the respective
polyhedron. A similar approach is taken for node partition and lifted r-cover inequalities.
More details can be found in Grotschel, Monma and Stoer (1992b) and in Stoer (1992).
Typically, in the rst few iterations of the cutting plane algorithm, the fractional solution y
consists of several disconnected components. So, y violates many cut and partition inequali-
ties but usually no lifted r-cover inequalities. We start to separate lifted r-cover inequalities
only after the number of violated partition inequalities found drops below a certain thresh-
old. Node partition inequalities are used only after all other separation algorithms failed in
nding more than a certain number of inequalities.
To keep the number of LP constraints small, all inequalities in the current LP with non-zero
slack are eliminated. But since all inequalities ever found by the separation algorithms are
stored in some external pool they can be added again if violated at some later point.

7.3 Computational Results for Low-Connectivity Problems

In this section we describe the computational results based on the practical heuristics de-
scribed in Section 4 and the cutting plane approach described earlier in this section. We
consider the low-connectivity case with node types in {0, 1, 2} here and the high connectivity
case in the next section. All running times reported are for a SUN 4/50 IPX workstation (a
28.5 MIPS machine). The LP-solver used is a research version of the CPLEX-code provided
to us by Bixby (1991). This is a very fast implementation of the simplex algorithm.
To test our code, network designers at Bellcore provided the data (nodes, possible direct
links, costs for establishing a link) of seven real LATA networks that were considered typical
for this type of application. The sizes ranged from 36 nodes and 65 edges to 116 nodes

49
and 173 edges; see Table 7.1. The problem instances LATADL, LATADS, and LATADSF
are dened on the same graph. The edges have the same costs in each case, but the node
types vary. Moreover, in LATADSF, 40 edges were required to be in the solution. (The
purpose was to check how much the cost would increase if these edges had to be used, a
typical situation in practice, where alternative solutions are investigated by requiring the use
of certain direct links.)
Table 7.1 provides information about the problems. Column 1 contains the problem names.
For the original graphs, columns 2, 3, and 4 contain the numbers of nodes of type 0, 1,
and 2, respectively; column 5 lists the total number of nodes, column 6 the number of edges
and the number of edges required to be in any solution (the forced edges). All graphs were
analysed by our preprocessing procedures described in Section 7.2. Preprocessing was very
successful. In fact, in every case, the decomposition and xing techniques ended up with a
single, much smaller graph obtained from the original graph by splitting o side branches
consisting of nodes of type 1, replacing paths where all interior nodes are of degree 2, by a
single edge, etc. The data of the resulting reduced graphs are listed in columns 6, . . . , 10 of
Table 7.1.

Original Graphs Reduced Graphs

Problem 0 1 2 Nodes Edges 0 1 2 Nodes Edges
LATADMA 0 12 24 36 65/0 0 6 15 21 46/4
LATA1 8 65 14 77 112/0 0 10 14 24 48/2
LATA5S 0 31 8 39 71/0 0 15 8 23 50/0
LATA5L 0 36 10 46 98/0 0 20 9 29 77/1
LATADSF 0 108 8 116 173/40 0 28 11 39 86/26
LATADS 0 108 8 116 173/0 0 28 11 39 86/3
LATADL 0 84 32 116 173/0 0 11 28 39 86/6
Table 7.1 Data for LATA Problems

To give a visual impression of the problem topologies and the reductions achieved, we show
in Figure 7.1 a picture of the original graph of the LATADL problem (with 32 nodes of type 2
and 84 nodes of type 1) and in Figure 7.2 a picture of the reduced graph (with 39 nodes and
86 edges) after preprocessing. The nodes of type 2 are displayed by squares, and the nodes
of type 1 are displayed by circles. The 6 forced edges that have to be in any feasible solution
are drawn bold.
LATA1 is a 2ECON problem, while the other six instances are 2NCON problems. All opti-
mum solutions of the 2ECON versions turned out to satisfy all node-survivability constraints
and thus were optimum solutions of the original 2NCON problems with one exception. In
LATA5L one node is especially attractive because many edges with low cost lead to it. This

50
node is an articulation node of the optimum 2ECON solution. In the following, LATA5LE
is the 2ECON version of problem LATA5L.
Table 7.2 contains some data about the performance of our code on the eight test instances.
The entries from left to right are:

IT number of iterations ( = calls to the LP-solver)

P number of partition inequalities (6.4) used in the cutting plane phase
NP number of node partition inequalities (6.5) used in the cutting plane phase
RC number of lifted r-cover inequalities (6.9) used in the cutting plane phase
C value of the optimum solution after termination of the cutting plane phase
COPT optimum value
GAP 100 (COPT C)/COPT ( = percent relative error at the end of the
cutting plane phase)
T total running time including input, output, preprocessing, etc.,
of the cutting plane phase (not including branch & cut), in rounded
BN number of branch & cut nodes generated
BD maximum depth of the branch & cut tree
BT total running time of the branch & cut algorithm including
the cutting plane phase, in seconds

Problem IT P NP RC C COPT GAP T BN BD BT

LATADMA 12 65 3 7 1489 1489 0 1
LATA1 4 73 0 1 4296 4296 0 1
LATA5S 4 76 0 0 4739 4739 0 1
LATA5LE 7 120 0 0 4574 4574 0 1
LATA5L 19 155 12 0 4679 4726 0.99 2 4 2 4
LATADSF 7 43 0 0 7647 7647 0 1
LATADS 17 250 0 4 7303.60 7320 0.22 4 28 9 17
LATADL 14 182 0 28 7385.25 7400 0.20 3 32 10 21
Table 7.2. Performance of branch & cut on LATA problems

We think that it is worth noting that each of these real problems, typical in size and structure,
can be solved on a 28-MIPS machine in less than thirty seconds including all input and output
routines, drawing the solution graph, branch and cut, etc.

51
A detailed analysis of the running times of the cutting plane phase is given in Table 7.3. All
times reported are in percent of the total running time TIME (without the branch & cut
phase). The entries from left to right are:

PT time spent in the preprocessing phase (in percent)

CT time spent in the separation routines (in percent)
LPT time used by the LP-solver (in percent)
MT miscellaneous time for input, output, drawing, etc. (in percent)
TT total time (in seconds)
TT\RED total time (in seconds) of the algorithm when applied to the original
instance without prior reduction by preprocessing

The last column TT\RED shows the running times of the cutting plane phase of our algo-
rithm applied to the full instances on the original graphs (without reduction by preprocess-
ing). By comparing the last two columns, one can clearly see that substantial running time
reductions can be achieved by our preprocessing algorithms on the larger problems.

PROBLEM PT% LPT% CT% MT% TT TT\RED

LATADMA 2.0 39.2 41.2 17.6 1 1
LATA1 3.8 34.6 34.6 26.9 1 4
LATA5S 3.8 34.6 34.6 26.9 1 1
LATA5LE 0.0 42.9 41.1 16.1 1 1
LATA5L 0.7 37.1 55.2 7.0 2 5
LATADSF 2.1 21.3 57.4 19.2 1 4
LATADS 0.0 44.7 49.0 6.4 4 17
LATADL 1.0 26.3 66.2 6.5 3 18
Table 7.3. Running times of cutting plane algorithm on LATA problems

A structural analysis of the optimum solutions produced by our code revealed that except
for LATADSF, LATA5LE, and LATA1 the optimum survivable networks consist of a long
cycle (spanning all nodes of type 2 and some nodes of type 1) and several branches connecting
the remaining nodes of type 1 to the cycle. The optimum solution of the LATADL instance
in shown in Figure 7.3, with the 2-connected part (the long cycle) drawn bold.
From the view of a telephone network designer, a long cycle connecting all nodes of type 2
is not a desirable feature in a communication network because routing paths are very long,
resulting in delays, and because each link has to carry a high trac load, resulting in high

52
costs for terminal electronics, multiplexers, etc. But since the network installation costs
form part of the whole network cost, the lowest network installation cost (as found by our
algorithm) provides a lower bound for the whole network cost, and the subgraph minimizing
the installation cost could be modied, e.g., by adding some more links of low cost, to
produce a network with shorter routing paths between each pair of nodes. This is the design
approach taken in the software package distributed by Bellcore (1988); rst a survivable
network topology of low cost in computed by heuristics, then this topology is modied to
account also for costs associated with the expected trac in the network.
We ran a few tests on randomly generated problems of higher density and 50-100 nodes.
Here our code did perform reasonably well but not as well as on sparse problems. (That is
not of great importance, since our goal was to solve real-world problems and not random
problems.) More serious is a dramatic increase in running time when many nodes of type 0
are added, as is the case in the ship problem treated in the next section. Here our cutting
plane code for the low-connectivity case takes very long before the intermediate fractional
solutions become connected. For such cases new separation heuristics have to be developed
that perform a more sophisticated structural analysis of the given instance. But the problems
that we address here mainly and that come up in the design of ber optic telephone networks
have very few nodes of type 0, if any.
Another motivation for our work was to nd out how well the heuristics of Monma and
Shallcross (1989) described in Section 4 perform. It turned out that they do very well.
Table 7.4 compares the values CHEUR of the solutions produced by the heuristics with the
optimum values COPT computed by our code. The percent relative error GAP ( = 100
(CHEUR - COPT)/COPT) is always below 1.5%. In three cases the heuristics found
an optimum solution. This result denitely justies the present use of these heuristics in
practice. We note that these heuristics are very fast, typically taking only a few seconds on
a IBM PC/AT.

PROBLEM COPT CHEUR GAP

LATADMA 1489 1494 0.34
LATA1 4296 4296 0
LATA5S 4739 4739 0
LATA5LE 4574 4574 0
LATA5L 4726 4794 1.44
LATADSF 7647 7727 1.05
LATADS 7320 7361 0.56
LATADL 7400 7460 0.81
Table 7.4: Comparison of heuristic values with optimal values

53
Figure 7.1: Original graph of LATADL-problem

54
Figure 7.2: Reduced graph of LATADL-problem

55
Figure 7.3: Solution of LATADL-problem

56
7.4 Computational Results for High-Connectivity Problems
At present, we have a rst preliminary version of a code for solving survivability prob-
lems with higher connectivity requirements. In order to test our code for general kNCON
problems, we rst used a set of random problems. Later, we also obtained test data for a
real-world 3NCON problem, which arose in the design of a communication network on a
ship. Both types of test problems have their drawbacks, however. The random problems
turned out to be too easy (most of them were already solved in the rst iteration), and the
ship problem confronted us with so many new diculties (with respect to space, running
time, and quality of solutions) that we have to redesign our separation strategies completely
to solve variants of the ship problem to optimality.

7.4.1 Random Problems

We rst report about our computational results on random kECON problems. We used the
same set of random data as Ko and Monma (1989) used for their high-connectivity heuristics.
So we will be able to compare results later.
The test set of Ko & Monma consists of ve complete graphs of 40 nodes and ve complete
graphs of 20 nodes, whose edge costs are independently drawn from a uniform distribution
of real numbers between 0 and 20. For each of these 10 graphs, a minimum-cost k-edge
connected subgraph for k = 3, 4, 5 is to be found. The next table reports the number of
iterations (minimum and maximum) and the average time taken by our code to solve these
problems for k = 3, 4, and 5, respectively. Only the time for the cutting plane phase is given.

# Nodes # Iterations Average Time (secs)

k= 3 4 5 3 4 5
20 nodes: 1-2 1-5 1-4 0.43 0.51 0.58
40 nodes: 1-2 1-2 1-4 1.54 1.95 2.36

All problems except one 3ECON instance on 20 nodes were solved in the cutting plane phase.
In fact 20 of the 30 problems were already solved in the rst iteration with the initial LP
(7.1). For the instances not solved in the rst iteration, at most four lifted r-cover inequalities
(6.9) had to be added to obtain the optimal solution. Except for one 3ECON instance, no
partition inequalities were added. So, the average solution time is mainly the solution time
for the rst LP.
All optimal solutions for the kECON problems were at the same time feasible and hence
optimal for the corresponding kNCON problems, except the one 3ECON problem which

57
could not be solved in the cutting plane phase. There the optimal solution (obtained by
branch & cut) is 3-edge connected, but not 3-node connected.
These excellent results were surprising, because we always thought high-connectivity prob-
lems to be harder than low-connectivity problems. But this does not seem to be true for
random costs.
The high-connectivity heuristics of Ko and Monma did not perform quite as well as the low-
connectivity heuristics, but still reasonably well. The relative gap between the heuristic (h)
and the optimal solution value (o), namely 100 (h o)/o, computed for the above set of
random problems, ranged between 0.8 and 12.8 with an average of 6.5 % error (taken over
all problems).

7.4.2 Ship Problems

One real-world application of survivable network design, where connectivities higher than
two are needed, is the design of a ber communication network that connects locations on
a military ship containing various communication systems. The reason for demanding high
survivability of this network is obvious.
The problem of nding a high-connected network topology minimizing the cable installation
cost can be formulated as a 3NCON problem. We will describe the characteristics of this
problem in the following.
We obtained the graph and edge cost data of a generic ship model. It has the following
features. The graph of possible link installations has the form of a three-dimensional grid
with 15 layers, 494 nodes, and 1096 edges, which is depicted in Figure 7.4. The problem to
be solved in this graph is a 3NCON problem with the following node types and costs.
Of the grids 494 nodes, only 33 are nonzero type, called special nodes. They are drawn by
lled circles or triangles. The 33 special nodes symbolize the various communication systems
to be interconnected by the network. To evaluate the dependence of network topology cost
on the required survivability, the ship problem appears in three dierent versions depending
on the node types of the 33 special nodes. The three nodes depicted by triangles in the
tower of the ship always have type 3, the other 29 special nodes are all given either type 1,
type 2, or type 3. We call the three resulting versions of the ship problem ship13, ship23,
and ship33, respectively. The remaining 461 nodes are nodes of type 0. They represent
possible ber junction boxes where the ber cable may be routed.
The cost structure is highly regular. The costs are proportional to the distances between
nodes, with the feature that horizontal distances are much higher than vertical distances.

58
(The grid shown in Figure 7.4 has been scaled. Also, contrary to the graphical representation,
the horizontal layers do not always have the same distance from each other.) With this cost
structure, it is much cheaper to route vertically than horizontally. Since there exist many
shortest paths between any two nodes, there will also exist many optimum solutions to the
survivable network problem. So the problem is highly degenerate. Degeneracy together
with the size of the ship problem caused us to run into diculties. In fact, when we rst
applied our code to the ship13 problem, with the initial LP consisting only of the degree
constraints for the special nodes, the fractional solutions did not get connected for a long
time.
Our rst idea was to heuristically reduce the size of the problem in some way. Unfortunately,
none of the decomposition techniques described earlier applied, except at the tower of the
ship, where nodes of type 3 are separated by a cut of size 3. We cut out some of the
unnecessary nodes of type 0 in the lower left and right hand corner of the grid, and also
deleted some of the horizontal layers of the grid containing only nodes of type 0.
It is not obvious at all that corners of a grid may be cut out and layers may be deleted
without aecting the optimum objective function value of the problem. We could prove
such a result only for Steiner tree problems (1NCON problems), not for 2NCON or 3NCON
problems. But nevertheless, we used these reductions heuristically to cut down problem sizes
in the hope that some optimal solution of the original graph is still contained in the reduced
graph. For the ship23 problem, the optimal solution of the reduced problem turned out to
be optimal for the nonreduced problem, too.
Figure 7.5 shows the reduced graph of the ship13 problem. The result of the reductions
can be seen from Table 7.5, whose columns list, from left to right, the problem names, and,
for the original ship graph and the reduced ship graphs, the number of nodes of type 0, 1,
2, and 3, the total number of nodes and the total number of edges/number of forced edges.
The forced edges are those edges contained in some cut of size 3 separating two nodes of
type 3, which must be contained in any feasible solution.
An optimal solution for the reduced ship23 problem is shown in Figure 7.6.

Original Graph Reduced Graph

Problem 0 1 2 3 Edges 0 1 2 3 Edges
ship13 461 30 0 3 1096/0 128 28 0 3 325/3
ship23 461 0 30 3 1096/0 249 0 30 3 607/3
ship33 461 0 0 33 1096/0 300 0 0 33 719/9

Table 7.5: Sizes of ship problems

59
Table 7.5 shows that the reductions are enormous, yet there are still many more nodes of
type 0 than nodes of nonzero type in each problem.
When we applied our code to the reduced graphs, the fractional solutions still looked fre-
quently like paths beginning at some special node and ending in some node of type 0. To
cure this problem, we made use of the following type of inequalities.
x((v)\{e}) xe
for all nodes v of type 0 and all e (v). These inequalities (we call them con0 inequalities)
describe algebracially that nodes of type 0 do not have degree 1 in an edge-minimal solution.
This is not true for all survivable networks, but it is true for the optimum solution if all costs
are positive. So, although these inequalities are not valid for the kNCON polytope, we used
them to force the fractional solutions into the creation of longer paths. Another trick that
we used to obtain good starting solutions was to use cuts of a certain structure in the start
LP.
Table 7.6 gives some preliminary computational results of our cutting plane algorithm on
the three reduced and not reduced versions of the ship problem. The entries from left to
right are:

PROBLEM problem name, where red means reduced

VAR number of edges minus number of forced edges
IT number of LPs solved
PART number of partition inequalities added
RCOV number of r-cover inequalities added
LB lower bound = optimal LP value
UB upper bound = heuristic value
GAP (ub-lb)/lb in percent
TIME in minutes:seconds

PROBLEM VAR IT PART RCOV LB UB GAP TIME

ship13 1088 3252 777261 0 211957.1 217428 2.58 10122:35
ship23 1088 15 4090 0 286274 286274 0 27:20
ship33 1082 42 10718 1 461590.6 483052 4.64 55:26
ship13red 322 775 200570 0 217428 217428 0 426:47
ship23red 604 12 2372 0 286274 286274 0 1:54
ship33red 710 40 9817 0 462099.3 483052 4.53 34:52

Table 7.6: Performance of cutting plane algorithm on ship problems

60
Although Table 7.6 shows that the code is still rather slow, it could at least solve two of the
ship problems. In order to obtain faster results, some more research most be done, especially
on nding better starting solutions, devising faster separations heuristics that exploit the
problem structure, and, maybe, inventing new classes of inequalities for high-connectivity
problems.
Table 7.7 shows the percentage of time spent in the dierent routines.

PROBLEM problem name where red means reduced

PT time spent for reduction of problem (in percent)
LPT time spent for LP solving (in percent)
CT time spent for separation (in percent)
MT time on miscellaneous items, input, output, etc (in percent)
TIME total time in minutes:seconds

PROBLEM PT LPT CT MT TIME

ship13 0.0 75.6 23.9 0.5 10122:35
ship23 0.0 13.1 86.4 0.4 27:20
ship33 0.0 31.2 68.2 0.6 55:26
ship13red 0.0 68.5 30.1 1.4 426:47
ship23red 0.1 39.2 58.6 1.9 1:54
ship33red 0.0 41.1 58.4 0.5 34:52

Table 7.7

We do not understand yet why our code solves the ship23 problem rather easily and why
there is still a gap after substantial running time of our cutting plane algorithms for the
ship33 problem. Probably, the small changes of a few survivability requirements result in
more dramatic structural changes of the polyhedra and thus of the inequalities that should
be used. It is conceivable that our code has to be tuned according to dierent survivability
requirements settings. We should mention that we did not attempt to solve ship13 and
ship33 by entering the branching phase of our code. The gaps are not small enough yet for
the enumerative stage to have a decent prospective. Further details of our attempts to solve
network design problems with higher connectivity requirements can be found in Grotschel,
Monma & Stoer (1992c).
Summarizing our computational results, we can say that for survivability problems with
many nodes of type 0 and highly regular cost structure (such as the ship problems) much

61
still remains to be done to speed up our code and enhance the quality of solutions. But for
applications in the area of telephone network design, where problem instances typically are
of moderate size and contain not too many nodes of type 0, our approach produces very good
lower bounds and even optimum solutions in a few minutes. This work is a good basis for the
design of a production code for the 2ECON and 2NCON problems coming up in ber optic
network design and a start towards problems with higher and more varying survivability
requirements and larger underlying graphs.

62
Figure 7.4: Grid graph of the ship problem

63
Figure 7.5: Reduced grid graph of the ship13 problem

64
Figure 7.6: Optimum solution of reduced ship23 problem

65
8 Directed Variants of the General Model
There are many possible variants of the general model described in Section 3 for the design
of networks with connectivity constraints. A natural variant is to consider networks with
directed links. As we will see below, there are practical and theoretical reasons for considering
survivability in directed graphs.

8.1 Survivability Models for Directed Networks

In order to model directed links, we let D = (N, A) denote a directed graph consisting of
a set V of nodes (just as in the undirected case) and a set A of directed arcs. Each arc
a = (u, v) A represents a link directed from node u to node v. For example, this could
model certain communications facilities that allow only the one-way transfer of information.
Of course, there may be arcs directed each way between any given pair of nodes. Each arc
a A has a nonnegative fixed cost ca of establishing the link connection. The directed
graph may have parallel arcs (in each direction). As before, the cost of establishing a network
consisting of a subset B A of arcs is the sum of costs of the individual links contained in B.
The goal is to build a minimum-cost network so that the required survivability conditions
are satised.
The survivability requirements demand that the network satisfy the same types of edge and
node connectivity requirements as in the undirected case. We simply replace the notion of
an undirected path by a directed one. The previous denitions and model formulations are
essentially unchanged.
The problem of designing a survivable directed network has not received as much attention
in the literature as the undirected case. We briey summarize some recent eorts along these
lines.
Dahl (1991) has given various formulations for the directed survivable network design prob-
lem with arc connectivity requirements. He mainly studies bi-Steiner problem which is the
problem of nding a minimum-cost directed subgraph that contains two arc-disjoint paths
from a given root to each node of a set of terminal nodes. This problem has applications in
the design of hierarchical subscriber networks, see Lorentzen & Moseby (1989).
Chopra (1990) found a directed version of the 2ECON problem, which becomes the 2ECON
problem after projection into a lower-dimensional space. He showed that all partition
inequalities and further inequalities can be generated by the projection of certain directed
cut inequalities. The idea for the directed model of the kECON problem relies on a theorem
of Nash-Williams (1960), that says that all undirected graphs can be oriented in such a way,

66
that the resulting directed graph contains, between each i and j, rij /2 arc-disjoint paths,
when the underlying undirected graph contains rij edge-disjoint paths. So Chopras results
can be generalized to higher (even) connectivities.

8.2 Projection
The last remarks show that directed versions of the kECON and kNCON problems are not
only interesting in their own right, but they are sometimes also useful in solving their undi-
rected counterparts. We will illustrate this now by pointing out the value of projections.
For many combinatorial problems, good polyhedral descriptions can be obtained by trans-
ferring the original problem into higher dimensions, that is, by formulating it with additional
(auxiliary) variables, which may later be projected away. This was done successfully for the
2-terminal Steiner tree problem in directed graphs, see Ball et al. (1987). There the formu-
lation with auxiliary variables contains a polynomial number of simple constraints, which by
projection are turned into an exponential number of weird constraints. The general idea
of projection was described by Balas and Pulleyblank (1983).
For the 2ECON problem, Chopra (1990) has found a formulation in directed graphs using
2|E| integer variables and directed cut constraints, which he called the DECON problem,
see (8.1) below. The directed cut constraints (see (8.1)(i)) used in the formulation of the
DECON problem have the advantage that they can be separated in polynomial time, whereas
the separation of the inequalities appearing in our undirected 2ECON problem is NP-hard.
Projection of the directed cut constraints and nonnegativity constraints of the DECON
problem gives a new class of inequalities for the 2ECON problem (we call these Prodon
inequalities) which contain as a subclass the partition inequalities (6.4). For the Steiner tree
problem (where rv {0, 1}), these new inequalities have been found by Prodon (1985). In
the following we show how the Prodon inequalities are derived from the DECON model by
projection.
In order to do this, we must rst introduce some terminology. Let a graph G = (V, E) and
node types rv {0, 1, 2} be given, where at least two nodes are of highest (positive) node
type. This may either be a 2ECON or a 1ECON problem. From G we construct a directed
graph D = (V, A) by replacing each undirected edge ij with two directed edges (i, j) and
(j, i). Furthermore, we pick some node w V of highest node type. Let (W ) be the set of
arcs directed into node set W . If (x, y) is a solution to the following system of inequalities
(where x RE and y RA ),

67
 (i) y( (W )) 1 for all W V, = W = V, with
con(W ) = 2 (or r(W ) = 1 and w  W );
(ii) y(i,j) 0 for all (i, j) A;
(iii) y(i,j) integral for all (i, j) A;
(iv) y(i,j) y(j,i) + xij = 0 for all ij E;
(v) xij 1 for all ij E;

then the integer vector x is feasible for the 2ECON problem, and vice versa: if some integer
vector x is feasible for the 2ECON problem, then an integer vector y can be found so
that (x, y) satised (8.1)(i)(v). So the projection of system (8.1) onto x-variables gives a
formulation of the 2ECON problem. (Originally, Chopra considered this system without the
upper bound constraints.) If no node is of type 2, a feasible vector y is just the incidence
vector of a subgraph of D containing a Steiner tree rooted at w. If all nodes are of type 2,
then y is the incidence vector of a strongly connected directed subgraph of D (strongly
connected means that between each distinct pair s, t of nodes there exists a directed (s, t)-
path and a directed (t, s)-path).
Without the integrality constraints (iii) and upper bound constraints (v), we obtain a relax-
ation, which, after projection onto x-variables, gives a relaxation of the 2ECON problem.
The projection works as follows. Let us dene

F as the set of those W V that appear in the formulation of inequalities (8.1)(i),

bW 0 as the variables assigned to each inequality (8.1)(i) for W F ,

aij R as the variables assigned to each equation (8.1)(iv) for ij E,

s(F ; b; i; j) as the sum of bW over all W F with i W and j  W , and

C as the cone of variables a RE and b := (bW )W F satisfying

aij s(F ; b; i; j), for all ij E and W F

aij s(F ; b; j; i), for all ij E and W F ,
b 0.

If (a, b) C, and if all inequalities of type (8.1)(i) and all inequalities of type (8.1)(iv) are
added with coecients bW and aij respectively, then we obtain an inequality

68
   
u(i,j)y(i,j) + aij xij bW ,
(i,j)A ijE W F

where the u(i,j) are non-positive coecients of the variables y(i,j). In fact, C was dened
exactly in such a way, that the u(i,j) are non-positive. The above inequality is valid for the
system given by all inequalities (8.1)(i), (i), and (iv). Since y 0,

 
aij xij bW ,
ijE W F

is valid for 2ECON(G; r). It can also be proved with the general projection technique of
Balas and Pulleyblank (1983) that

 
(8.2) aij xij bW , for all (a, b) C
ijE W F
x 0

is exactly the projection of system (8.1)(i), (ii) and (iv) onto the x-variables. Not all (a, b)
C are needed in the formulation of (8.2). The following system is clearly sucient to describe
the projection of (8.1) (i), (ii) and (iv) onto x-variables:

 
(8.3) (i) aij xij bW , for all b 0 and
ijE W F aij := max{s(F ; b; i; j), s(F ; b; j; i)}
for all ij E
(ii) x 0.

We call inequalities (8.3)(i) Prodon inequalities (induced by b), because this class of
inequalities was discovered by Prodon (1985) for 1ECON(G; r).
The class of Prodon inequalities properly contains the class of partition inequalities (7.4).
Namely, a partition inequality

p if at least two Wi contain nodes of type 2
x[W1 : . . . : Wp]
p 1 otherwise

69
(where W1 , . . . , Wp is a partition of V into p node sets with r(Wi ) 1) can also be written
as a Prodon inequality, if bW is set to 1 for all Wi that are in F and bW := 0 for all other sets
in F . By denition of F , if at least two sets Wi contain nodes of type 2, then Wi F for
all Wi , and if only one set, say Wp , contains nodes of type 2 (and therefore the root w),
then W1 , . . . , Wp1 are in F , but Wp is not. This explains the diering right-hand sides in
both cases.
But not every facet-dening Prodon inequality is also a partition inequality. For instance,
the inequality depicted in Figure 8.1 is not a partition inequality, but can be written as a
Prodon inequality induced by bW := 1 for the sets {1}, {2}, {5}, {7}, {3, 5, 6}, {4, 6, 7}, and
bW := 0 for all other sets W in F . So the coecients on all depicted edges are 1, and the
right-hand side is 6. Here, nodes 1 and 2 are nodes of type 2; nodes 5 and 7 are nodes of
type 1; all others are of type 0. The Prodon inequality of Figure 8.1 can be proved to be
facet-dening for 2NCON(G; r), where G consists exactly of the depicted nodes and edges.

1 2
@ ;
@ ;
@
;
; @
; @
3 4
 A  A
 A  A
 A  A


5

 A
A 
6
 A

A
7

Figure 8.1: Prodon inequality

We show in the following remark that no Prodon inequality except the cut inequalities are
facet-dening if there are no nodes of type 1.
(8.4) Remark. If (G, r) is an instance of the 2ECON problem, where node types rv only
take values 0 and 2 for all v V , then no Prodon inequalities except the cut constraints
dene facets of 2ECON(G; r).
 
Proof. Let aij xij bW be a Prodon inequality. By denition,
ij W F

1 1
aij s(F ; b; i; j) + s(F ; b; j; i),
2 2

70
 which is the same as 1/2 times the sum of all b W over W F with ij (W ). Therefore,

1 
aT x bW x((W )).
2 W F


Since x((W )) con(W ) = 2 for all W F , this expression is at least bW for all
W F
x 2ECON(G; r). So our Prodon inequality is implied by the sum of some cut inequalities,
and must itself be a cut inequality, if it is to be facet-dening. 2
Inequalities of Prodon type can similarly be dened for kECON problems, if we take as a
basis the formulation of the DECON problem for node types r v {0, 1, 2, 4, 6, . . .}.
We close this section by observing that the separation problem for Prodon inequalities can
be performed in polynomial time. The separation algorithm also makes use of projection
and works in the same way as one iteration of Benders decomposition method, see Benders
(1962). This observation is meant to show that projection is not only of theoretical value
but also of computational interest.
Suppose a point x with 0 x 1 is given, for which it has to be decided, whether there
is a Prodon inequality violated by this point or not. This can be decided by solving the
following LP derived from (8.1):

(8.5) min z
subject to
(i) y( (W )) +z 1 for all W F ;
(ii) y(i,j) 0 for all (i, j) A;
(iii) y(i,j) y(j,i) zxij
= xij for all ij E;
(iv) z 0

This LP has the feasible solution y = 0 and z = 1. If its optimal value is 0, and y is an
optimal solution, then (x , y ) is feasible for the system (8.1), hence x satises all Prodon
inequalities (by the projection result). If the optimal value is non-zero, then the optimal
dual variables bW for inequalities (i) and aij for equations (iii) dene a Prodon inequality
violated by x . More explicitly, the optimal dual variables bW (W F ) and a RE satisfy

(8.6) + bW 0
aij for all ij E and all W separating i and j

a x +
T
bW > 0.
wF

71
The rst inequality implies that aij is at least the maximum of s(F ; b; i; j) and s(F ; b; j; i)
for each ij E. The optimality of a and b implies that a ij is exactly the maximum of these
two numbers. By multiplying a and b with some number, b can be made integral without
changing (8.6). So b induces a Prodon inequality violated by x.
The LP (8.5) can be solved in polynomial time, since there exist polynomial separation
algorithms for the directed cut inequalities (8.5)(i). Therefore, the Prodon inequalities can
also be separated in polynomial time. We have, however, not made use of these inequalities
yet.

72
 REFERENCES

1. A. Agrawal, Ph. Klein and R. Ravi (1991), When trees collide: An approximation
algorithm for the generalized Steiner tree problem on networks, in the Proceedings of
the Symposium of the Theory of Computing 1991.
2. E.Balas and W.R. Pulleyblank (1983), The perfect matchable subgraph polytope of
a bipartite graph, Networks 13, 495-516.
3. M.O. Ball, W.G. Liu and W.R. Pulleyblank (1987), Two-terminal Steiner tree poly-
hedra, Technical Report 87466-OR, University of Bonn.
4. Bellcore (1988), FIBER OPTIONS: Software for designing survivable optical ber
networks, Software Package.
5. J. F. Benders (1962), Partitioning procedures for solving mixed-variable programming
problems, Numerische Mathematik 4, 238-252.
6. D. Bienstock, E.F. Brickell and C.L. Monma (1990), On the structure of minimum-
weight k-connected spanning networks, SIAM J. Disc. Math. 3, 320-329.
7. R. E. Bixby (1991), Implementing the simplex method: The initial basis, Technical
Report TR 90-32, Department of Mathematical Sciences, Rice University, Houston,
Texas.
8. G-R. Cai and Y-G. Sun (1989), The minimum augmentation of any graph to a k-edge
connected graph, Networks 19, 151-172.
9. R.H. Cardwell, C.L. Monma and T.H. Wu (1989), Computer-aided design procedures
for survivable ber optic networks, IEEE J. Selected Areas of Communications 7,
1188-1197.
10. R.H. Cardwell, T.H. Wu and W.E. Woodall (1988), Decreasing survivable network
cost using optical switches, Proc. GLOBECOM 88.
11. S. Chopra (1990), Polyhedra of the equivalent subgraph problem and some edge con-
nectivity problems, J.L. Kellogg Graduate School of Management, Northwestern Uni-
versity, Evanston, Illinois.
12. S. Chopra (1991), The k-edge connected spanning subgraph polyhedron, J.L. Kel-
logg Graduate School of Management, Northwestern University, Evanston, Illinois.
13. W. Chou and H. Frank (1970), Survivable communication networks and the terminal
capacity matrix, IEEE Transactions on Circuit Theory, CT-17 (2) 192-197.

73
14. V. Chvatal (1973), Edmonds polytopes and a hierarchy of combinatorial problems,
Discrete Mathematics 4, 305-337.

15. G. Cornuejols, J. Fonlupt and D. Naddef (1985), The traveling salesman problem on
a graph and some related integer polyhedra, Mathematical Programming 33, 1-27.

16. G. Dahl (1991), Contributions to the design of survivable directed networks, Ph.D.
Thesis, University of Oslo, Technical Report TF R 48/91, Norwegian Telecom, Re-
search Dept., Kjeller, Norway.

17. J. Edmonds (1965), Maximum matching and a polyhedron with 0,1-vertices, Journal
of Research of the National Bureau of Standards B 69, 125-130.

18. K.P. Eswaran and R.E. Tarjan (1976), Augmentation problems, SIAM Journal on
Computing 5, (4) 653-665.

19. A. Frank (1990), Augmenting graphs to meet edge-connectivity requirements, sub-

mitted to SIAM Journal on Discrete Mathematics.

20. A. Frank (1992), Connectivity and network ows, Chapter 2 of the Handbook of
Combinatorics, R. Graham, M. Grotschel, L. Lovasz (eds.), North Holland, Amster-
dam.

21. H. Frank and W. Chou (1970), Connectivity considerations in the design of survivable
networks, IEEE Transactions on Circuit Theory, CT-17 (4) 486-490.

22. G.N. Frederickson and J. JaJa (1982), On the relationship between the biconnectivity
augmentation and traveling salesman problem, Theoretical Computer Science 19, 189-
201.

23. M.R. Garey and D.S. Johnson (1979), Computers and Intractability: A Guide to the
theory of NP-completeness, Freeman, San Francisco.

24. M.X. Goemans and D.J. Bertsimas (1990), Survivable networks, linear programming
relaxations and the parsimonious property, OR Center Report OR216-90, MIT, to
appear in Mathematical Programming.

25. M. X. Goemans (1991a), Arborescence polytopes for series-parallel graphs, to appear

in Discrete Applied Mathematics.

26. M. X. Goemans (1991b), The Steiner tree polytope and related polyhedra, Depart-
ment of Mathematics, MIT, Cambridge.

74
27. R.E. Gomory and T.C. Hu (1961), Multi-terminal network ows, Journal of the
Society for Industrial and Applied Mathematics 9, 551-570.

28. M. Grotschel, L. Lovasz and A. Schrijver (1988), Geometric Algorithms and Combi-
natorial Optimization, Springer, Berlin.

29. M. Grotschel and C.L. Monma (1990), Integer polyhedra associated with certain
network design problems with connectivity constraints, SIAM Journal on Discrete
Mathematics 3, 502-523.

30. M. Grotschel, C.L. Monma and M. Stoer (1992a), Facets for polyhedra arising in the
design of communication networks with low-connectivity constraints, SIAM Journal
on Optimization 2, 474-504.

31. M. Grotschel, C.L. Monma and M. Stoer (1992b), Computational results with a
cutting plane algorithm for designing communication networks with low-connectivity
constraints, Operations Research 40, 309-330.

32. M. Grotschel, C.L. Monma and M. Stoer (1992c), Polyhedral and computational
investigations for designing communication networks with high survivability require-
ments, ZIB-Preprint SC 92-24, Konrad-Zuse-Zentrum fur Informationstechnik Berlin,
submitted to Operations Research.

33. M. Grotschel and M.W. Padberg (1985), Polyhedral theory, in: E.L. Lawler, J.K.
Lenstra, A.H.G. Rinnooy Kan & D. Shmoys (eds.), The traveling salesman problem,
Wiley, Chichester, 251-305.

34. J. Hao and J. B. Orlin (1992) A faster algorithm for nding the minimum cut in a
graph, Proceedings of the Third Annual ACM-SIAM-Symposium on Discrete Algo-
rithms, Orlando, Florida, 165-174.

35. F. Harary (1962), The maximum connectivity of a graph, Proceedings of the National
Academy of Sciences, USA 48, 1142-1146.

36. S. Khuller and U. Vishkin (1992), Biconnectivity approximations and graph carvings,
to appear in STOC 1992.

37. C-W. Ko and C.L. Monma (1989), Heuristic methods for designing highly survivable
communication networks, Technical report, Bellcore.

38. D.J. Kolar and T.H. Wu (1988), A study of survivability versus cost for several ber
network architectures, Proc. ICC 88, 61-66.

75
39. E. Lawler (1976) Combinatorial optimization: networks and matroids, Holt, Rein-
hart and Winston, 1976.

40. E.L. Lawler, J.K. Lenstra, A.H.G. Rinnooy Kan & D. Shmoys (1985), The traveling
salesman problem, Wiley, Chichester, 1985.

41. F. Ling and T. Kameda (1987), Complexity of graph connectivity functions, Tech-
nical Report, School of Computing Science, Simon Fraser University, Burnaby, British
Columbia, Canada.

42. R. Lorentzen and H. Moseby (1989), Mathematical models and algorithms used in
the subscriber network planning tool ABONETT, Norwegian Telecommunications
Research Dept., TF-report 66/89.

43. L. Lovasz (1976), On some connectivity properties of Eulerian graphs, Acta Math.
Acad. Scient. Hung. 28, 129-138.

44. L. Lovasz and M.D. Plummer (1986), Matching Theory, Annals of Discrete Mathe-
matics 29.

45. W. Mader (1978), A reduction method for edge-connectivity in graphs, Ann. Dis-
crete Math. 3, 145-164.

46. A.R. Mahjoub (1988), Two edge connected spanning subgraphs and polyhedra, Uni-
versity of Bonn, Technical Report No. 88520-OR.

47. C.L. Monma, B.S. Munson and W.R. Pulleyblank (1990), Minimum-weight two-
connected spanning networks, Mathematical Programming 46, 153-171.

48. C.L. Monma and D.F. Shallcross (1989), Methods for designing communication net-
works with certain two-connected survivability constraints, Operations Research 37,
531-541.

49. D. Naor, D. Guseld and C. Martel (1990), A fast algorithm for optimally increasing
the edge-connectivity, Proceedings of the Foundation of Computer Science 90, 698-
707.

50. C. St. J. A. Nash-Williams (1960), On orientations, connectivity, and odd vertex

pairings in nite graphs, Canadian Journal of Mathematics 12, 555-567.

51. Newark Star Ledger (1987), Damage to ber cable hinders phone service, Newark
Star Ledger (September 22, 1987).

76
52. Newark Star Ledger (1988a), Cable snaps, snags area phone calls, Newark Star
Ledger (February 26, 1988).
53. Newark Star Ledger (1988b), Phone snafu isolates New Jersey; long-distance cable
snaps, Newark Star Ledger (November 19, 1988).
54. New York Times (1988), Phone system feared vulnerable to wider disruptions of
service, New York Times (May 26, 1988).
55. New York Times (1989), Experts say phone system is vulnerable to terrorists, New
York Times (February 8, 1989).
56. M.W. Padberg and M.R. Rao (1982), Odd minimum cut sets and b-matchings, Math-
ematics of Operations Research 7, 67-80.
57. C.H. Papadimitriou and K. Steiglitz (1982), Combinatorial optimization: algorithms
and complexity, Prentice-Hall, Englewood Clis, NJ.
58. C.H. Papadimitriou, M. Yannakakis (1982), The complexity of facets and some facets
of complexity J. Assoc. Comp. Mach. 29, 285-309.
59. A. Prodon, Liebling and Groin (1985) Steiners problem on two-trees, Technical
Report RO-830315, Ecole Polytechnique Federale de Lausanne, Switzerland.

60. A. Prodon (1985), A polyhedron for Steiner trees in series-parallel graphs, Technical
Report, Ecole Polytechnique Federale de Lausanne, Switzerland.
61. W.R. Pulleyblank (1990), Polyhedral combinatorics, in Nemhauser, Rinnoy Kan,
Todd (eds.), Optimization Handbook in Operations Research and Management Sci-
ence, 1991.
62. A. Rosenthal and A. Goldner (1977), Smallest augmentation to biconnect a graph,
SIAM Journal of Computing, 6(1) 55-66.
63. A. Schrijver (1986), Theory of linear and integer programming, Wiley, New York.
64. K. Steiglitz, P. Weiner and D.J. Kleitman (1969), The design of minimum cost sur-
vivable networks, IEEE Transaction on Circuit Theory 16, 455-460.
65. M. Stoer (1992), Design of survivable networks, Ph.D. thesis, University of Augsburg,
Lecture Notes in Mathematics, 1531, Springer, Heidelberg.
66. K. Takamizawa, T. Nishizeki, and N. Saito (1982), Linear-time computability of com-
binatorial problems on series-parallel graphs, Journal of the Association for Comput-
ing Machinery 29, (3) 623-641.

77
67. S. Ueno, Y. Kajitani, and H. Wada (1988), Minimum augmentation of a tree to a
k-edge-connected graph, Networks 18, 19-25.

68. J.A. Wald and C.J. Colbourn (1983), Steiner trees, partial 2-trees and minimum IFI
networks, Networks 13, 159-167.

69. Wall Street Journal (1988), Fire in ber gateway sparks ight delays, problems at
brokerages, Wall Street Journal, (May 11, 1988).

70. T. Watanabe and A. Nakamura (1987), Edge-connectivity augmentation problems,

Computer and System Sciences 35, (1) 96-144.

71. P. Winter (1985a), Generalized Steiner tree problem in Halin networks, In Proc.
12th Intern. Symp. on Mathematical Programming, MIT.

72. P. Winter (1985b), Generalized Steiner problem in outerplanar networks, BIT 25,
485-496.

73. P. Winter (1986), Generalized Steiner problem in series-parallel networks, Journal

of Algorithms 7, 549-566.

74. P. Winter (1987), Steiner problem in networks: A survey, Networks 17, (2) 129-167.

75. T.H. Wu and R.H. Cardwell (1988a), Optimum routing in ber network design: mod-
els and applications, Proc. ICC 88.

76. T.H. Wu, D.J. Kolar and R.H. Cardwell (1988b), Survivable network architectures for
broadband ber optic networks: model and performance comparison, IEEE Journal
Lightwave Technology 6, 1698-1709.

77. G. Zorpette (1989), Keeping the phone lines open, IEEE Spectrum (June 1989) 32-36.

78