0% found this document useful (0 votes)
95 views51 pages

Using Link Cost As A Metric

Link state routing uses Dijkstra's algorithm to calculate the shortest path to all destinations based on the network topology. Each router broadcasts link state packets (LSPs) containing neighboring router addresses and link costs. When a router receives LSPs from all other routers, it runs Dijkstra's algorithm to build forwarding tables mapping destinations to next hops with minimum cost paths. If the topology changes, such as new links or cost changes, routers recalculate routes by running Dijkstra's algorithm on the updated link state database.

Uploaded by

poojaq
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
95 views51 pages

Using Link Cost As A Metric

Link state routing uses Dijkstra's algorithm to calculate the shortest path to all destinations based on the network topology. Each router broadcasts link state packets (LSPs) containing neighboring router addresses and link costs. When a router receives LSPs from all other routers, it runs Dijkstra's algorithm to build forwarding tables mapping destinations to next hops with minimum cost paths. If the topology changes, such as new links or cost changes, routers recalculate routes by running Dijkstra's algorithm on the updated link state database.

Uploaded by

poojaq
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
You are on page 1/ 51

Using Link Cost as a Metric

Link State Routing


Link State Routing
 Also called shortest path first (SPF)
forwarding
 Named after Dijkstra’s algorithm (1959)
which it uses to compute routes
 All routers have tables which contain
a representation of the entire network
topology
 In the form of lists of routers and
information about each router’s
neighbours and the connection between
the two
Link State Routing
 Each router creates a link state packet
(LSP) which contains names (e.g. network
addresses) and cost to each of its
neighbours
 The LSP is transmitted to all other routers, who
each update their own records
 When a routers receives LSPs from all routers,
it can use (collectively) that information to
make topology-level decisions
Link State Packets
 LSPs are generated and distributed
when:
 A time period passes
 New neighbours connect to the router
 The link cost of a neighbour has
changed
 A link to a neighbour has failed (link
failure)
 A neighbour has failed (node failure)
Link State Packets
 LSP are essentially a list of tuples,
containing:
 The name of a neighbour to a router
 Which may be a router or a network
 The cost of the link to that neighbour
Link State Packets
 Distribution of LSPs can be difficult
 Routers themselves are the means for
delivering messages
 How do routers deliver their own
messages, particularly when routers are
in an inconsistent state
 e.g.
During link failure, before each router
has been notified of the problem
Link State Packets
 One method for LSP distribution: Flooding
 Each LSP received is transmitted to every direct
neighbour (except the neighbour where the
LSP came from)
 This creates an exponential number of packets
on the network (similar to O(2R), where R is the
number of routers)
 It does, however, guarantee that the LSP will
be received by every router
 Assuming that node or link failure does not occur,
and LSPs are not somehow lost
Link State Packets
 An improvement on this scheme is as follows:
 When an LSP is received, it is compared with the stored
copy
 If it is identical to the stored copy, it is dropped
 If it is different, the stored LSP is overwritten with the new
LSP and the LSP is transmitted to every direct neighbour
(except the source of the LSP)
 This scheme works because if a given router has
already received a LSP from another neighbour, it will
have also already distributed the LSP to all of its
neighbours
 This scheme has a network complexity similar to O(R 2)
Link State Routing Algorithm
 Ok, now that we know how to
distribute LSPs, how are they used to
determine routes?
 The algorithm used (mostly) was
developed by Dijkstra
 Essentially, the algorithm runs at each
router, computing each possible path to
the destination, adding up each cost
 The path with the lowest cost is used
Link State Routing Algorithm
 The algorithm requires the following information:
 Link state database: List of all the latest LSPs from each router on the network
 Path: Tree structure storing previously computed best paths
 Consider this a sort of cache
 Data type for nodes: (ID, path cost, port)
 Tent: Tree structure storing paths currently being tested and compared
(tentative)
 Consider this a sort of rough workspace
 Data type for nodes: (ID, path cost, port)
 Forwarding database: Table storing all IDs that can be reached, and the port
to which messages should be sent
 This is simply a reduced version of the ‘Path’, which contains (destination,port) pairs
 This can be used by the router to quickly forward packets for which the best path
has already been determined
 Data type for table rows: (ID, port)
Dijkstra’s LSR Algorithm
 Initially, PATH is just a root containing (this
router’s ID, 0, 0)
 For every node placed into path, N:
 For all neighbours M of node N:
 If M is not in TENT, add a node to TENT for M (use
the LSP for N to determine link cost)
 If M is in TENT already, and its cost is lower than an
existing entry for M, replace that entry with
information from N’s LSP
 If M is in TENT already, but its cost is higher, ignore
N’s link to M
 Calculate the shortest route in TENT
 If the shortest route has lower cost than the route in
PATH, overwrite the route in PATH with the route in
TENT
Dijkstra’s LSR Algorithm
 Consider the following network:

6 2
A B C 5
2 1 2 G
2 4
D E F 1

Link state database:


A B C D E F G
B 6 A 6 B 6 A 2 B 1 C 2 C 5
D 2 C 2 F 2 E 2 D 2 E 4 F 1
E 1 G 5 F 4 G 1
Dijkstra’s LSR Algorithm
 Now, if we want to generate a PATH for C:
 First, we add (C,0,0) to PATH

C (0)
Dijkstra’s LSR Algorithm
 Examine C’s LSP
 Add F, G, and B to TENT

C (0)

(2) F G (5) B (2)


Dijkstra’s LSR Algorithm
 Place F in PATH (shown as solid line)
 Add G and E to TENT (adding costs)

C (0)

(2) F G (5) B (2)

(3) G (6)
E
Dijkstra’s LSR Algorithm
 G exists in TENT twice, keep only the best
 The new G is a better path than the old (3 < 5)

C (0)

(2) F G (5) B (2)

(3) G (6)
E
Dijkstra’s LSR Algorithm
 Put B into path (shown as solid line)
 Add A and E to TENT

C (0)

(2) F B (2)

(3) G (6) (3) A (8)


E E
Dijkstra’s LSR Algorithm
 E exists in TENT twice, keep only the best
 The new E is better than the old (3 < 6)

C (0)

(2) F B (2)

(3) G (6) (3) A (8)


E E
Dijkstra’s LSR Algorithm
 Place E in PATH (shown as solid line)
 Add D to TENT

C (0)

(2) F B (2)

(3) G (3) A (8)


E

(5)
D
Dijkstra’s LSR Algorithm
 Place G in PATH (shown as solid line)
 All G’s LSP elements already exist in TENT

C (0)

(2) F B (2)

(3) G (3) A (8)


E

(5)
D
Dijkstra’s LSR Algorithm
 Place D in PATH (shown as solid line)
 Add path to A since it is better than old A

C (0)

(2) F B (2)

(3) G (3) A (8)


E

(5)
D

A (7)
Dijkstra’s LSR Algorithm
 Place A in PATH (shown as solid line)
 All A’s LSP elements already exist in PATH

C (0)

(2) F B (2)

(3) G (3)
E

(5)
D

A (7)
Dijkstra’s LSR Algorithm
 We are done since all routes from TENT
were placed into PATH

C (0)

(2) F B (2)

(3) G (3)
E

(5)
D

A (7)
Dijkstra’s LSR Algorithm
 We can now create a forwarding
database:
Forwarding Database
C (0) Destination Port

(2) F C C
B (2)
F F
(3) G (3) G F
E
B B
(5) E B
D
D B
A (7) A B
LSR Topology Changes
 LSR forwarding tables must be recalculated
whenever a topology change occurs
 For example, a new router and/or link is added
to the network
 This new link may provide a more efficient route to
one or more other nodes
 For example, a given link’s cost is reduced
 This new link may now provide the lowest total cost
route to a destination that was previously forwarded
in another direction
 For example, a given link’s cost is increased
 This new link may no longer provide the lowest total
cost route to a given destination, and another route
should now be chosen
LSR Topology Changes
 In a nutshell, LSR routers should
invalidate (indicate that it needs to be
regenerated) its PATH data structure,
and thus its forwarding table
 The entire PATH generation algorithm
(e.g. Dijkstra’s algorithm) should be
reapplied
Topology Change Example
 Let’s consider our previously generated
PATH structure for the router C

C (0)

(2) F B (2)

(3) G (3)
E

(5)
D

A (7)
Topology Change Example
 Say we receive an LSP from router B,
indicating the link cost from B to E is
now 6
C (0)

(2) F B (2)

(3) G (3)
E

(5)
D

A (7)
Topology Change Example
 The total route costs are different in
PATH:

C (0)

(2) F B (2)

(3) G (8)
E

(10)
D

A (12)
Topology Change Example
 Consider for now, only the cost to A

C (0)

(2) F B (2)

(3) G (8)
E

(10)
D

A (12)
Topology Change Example
 Recall that another path to A existed
 Now, that path is more efficient

C (0)

(2) F B (2)

(3) G (8)
E A (8)

(10)
D

A (12)
Topology Change Example
 The PATH data structure is complete,
the forwarding table can now be
regenerated
C (0)

(2) F B (2)

(3) G (8)
E A (8)

(10)
D
Topology Change Example

 In a router, which will be running


as a computer program, finding if
a new path exists essentially
requires complete re-execution of
Dijkstra’s algorithm
 For example, there could have
been many routes to A, each of
which would have to be compared
to find the most efficient route
Open Shortest Path First
OSPF
OSPF
 Open SPF protocol
 SPF: Shortest path first
 Essentially the OSPF specification is one specification describing an
algorithm implementing shortest path forwarding
 It is open, meaning anyone can implement the specification
at no cost
 Since OSPF is a link state routing (LSR) protocol, it
can have load balancing
 However, to be deployed on large-scale WANs, the LSP
propagation must be limited
 OSPF partitions networks into regions called ‘areas’ which contain a
subset of the routers
 This is similar to schemes used in other LSR protocols
OSPF Autonomous Systems
 An OSPF AS allows a multi-level
routing strategy to be employed
 The complete system is partitioned
into autonomous systems (AS)
 Within each AS, OSPF routing is used
 Inthis way, the limitation of the number of
routers in link state routing is not a factor
 Messaging between AS can use ATM
OSPF Messages
 There are 5 types of messages in OSPF:
 Hello messages
 Allow routers to test if a node is reachable
 Link State Advertisement (LSA)
 Topology information from a router (i.e. LSPs)
 Link status request (LSR)
 Requests send to another router to determine the
status of one or more links
 Link status update (LSU)
 Responses to a link status request message
 Link status acknowledgement
 Used to indicate that the LSU was received (reliable
transfer)
OSPF Hello Packets
 When a router wants to test if a node is
reachable, it sends a Hello packet
 If the node is reachable, it will respond
with its own Hello packet
 Hello packets contain a list of reachable
addresses (among other things)
 A router might query the node about one of
those addresses
 The node will respond with topology
(connectivity) information about the host at
that address
 In the form of an LSA
OSPF LSA Packets
 When requested, these packets
contain a list of links (forming a
complete route) to a destination
 This can be used (including the link
cost information) to determine the
‘shortest’ path
OSPF Link Status Requests
 This packet represents a request for
information about one or more links
 A router may be given this request if
another router has outdated
information
OSPF Link Status Updates
 This is a response to a link status
request
 It contains information about each
link requested
 Most importantly: the length (cost) of
the link
Multicast Routing in OSPF
 In OSPF and other link state
implementations, multicast routing is
supported
 For a multicast message, Dijkstra’s path
tree is also created
 However, in this case, the sender is the root
of the tree, not the current router
 The current router sends the message to all
of its direct child nodes in the tree
PNNI Switches
 Used in ATM switches
 PNNI is a link state algorithm used for ATM
switching
 PNNI is multi-level routing
 Areas are called peer groups
 Peer groups can be arbitrarily connected
 Despite the fact that ATM networks are
connection-oriented, the process of finding
a route for packets/cells is the same
 Thus PNNI switching is very similar to OSPF
and IS-IS routing
LSR and Load Balancing
 At least one scheme for load
balancing is possible with LSR:
 Multiple routes could be calculated for
each destination
 As a result, the forwarding table would
contain more than one entry for nodes
 This technique is known as load splitting
LSR and Load Splitting
 Since the forwarding table contains more than
one entry for each destination, some scheme
must be used to choose one of the entries for
each incoming packet
 Each entry could have its turn (in a round-robin
fashion)
 This evenly distributes packets among the different
routes
 Randomly choose one of the forwarding entries
 Use one of the above schemes, but use some
preference method to ensure the entry which has
more congestion on its link is used less often
 Have routers keep an on-going database of link
congestion, allowing it to choose the least
congested link for each packet (which may change
from one packet to another)
LSR and Load Splitting
 Advantages
 Distributing network load among different routes
directly results in more optimal use of network
bandwidth
 i.e. Network capacity is increased
 Disadvantages
 The number of out-of-order packets is increased
when load splitting is employed
 Network delivery times are difficult to estimate
 As is frequently important with streaming data (such as
streaming audio) where Quality of Service (QoS) is
important
LSR and Load Balancing
 LSR provides another, more natural,
scheme for load balancing
 When links become too saturated
with packets, alternate routes can be
used
 For example, a link could be assigned a
higher cost due to the amount of traffic
 The higher cost could ensure that some
of the routers use alternate routes for
forwarding
LSR and Load Balancing
 The other scheme, varying the link
cost as network traffic changes, is
another interesting idea
 Increasing the link cost as traffic through
that link increases allows automatic load
balancing to occur
 Similarly, link cost can be lowered as
traffic through the link decreases
 This technique is called ‘variable link
cost load balancing’
Variable Link Cost LB
 Advantages
 The automatic load balancing that results would offer
similar improvement in network capacity as when explicitly
employing load splitting
 Disadvantages
 Having link cost change with network congestion results in
a need for LSP propagation whenever a significant change
in network congestion occurs
 As a result, there are more LSPs required than when only network
topology changes should force them
 Network topology changes happen very rarely, but network
congestion varies often
 By the time a link cost change is propagated to all routers
in the network, the network congestion may have (again)
changed
LSR vs. DVR
 Bandwidth used by each:
 This is dependent upon network topology
 Some networks use less bandwidth for LSR than DVR (and vice
versa)
 Computation used by each:
 LSR (Dijkstra): O(n*k*log n)
 n: number of nodes on the network
 k: average number of links per node
 Therefore, n*k is the total number of links
 DVR: O(n * k)
 However, sometimes the list of distance vectors (n*k of them) must be
scanned more than once
 It should be fairly obvious that LSR uses more computation than
DVR
Problems w/ LSR & DVR
 Propagation of routing information
(distance vectors in DVR, LSPs in LSR) is
costly in terms of network bandwidth
 As the number of routers increases, the inter-
router communication increases rapidly
 e.g. LSR with reduced flooding: O(N2)
 By the time we reach a network the size of the
Internet, the inter-router traffic uses a very
large proportion of the total network bandwidth
 Multi-level routing can be used to solve this
problem
 This is beyond the scope of this course

You might also like