COMPNW 4771

No. of Pages 13, Model 3G

9 May 2012
Computer Networks xxx (2012) xxxxxx
1

Contents lists available at SciVerse ScienceDirect

Computer Networks
journal homepage: www.elsevier.com/locate/comnet

Trading availability among shared-protected dynamic connections in WDM

networks

2
3
4

Q1

5
6
7

Q2

Diego Lucerna a, Massimo Tornatore b,, Biswanath Mukherjee c, Achille Pattavina b

a

Huawei Technologies Italia, Italy

Department of Electronics and Information, Politecnico di Milano, Via Ponzio 34-35, 20121 Milan, Italy
c
Department of Computer Science, University of California, Davis, CA 95616, USA
b

a r t i c l e

1 4
0
2
11
12
13
14
15

i n f o

Article history:
Received 14 October 2011
Received in revised form 17 April 2012
Accepted 18 April 2012
Available online xxxx

16
17
18
19
20
21
22
23

Keywords:
Optical network
WDM
Elastic availability
Holding time
Shared protection
SLA Violation Risk

a b s t r a c t
Novel automatized management systems for optical WDM networks promise to allow customers asking for a connection (i.e., a bandwidth service) to specify on-demand the terms
of the Service Level Agreement (SLA) to be guaranteed by the Network Operator (NO). In
this work, we exploit the knowledge, among the other Service Level Specications (SLS),
of the holding time and of the availability target of the connections to operate shared-path
protection in a more effective manner.
In the proposed approach, for each connection we monitor the actual downtime experienced by the connection, and, when the network state changes (typically, for a fault occurrence, or a connection departure or arrival), we estimate a new updated availability target
for each connection based on our knowledge of all the predictable network-state changes,
i.e., the future connection departures. Since some of the connections will be ahead of the
stipulated availability target in their SLA (credit), while other connections will be behind
their availability target (debit), we propose a mechanism that allows us to trade availability credits and debits, by increasing or decreasing the shareability level of the
backup capacity. Our approach permits to exibly manage the availability provided to living connections during their holding times.
The quality of the provided service is evaluated in terms of availability as well as probability of violation of availability target stipulated in the SLA (also called SLA Violation Risk),
a recently-proposed metric that has been demonstrated to guarantee higher customer satisfaction than the classical statistical availability. For a typical wavelength-convertible US
nationwide network, our approach obtains signicative savings on Blocking Probability
(BP), while reducing the penalties due to SLA violations. We also analytically demonstrate
that proposed scheme can be highly benecial if the monitored metric is the SLA Violation
Risk instead of the availability.
 2012 Published by Elsevier B.V.

25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50

51
52

1. Introduction
In optical WDM networks survivability mechanisms are
needed to avoid that a failure of a network element may
cause signicant losses of revenue for those customers that

53
54
55

Q3
Q1

Corresponding author.
E-mail addresses: [email protected] (D. Lucerna), tornator@
elet.polimi.it (M. Tornatore), [email protected] (B. Mukherjee),
[email protected] (A. Pattavina).

run their service over the bandwidth provided by means of

the optical paths. These revenue losses are then reclaimed
in the form of penalties to be paid by the Network Operator
(NO). Different protection mechanisms to ensure survivability in WDM networks have been proposed [1]: among
them, Shared-Path Protection (SPP) is one of the mostadopted options, because of its desirable resource efciency [2].
Recently, many new applications are emerging with
requirements of large bandwidth over relatively short

1389-1286/$ - see front matter  2012 Published by Elsevier B.V.

http://dx.doi.org/10.1016/j.comnet.2012.04.021

Q1 Please cite this article in press as: D. Lucerna et al., Trading availability among shared-protected dynamic connections in WDM networks,
Comput. Netw. (2012), http://dx.doi.org/10.1016/j.comnet.2012.04.021

56
57
58
59
60
61
62
63
64
65

COMPNW 4771

No. of Pages 13, Model 3G

9 May 2012
Q1
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126

D. Lucerna et al. / Computer Networks xxx (2012) xxxxxx

and predictable periods of time: let us consider, e.g., video

distribution of important sport or social events, or the massive data transfer for backup, storage or e-science purposes. Network technology and the bandwidth market
are developing to provide the exible platform the new
applications are asking for. In particular, new architectures
and routines for user-controlled on-demand optical circuit
provisioning [3], typically based on automatic or webbased interfaces at the management plane (MP) [1], will
enable the on-line specication of the Service Level Agreement (SLA) terms to be guaranteed (with different price
range) by the Network Operator (NO). In other words,
users may be able to specify the QoS terms [3] of their connection requests, e.g., the availability target (AT) or the
holding-time.
In particular, the NO should be able to guarantee with a
high probability that the stipulated AT is respected, in order
to avoid penalties; at the same time, the NO aims at
increasing its prot, i.e., it wants to maximize the number
of connections (or bandwidth) provisioned. But, in case of
SPP and dynamic trafc, the NO must carefully monitor
the actual availability provided to the customer. In fact,
whenever a new SPP connection is routed, the NO must
not only verify if the AT of the incoming connection is satised, but also it must check if the ATs of the existing connections are still respected despite the increased sharing of
backup resources. A typical solution to avoid penalties is to
employ an availability-guaranteed provisioning approach
[4,2]: in this case, the NO provisions a connection only if
the network can provide a path with a long-term theoretical availability (that can a priori evaluated) that is equal or
larger than the AT target.
In this work, we present a novel availability-guaranteed
provisioning method that dynamically manages the availability provided to SPP-protected connections during their
holding time. Our approach (i) leverages on the information about connection departures (future departures imply
decreased sharing and in turn more availability) and (ii) is
able to dynamically trade availability from connections
ahead of their AT to connections behind their AT. Note that
the proposed scheme do not reprovision backup resources
(and, consequently, it avoids the additional control overhead required by backup reprovisioning), but it only operates on the sharing of the backup resources.
When a new connection has to be routed, the target
availability of each living connection in the network is
re-estimated by considering the experienced downtime
and the remaining holding time: e.g., if a connection has
not been affected by failures, the NO can be considered
in credit of availability with respect to the customer
and the customers availability requirements can be opportunely decreased by the NO, as long as the original target
AT is still respected. On the other hand, if a connection
has undergone an outage period and it is getting close to
its maximal acceptable downtime, then it can be considered in debit of availability with respect to the NO, and
its current availability requirement should be increased
by the NO in the attempt to match the original stipulated
AT. In brief, the proposed approach dynamically transfers
availability from connections that have an availability debit to connections in credit of availability by allowing more

or less sharing of backup resources and it helps the NO to

meet connections SLAs. Although the proposed methodology can be applied to any SPP algorithm, in the following
we show the effectiveness of our approach through simulative experiments using the Availability-Guaranteed Provisioning algorithm (AGP) presented in [2]. Since our
proposed method requires the knowledge of the connection holding-time we will refer to it as Holding-TimeAware (HTA) method.
Furthermore, recent studies [57] have shown that
using the theoretical long-term availability to evaluate
the quality of a connection provided over a short period
of time (e.g., a period comparable with the average failure
interval) may not be enough to characterize the quality of
provisioning in terms of SLA satisfaction (more details will
be given in Section 2). So, in this paper, we also analytically
discuss how to extend the proposed approach to the case
where the monitored metric is the SLA Violation Risk
(i.e., the probability that, given a certain availability target
and certain statistical availability associated to the path,
the offered connection does satisfy the availability target)
when our trading-availability method is adopted. Some
preliminary results are also provided.
The rest of the paper is organized as follow. Section 2
overviews some background work on availability-guaranteed SPP strategies. In Section 3 we formally state the
availability-guaranteed SPP problem and we briey describe an existing solution for availability-guaranteed
SPP, called AGP. In Section 4 our new HTA method to
dynamically trade availability among SPP connections is
presented. Section 5 discusses how to extend the HTA approach to include the new SLA Violation Risk metric. In
Section 6 we compare and evaluate by means of simulations our HTA methodology vs. the basic AGP approach.
In Section 7, we draw the conclusion. In the Appendix A,
a rigorous approach to evaluate the availability in a SPP
network scenario is presented.

127

2. Prior work

164

This paper provides novel contributions on two complementary, bus distinct, lines of research in the eld of
shared-path protection: (1) how to route of availabilityguaranteed shared-path-protected connections and (2)
how to evaluate the SLA Violation Risk, or interval availability, for availability-guaranteed SPP routing.

165

2.1. Availability-guaranteed SPP

171

We start considering the problem of dynamic routing

with Shared-Path Protection (SPP). To enable dynamic provisioning of SPP connections, a network-control element
(say, e.g., the Path Computational Element, PCE) needs to
compute two link-disjoint paths, a dedicated working path
and a shared backup path, for each incoming connection
request, based on the current network state. SPP routing
algorithms are usually based on two-step approaches
(e.g., [8]), which compute separately the working and the
backup path, using shortest-path or K-shortest-path algorithms [9] that minimize the total link costs. Generally, link

172

Q1 Please cite this article in press as: D. Lucerna et al., Trading availability among shared-protected dynamic connections in WDM networks,
Comput. Netw. (2012), http://dx.doi.org/10.1016/j.comnet.2012.04.021

128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163

166
167
168
169
170

173
174
175
176
177
178
179
180
181
182

COMPNW 4771

No. of Pages 13, Model 3G

9 May 2012
Q1

D. Lucerna et al. / Computer Networks xxx (2012) xxxxxx

216

costs are assigned according to optimized metrics such as

ber distance, hop count, and link load.
In [8,10], meaningful link-cost assignment approaches
have been proposed to increase the sharing of resources
which are already reserved by the backup paths of other
working connections, instead of reserving new resources.
Ref. [11] shows that backup capacity for SPP can be decreased by exploiting the connection holding-time information. However, these approaches do not take into
consideration that incoming connections may have different availability requirements.
Availability-aware routing has been extensively investigated. In the following we comment only on those works
where availability awareness is coupled with shared path
protection [4,2,1217]. The works in [2,12] propose new
routing algorithms which support service differentiation
under static and dynamic trafc conditions, respectively.
The primary objective is to route connections that comply
with their target availability. A secondary objective is to
minimize resource usage. In [18], it is shown that, if the
cost of a link is dened as a function of its availability, nding a shortest path traversing these links becomes equivalent to nding the Most-Reliable Path (MRP). Authors in [4]
look at the case of SPP with node-disjointness. The work in
[13] also investigates SPP with guaranteed availability
requirements in a dynamic environment and uses a matrix-based approach for availability analysis. To the best
of our knowledge, Ref. [17] is the rst work where connection-availability and connection holding time are jointly
considered. However, neither the outage history of connections nor the connections holding time information are
exploited to dynamically manage the availability targets
during connections lifetime. Both these aspects will be addressed in this paper.

217

2.2. SLA Violation Risk

218

As a second aspect of the overview, consider that, in order to apply availability-aware routing, one has to rely on
rigorous analytical approaches to evaluate the long-term
theoretical availability of an SPP connection. Different availability analysis methods can be found in literature [1822].
An exhaustive comparison of these approaches can be
found in [23]. While the analytical estimation of the availability of an SPP connection is a mature topic, recent literature has shown that the theoretical long-term availability is
not enough to evaluate the quality of a connection provided
over a short period of time (e.g., a period comparable with
the average failure interval) in terms of SLA satisfaction.
More specically, different works have referred to the
concept of SLA Violation Risk, i.e., the probability that, given
a certain availability target AT and certain theoretical longterm availability associated to the path, the offered connection does satisfy the AT. The authors in [24,5] were the rst
to propose to quantify the uncertainty of optical-layer provisioning based on service settings and failure proles. In
[24], the probability of SLA violation is examined based on
simulation. By running a large number of connections in a
given network, the ratio of SLA violations over the total
admitted connections is determined, which essentially corresponds to our SLA Violation Risk at statistical level. How-

183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215

219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241

ever, when the network setting changes, the simulation

must be re-run and hence generality is reduced. In [5], the
authors examine a safety factor to guarantee the SLA Violation Risk focusing mainly on the randomness of the number
of failures in dedicated backup systems. Ref. [25,7] provide
an analytical analysis to compute the probability of SLA
violation considering both the number of failures and the
failure repair time as random variables for the case of a single (unprotected) path and provide a routing algorithm that
minimizes the probability of SLA violation. In [26], the
authors use service continuity (i.e., the probability of
obtaining an uninterrupted service) other than availability
for some classes of service. In [27], the authors show that,
by using Markov models, accurate estimates for the interval
availability distribution (a metric that is strictly related to
SLA Violation Risk) can be achieved, and the authors derive
analytical approximations to the interval availability distribution. Finally, Ref. [6] provides a formulation for interval
availability also for the dedicated path protected case.

242

2.3. Elastic QoS

261

Authors in [28] introduce the concept of Elastic QoS for

a connection, as the possibility to vary the target of QoS for
a connection (e.g., in terms of number of protection paths
for that connection) according to the network state
(mainly according to congestion). In our case we show
how this elasticity can also be achieved in the case of
shared backup resources, by intelligently granting more
or less sharing over the backup resources.
In conclusion, for the rst time to the best of our knowledge, we apply the concept of SLA Violation Risk and QoS
elasticity to SPP. Our new method for availability trading allows us to elastically manage availability and we investigate
how this elasticity affects the SLA-violation-risk properties.

262

3. Notation and problem statement

275

3.1. SPP provisioning problem

276

We rst dene the notation and formally state the SPP

routing problem. A network is represented as a weighted,
directed graph G = (V, E, C, We), where V is the set of nodes,
E is the set of unidirectional bers (referred to as links),
C:E ? R+ is a function that maps the elements in E to positive real numbers representing the link costs, and
We:E ? Z+ species the number of wavelengths on a generic link e (where Z+ denotes the set of positive integers).
We use W ef to denote the number of free wavelengths on
link e 2 E. We denote the set of existing lightpaths in the netn
o
i
i
work at any time by L
lw ; lb ; tia ; tih , where the quadru

i
i
ple lw ; lb ; t ia ; t ih species the working path, the backup path,

277

the arrival time and the holding time for the ith lightpath.
We associate a link vector [8] with each link in the network, to identify the sharing potential between backup
paths. The linkvector me for link e can be represented as
0
0
0
an integer set, mee j8e0 2 E; 0 6 mee 6 W e0 , where mee species the number of working lightpaths that traverse link
e0 and are protected by link e (i.e., their corresponding

289

Q1 Please cite this article in press as: D. Lucerna et al., Trading availability among shared-protected dynamic connections in WDM networks,
Comput. Netw. (2012), http://dx.doi.org/10.1016/j.comnet.2012.04.021

243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260

263
264
265
266
267
268
269
270
271
272
273
274

278
279
280
281
282
283
284
285
286
287
288
290
291
292
293
294
295

COMPNW 4771

No. of Pages 13, Model 3G

9 May 2012
Q1
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313

D. Lucerna et al. / Computer Networks xxx (2012) xxxxxx

backup lightpaths traverse link e). Through such a simple

data structure, the link vector captures the necessary information on the sharing potential offered by each link. The
number of wavelengths which need to be reserved
 for
0
backup lightpaths on link e is thus me maxe0 2E mee .
Therefore, using the link vector, we can simply reserve me
wavelengths on link e as backup wavelengths.
Based on the information contained in me, an SPP procedure has to nd two Shared-Link-Risk-Group (SRLG)-disjoint paths for the incoming request (lw, lb, ta, th), so that:
(C.1) the working and backup lightpaths, lw and lb are
link disjoint;
i
(C.2) lw and lw do not utilize the same wavelength on any
common link they traverse;
i
(C.3) lw does not share any wavelength with lb on any
common link they traverse;
i
(C.4) lb and lb can share a wavelength on a common link
i
only if lw and lw are link disjoint.

314

318

In the following Section 3.3, we will describe how this

SPP provisioning process can be upgraded to enforce availability targets. In the Appendix A a rigorous approach to
evaluate the availability of an SPP connection is shown.

319

3.2. Availability-guaranteed SPP provisioning problem

320

Now we extend the formal problem statement to the

availability-guaranteed SPP provisioning.nLet us redene
i
i
the set of the existing connections L
lw ; lb ; tia ; t ih ; AT i ;
Ai g, where the sextuple species the working path, the
backup path, the arrival time, the holding time, the stipulated availability target specied in the SLA, and the longterm theoretical availability provided to a connection. Similarly, l = (ta, th, SLA) denes the arrival time, the holding
time and the stipulated availability target specied in the
SLA of a new incoming connection. As a difference from
the traditional provisioning approach, the working path lw
and the backup path lb of the incoming connection must satisfy two additional conditions with respect to the other
existing connections in L:

315
316
317

321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337

(C.5) the availability target AT of the incoming connection must be satised (AT 6 A);
(C.6) the availability target AT i of the existing connections must be satised (AT i 6 Ai ).

338

344

Note that the provisioning of the new connection may

reduce the availability of other existing connections due
to the increased sharing of backup capacity. So, if the SLA
of one of the existing connections gets violated due to the
increased sharing, then the incoming connection is blocked
(even if its own availability meets the requirement).

345

3.3. Availability-guaranteed SPP provisioning algorithm

346

Several algorithms have been proposed to solve the

availability-guaranteed SPP provisioning problem. In
Algorithm 1, we describe a baseline approach, called AGP,
for availability-guaranteed SPP dynamic provisioning,
which is a modied version of the algorithm in [2], that

339
340
341
342
343

347
348
349
350

we will use as a holding-time-agnostic counterpart of our

approach. In AGP, a connection can be either unprotected
or shared-path-protected, such that its SLA requirement
is met and network resource usage is minimized. The formulas used in this paper to calculate the long-term availability of the connections is reported in the Appendix.
The routing algorithm follows this two-step procedure:
rst the MRP is computed as the working path. If the SLA
target is not met, then the connection is also provided by
a shared-protected backup path. To compute the backup
path, a new cost function Ce = log(Ae  ae) is applied to
each link e, so that the path with minimum cost will be
the path with maximum availability. Note that Ae is the
availability of link e, while ae represents the availability
product of the links which, in case of double failure, contend on link e backup resources with links belonging to
the working path. If the backup wavelengths on a link e
can be shared by the incoming connection, ae is evaluated
considering only the v e wavelengths already existing on
link e until the new arrival. Otherwise, a link is usable
but not sharable, if no existing backup v e wavelengths
can be shared on it but there is at last one free wavelength
on this link.In thislatter case, ae is evaluated considering e
as formed v e 1 backup wavelengths. ae is rigorously
dened in Eq. (2), where W ef denotes the number of free
wavelengths on link e 2 E.

351

Algorithm 1. Availability-Guaranteed Provisioning (AGP)

377

Input: G = (V, E, C, We), m = {meje 2 E}, the set of existing

n
o
i
i
connections L
lw ; lb ; tia ; t ih ; AT i ; Ai , an

379

incoming connection l = (ta, th, AT)

353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376

380
381

i
lw

Output: A working path or a pair of working/backup



i
i
path lw ; lb for the incoming connection with

382

overall long-term availability A satisfying

constraints C.1-C.6.
1. Compute the MRP for the incoming connection
request, as the working path lw. If Alw P AT , then
A P AT and go to step 4. Block the incoming
connection if path lw is not found.
2. Compute backup path lb with minimal cost
according to the following cost function:

384

Ce

8
>
<1
>
:

383
385
386
387
388
389
390
391
392

393

if

W ef

0^
0

9e0 2 lw jv ee v e

logAe ae otherwise

Compute A (see Eq. (A.5)). If A < AT (i.e., C.5 is

violated), or if path lb is not found, block the
incoming connection request.
3. Re-compute the availabilities for all the
connections in L, which share backup resources
with lb. If there is any connection i 2 L whose recomputed availability does not meet its AT i
requirements (i.e., C.6 is violated), block the
incoming connection request.
4. The connection is accepted and the path lw or the
path pair lw and lb is set-up.

Q1 Please cite this article in press as: D. Lucerna et al., Trading availability among shared-protected dynamic connections in WDM networks,
Comput. Netw. (2012), http://dx.doi.org/10.1016/j.comnet.2012.04.021

352

395
396
397
398
399
400
401
402
403
404
405
406
407

COMPNW 4771

No. of Pages 13, Model 3G

9 May 2012
Q1
408

D. Lucerna et al. / Computer Networks xxx (2012) xxxxxx

8
Q
< ae 8e00 Rflw S egj9e0 2lw ^De0 ;e00 Pv  Ae00
e
e
Q
S
: ae
e0 ;e00
 Ae00
00
0
8e Rflw

egj9e 2lw ^De

>v e

if 8 e0 2 lw jv ee < v e
if W ef 0 ^ 9e0 2 lw jv ee v e
0

410

411

In the next two section, we rst provide in Section 4 a

new version of the AGP algorithm which is able to manage
the availability trading among dynamic connections,
then in Section 5 we mathematically show how the same
concept of trading, initially applied the availability metric
in Section 4, can be extended to the more practical (and recently subject of investigation in various works in the
availability eld) concept of SLA Violation Risk.

412
413
414
415
416
417
418

419

4. Availability-guaranteed SPP with availability trading

420

Let us now discuss the new concept of SPP with availability trading and what are the upgrades to AGP needed
to obtain an algorithm for the new approach.
In the proposed algorithm, whenever a new connection
is offered to the network, we exploit three crucial pieces of
information: (i) the outage history of the connection (i.e., if
an existing connection has been already subject to
outages), (ii) for how long the connection is going to remain
in the network, (iii) for how long the other existing
connections are going to remain in the network. Hence, we
will refer to this algorithm as Holding-Time-Aware (HTA)
algorithm.
The basic idea is that, with respect to a specic connection, NO may pass (with respect to its customers) from situations of availability credit to conditions of availability
debit, and vice versa, as long as the overall SLA target is
guaranteed. There are four possible availability transactions that can be managed by the NO:

421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462

1. NO credit: in an availability-guaranteed provisioning approach, initially the NO is always in credit

with its customers. In fact, a connection is accepted
only if the statistical estimation of the availability
provided in the routing phase is larger than the
availability target AT. But topological constraints
and link-availability granularity of the network
force the NO to provide its customer a long-term
availability level that is higher the target has been
paid for.
2. Reducing NO credit: the NO has a means to provide
an actual availability level closer to the stipulated
value by decreasing the availability targets of living
connections. E.g., if a connection has not been
affected by failures, at the next network change
(typically the next connection arrival) its current
availability requirement can be decreased with
respect to the initial AT.
3. NO debit: two main causes can lead to an NO availability debit: one can be an outage affecting the
connection and putting its SLA at violation risk;
alternatively the NO can voluntarily provide an initial availability level lower than AT. In both cases,
the NO may take availability debit which could be
extinguished by providing a service with larger

availability statistical availability in the future

(when other sharing connections will leave the
network).
4. Paying off NO debit: a service interruption, if not
opportunely managed, may lead to an SLA violation. To avoid this situation when the outage is terminated, NO may extinguish its debit reducing the
sharing degree of the backup resources along the
backup path of the interested connection. This
can be obtained avoiding to share the backup
resources with backup paths of new incoming connections. Similarly, NO may be in debit with an
incoming as a voluntary action to accept more connections. In this case, the NO goes into debit only if
it can be someway guaranteed that, in the future,
the required availability will be provided. In conclusion, an availability debit should be acceptable
only if recoverable by the maximum amount of
future suppliable availability.

464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482

4.1. NO credit

483

Let us refer to the following example to show how an

NO can exploit the connection-holding-time knowledge
to take advantage of an availability credit. In Fig. 1, we
show the state of a network (consider, e.g., "e 2 E:We = 8)
at an instant t c t 2a 10, when connection r2 has to be
provisioned between nodes E and F with t2h 30. A connection r1 has already been routed into the network between
nodes A and B at the instant t1a 0 and it is characterized
by an holding time t1h 20. Both connections require an
availability target AT AT 1 AT 2 0:99.
In accord to the AGP approach, at time t1a 0, we x the
route of the working path of connection r1 along the MRP
(link AB); as a second step, since the availability A1 provided by the working path was less than the availability
target AT1, a backup path was routed on nodes ACDB
(dashed line in gure) utilizing the link cost assignment
in Eq. (1). The connection r1 is accepted because
A1 = 0.99152 > AT1 = 0.99. In this situation we say that the
NO is in availability credit with respect to the customer.
At time t2a 10, the AGP approach xes the route of the
working path of r2 connection along the MRP (link EF);

484

Fig. 1. In this simple network example, traditional availability evaluation

fails to route the second connection, while the availability target
redenition succeeds.

Q1 Please cite this article in press as: D. Lucerna et al., Trading availability among shared-protected dynamic connections in WDM networks,
Comput. Netw. (2012), http://dx.doi.org/10.1016/j.comnet.2012.04.021

463

485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504

COMPNW 4771

No. of Pages 13, Model 3G

9 May 2012
Q1
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527

D. Lucerna et al. / Computer Networks xxx (2012) xxxxxx

as second step, since the availability provided A2 with only

the working path is less than the availability target AT 2 , a
backup path is routed on nodes ECDF (dashed line in gure) utilizing the link cost assignment in Eq. (1). Since r1 and
r2 share a wavelength on link CD along their backup paths,
connection r2 is accepted if and only if conditions C.5 and
C.6 are both not violated, i.e. the availability target of connection r2 is respected and the availability target of connection r1 is still guaranteed. We recall that backup sharing
reduces the availability of a SPP connection. In this case, utilizing the traditional approaches, both the two previous
conditions are not respected, because A1 A2 0:98945
< AT 1 AT 2 . So, request r2 can be either refused or dedicated
protection have to be utilized, which will induce high resource consumption.
However, in our HTA method, we may notice that connection r1 has not been affected by failures during its previous lifetime, from time 0 to time 10. Then, a new
availability target g
AT 1 for r1 can be set, taking into account
that from t1a to tc, the connection has been served with
previous availability A1p 1 and that the connection will
remain in the network from tc to t 1a t1h . In general, the
availability target can be redened as:

528

530

Af
Ti

i

AT i t ih Ap t ia  t c
t ia t ih  t c


3

539

and a new target Af

T i can substitute the previous ATi target
that we were using during the check of condition C.6 in
Algorithm 1. In this specic case, the availability target of
connection r1 will be reduced, i.e. the NOs availability
credit is decreased, given that r1 has not been subject to
any service outage. As result, Af
T i 0:98 < A1 0:98945
and condition C.6 is now respected. However, connection
r2 could not be accepted because condition C.5 is still
violated.

540

4.2. NO debit

531
532
533
534
535
536
537
538

541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562

The availability redenition reported in Eq. (3) may also

be utilized when the NO goes in availability debit due to
network failures. In this case, applying availability redenition allows us to increase the availability target achieving more conservative treatment of the connection and,
in turn, reducing the availability debit. The NO will still accept new connections, but it will avoid that their backup
paths share backup resources with existing connections
which have experimented an outage.
As mentioned before, an NOs availability debit may be
induced voluntarily to accept more connections in the network; in the following we provide an example of a voluntarily induced availability debit. In order to follow the time
evolution of the network links state, we introduce the new
0 ;e00
0
e0 ;e00
symbols me Dsk ; mee Dsk ; ze
Dsk and De
Dsk
e
0 ;e00
e0 ;e00
e0
which express the values of me ; me ; ze
and De
e
respectively, in the time interval Dsk.
Let us expressly dene Dsk rst. According to connection
holding times, the ths can be ordered so that tia tih 6 t i1
a
i1
t
s fs0 ; . . . ; sjLj g
n h ; i 1; 2; . . . ; jLj. As a consequence,
o
jLj
will indicate the departure
0; t1a t 1h ; t 2a t 2h ; . . . ; t jLj
a th
events and Dsk = sk  sk1 expresses the time interval

Fig. 2. Time evolution on link CD of the network in Fig. 1.

e0 ;e00

00

;e
between two departures. mee Dsk ; ze
Dsk and De
e
Dsk will be updated according to the kth connection
departure. In other words, we have divided the time into a
series of intervals Ds which express the distance between
two departures. In Fig. 2 we focus on the departure events
on link CD of the network in Fig. 1, assuming that also connection r2 has been provisioned:

563

 Ds1 (from time 10 to time 20): backup paths of connection r1 and connection r2 share a wavelength on
link CD. During this time interval the provided
availability A1(Ds1) = A2(Ds1) is low and equal to
0.98945.
 Ds2 (from time 20 to time 40): r2 has a dedicated
resource on link CD because connection r1 has left
the network. In this time interval the suppliable
availability A2(Ds2) is equal to 0.99152.

570

565
566
567
568
569

571
572
573
574
575
576
577
578
579

As examined in previous section, during Ds1, NO can reduce its availability credit with connection r1. Moreover, at
the same time interval, NO can go voluntarily into debit
with r2 because it will be paid off during the r2s residual
lifetime, i.e., Ds2. Likewise, the NO will be able to guaranf2 0:99083 and also connectee an overall availability of A
tion r2 will be accepted.
More generally, the state of a link can vary in time, passing, e.g., from shared to dedicated. and the availability provided to the connection could consequently change. Each
of these availability contributions Ai(Dsk) can then be
weighted proportionally over each time interval according
to the following equation:

Aei

580
581
582
583
584
585
586
587
588
589
590
591
592

593

Dsk Ai D k  D k
t ia t ih  tc

where the new Aei can substitute the previous Ai in

conditions C.5 and C.6 of Algorithm 1. In other words, Aei
expresses the maximum suppliable availability for the connection ri if its backup path will be not shared with any other
future incoming connections. This condition can be easily
enforced by preventing the backup paths of future
incoming connection from sharing backup capacity of
connection ri.1
1
Note that, in order to pay back its debit, usually the NO needs to supply
a future availability lower than the maximum future suppliable availability
derived by Eq. (4). As a matter of fact, the debit may be dynamically
compensated by iterative credit reductions if a connection has not
experimented any outages.

Q1 Please cite this article in press as: D. Lucerna et al., Trading availability among shared-protected dynamic connections in WDM networks,
Comput. Netw. (2012), http://dx.doi.org/10.1016/j.comnet.2012.04.021

564

595

596
597
598
599
600
601
602
603

COMPNW 4771

No. of Pages 13, Model 3G

9 May 2012
Q1

D. Lucerna et al. / Computer Networks xxx (2012) xxxxxx

Fig. 3. Flow chart of the AGP and HTA algorithms.

604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620

Exploiting this holding-time-aware (HTA) approach

[11], the NO is able to accept more connections into the network and violates less SLA availability targets by dynamically adapting them to network state evolution. In Fig. 3 a
conceptual description of the main steps in HTA is reported.
Note that without applying the SLA redenition and the
maximum future availability estimation, we can easily
downgrade the approach from HTA to AGP.
It is worth noting at this point that our approach does
not involve reprovisioning of backup capacity, but the NO
simply manages in a more exible manner the sharing of
backup resource. Finally, our approach does not re-evaluate the availability status of all existing connections in
the network, but it is applied only to the incoming connection and to connections that share backup resources with
it. This, for common network scenarios, relevantly limits
the computational complexity.2

621

5. SLA Violation Risk in case of availability trading

622

As often mentioned throughout the paper, availabilityguaranteed provisioning in a optical WDM network must
typically satisfy the condition that the theoretical longterm availability A provided to a connection is greater or
equal to the stipulated availability target SLA. However,
due to the stochastic nature of network failures, even if
A 6 SLA, over a limited time period, there is a not-negligible probability that the actual provided availability turns
out to be less than the availability target, and so the stipulated contracts are usually at risk. Different works have referred to concept of calculating the SLA-Violation Risk
(SLA-VR), i.e., the probability that, given a certain availability target SLA and a certain theoretical long-term availability A associated to a connection, the provisioned path
satises the SLA target (see e.g., [6,7]).
In this section, we investigate an analytical approach
that allows us to utilize the HTA trading approach
proposed considering the SLA-Violation Risk (SLA-VR)

623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639

We remark here that the optimality of our method is not discussed

here, since our method is applied in a dynamic context, where more
emphasis is devoted to scalable and distributed approaches.

instead of the long-term availability. Note that, to the best

of our knowledge, this is the rst time the concept of SLA
Violation Risk is applied in the context of shared path protection (closed form analytical formulation have been provided for unprotected connections and dedicated path
protected connections, e.g., in [6,7,5]).
Let us consider a connection i dened by the sextuple
n
o
i
i
lw ; lb ; tia ; t ih ; AT i ; Ai , as in Section 3.3). If we assume

640

we know the actual availability AAi (i.e., AAi here do not represent the long-term availability, but the actual experienced availability), we can easily calculate the Provided
DownTime (PDTi) and the Stipulated maximum allowable DownTime (SDTi) as:

648

642
643
644
645
646
647
649
650
651
652

653

PDT i 1  AAi  t ih ;

SDT i 1  AT i  tih :

655

While SDTi is the actual maximum allowable downtime

that the customer can experience before the NO must pay a
penalty, the PDTi is the downtime that the customer actually experiences. Therefore, the contract concerning the
connection i is violated when the effective provided availability AAi is lower than the stipulated availability target
SLAi, or alternatively when PDTi > SDTi.
However, the actual availability AAi (and, consequently,
the value of PDTi) is not know a priori, since it depends by
the specic occurrence of randomly-distributed network
failures. It follows that the SLA-VR can only be probabilistically dened and evaluated as:

656

SLA  VR i PrPDT i > SDT i ;

657
658
659
660
661
662
663
664
665
666
667
668
670

which is the probability that the stipulated contract will be

violated. So, in order to consider SLA-VR instead of longterm availability, we must substitute conditions C.5 and
C.6 in Section 3 with the following two conditions to be
satised by a new incoming connection:

671

(C.5bis) The availability target A of the incoming connection must be guaranteed with an SLA-VR lower or
equal than a Prexed Risk Probability (PRP).
(C.6bis) The availability target Ai of the existing connections must be guaranteed with a SLA-VRi lower or equal
than a Prexed Risk Probability (PRP).

676

Q1 Please cite this article in press as: D. Lucerna et al., Trading availability among shared-protected dynamic connections in WDM networks,
Comput. Netw. (2012), http://dx.doi.org/10.1016/j.comnet.2012.04.021

641

672
673
674
675

677
678
679
680
681

COMPNW 4771

No. of Pages 13, Model 3G

9 May 2012
Q1

D. Lucerna et al. / Computer Networks xxx (2012) xxxxxx

682
683
684
685
686
687
688
689
690
691
692
693
694

695
696
697
698
699
700
701
702
703

704
706

707
708

709

In the following, extending the availability redenition

technique presented in Section 4, we dene a new methodology to trade availability credits and debits, under the
new constraint that the Network Operator (NO) only permits
to provision a connection with an SLA-VR 6 PRP.
To compute the SLA-VRi for a generic connection i, we
have to estimate i) its long-term availability Ai and ii) the
Mean Time To Repair MTTRi of that connection. Both estimations depend on the protection scheme adopted. Ai is
computed according to Eq. (A.5) (see Appendix). As for
the MTTRi, we consider here the common assumption that
the failure rate ke is much smaller than the repair rate
le = 1/MTTRe on a generic link e; thus, we can approximate
the MTTRi for a unprotected connection i or for a protected
connection i to MTTRe and MTTRe/2, respectively. These
assumptions have also been discussed and validated in
[5] for dedicated path protection, but we do not expect
the choice of shared protection to signicantly inuence
the MTTR. Therefore the connection failure rate ki can be
derived as (formula obtained inverting Eq. (A.1) in the
Appendix):

ki

1  Ai
:
MTTR i  1  Ai  MTTR i

Now, the SLA-VRi calculation reported in Eq. (7) can be

expressed as the following probability:

SLA  VR i
711
712
713
714
715
716
717
718
719
720
721
722

xjf1AT t ih <xMTTR
i

x
i 
eki th ki tih
x!

Note that the values of MTTR in

 this analysis is assumed to

be constant, so the values of xj 1  AT i  tih < x  MTTR i
can be easily a priori evaluated.
Once the formula in Eq. (7) for SLR-VR has been devised,
in order to extend the SPP availability-trading mechanism
proposed in Section 4 to SLA-VR, the next step is to modify
the computation of the NO credit reduction and the NO
debit payment, so that, not only the target availability SLAi
is redened, but also the SLA-VRi is re-evaluate step-bystep. In the next two subsections we show how.
5.1. NO credit with SLA Violation Risk

725

Referring to the analytical expression in Section 4.1, at

each instant tc (in which a network state change occurs)
we can evaluate the SLA-VRi for a generic connection i as:

728

 

X eki tia tih tc ki t ia t ih  tc x
SLA g
 VR i
x!
x

723
724

726

10

732

i

i
where xjf1  Af
T i  t a t h  t c < x  MTTR i g and a new
couple of availability target and SLA Violation Risk
g
f Af
can substitute the previous one
T i ; SLA  VRi g
fAT i ; SLA  VRi g during the check of condition C.6bis.

733

5.2. NO debit with SLA Violation Risk

734

Referring to the analytical expression in Section 4.2, we

introduce the new symbols ki(Dsk) and MTTRi(Dsk) that express the values of the connection failure rate and the connection Mean Time To Repair, respectively, in the time

729
730
731

735
736
737

interval Dsk. Each of these ki(Dsk) and MTTRi(Dsk) can then

be weighted proportionally over each time interval according to the following equations:

P
ki

Dsk ki D k  D k
t ia t ih  t c

MTTR i

s  Dsk

12

740

741

743

Finally, for the NO-debit calculation, we substitute ki

and MTTRi as computed in Eqs. (11) and (12) in the SLAVRi formula given in Eq. (10). A new couple of effective provided availability and SLA Violation Risk f Aei ; SLA g
 VRi g
can substitute the previous couple {Ai,SLA  VRi} during
the check of conditions C.5bis and C.6bis. In other words,
Aei expresses the maximum supplied availability for the
connection i with a probability 1- SLA g
 VRi under the
assumption that its backup path will not be shared by
any other future incoming connections.

744

6. Illustrative numerical examples

754

We now quantitatively evaluate the performance of the

two approaches: (1) AGP and (2) HTA, the holding-timeaware provisioning approach with availability trading.
We simulate a dynamic network environment with the
assumptions that the connection-arrival process is Poisson
and the connection-holding time follows a negative exponential distribution. Average connection-holding time is
normalized to unity. For the illustrative results shown
here, in every experiment, 105 connection requests are
simulated. All the plotted values have a 95% condence
interval not larger than 5% of the plotted value. Requests
are uniformly distributed among all node pairs; availability
requirements of the requests are uniformly distributed
over the three classes {0.99, 0.999, 0,9999}, denoted as C1,
C2, C3, respectively. The example network topology with
32 wavelengths per ber is shown in Fig. 4. In order to generate the failures in our simulations, the MTTR is considered constant and normalized to 0.032,3 while MTBF
follows a Poisson distribution that guarantees a link availability value of 0.999.
We employ four performance metrics: Blocking Probability, SLA Success Ratio, Resource Distribution and Availability Gap.

755

6.1. Blocking probability

778

The Blocking Probability (BP) indicates the ratio of the

blocked connections over the offered connections to the network. Exploiting the HTA approach, the NO will be able to
accept more connections into the network, because, by
periodically redening the availability target of a connection, more effective backup sharing is allowed in the network. Table 1 compares the BP achieved by HTA and AGP.
For sake of completeness, we also considered our previous
approach reported in [29] which is similar to HTA but only
considers the credit case. The approach in [29] outper-

779

3
With an average connection holding time of 15 days, the MTTR results
equal to 12 h.

Q1 Please cite this article in press as: D. Lucerna et al., Trading availability among shared-protected dynamic connections in WDM networks,
Comput. Netw. (2012), http://dx.doi.org/10.1016/j.comnet.2012.04.021

739

11

Dsk MTTR i D k
t ia tih  tc

738

745
746
747
748
749
750
751
752
753

756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777

780
781
782
783
784
785
786
787
788

COMPNW 4771

No. of Pages 13, Model 3G

9 May 2012
Q1

D. Lucerna et al. / Computer Networks xxx (2012) xxxxxx

19

11

20

21

16

12

15

22

4
13

10

17

14

18

23

24

Fig. 4. A carriers US nationwide backbone network topology.

Table 1
Blocking probability, BP (%): AGP vs. HTA vs. [29].
Arrival rate

AGP
[29]
HTA

789
790
791
792
793
794
795
796
797
798
799
800
801
802
803

20

40

60

80

100

0.526
0.250
0.151

4.38
2.567
1.028

9.0411
5.745
2.6190

14.092
9.034
4.9650

17.677
12.270
7.2391

forms AGP, especially for high loads: e.g., at around 100 Erlangs, BP decreases from 17% to 12%. The HTA approach further reduces the BP respect to the previous approach
reported in [29]: at around 100 Erlangs, BP decreases from
12% to 7%. The connection blocking may be due to four different causes: lack of resources, violation of condition C.5,
violation of condition C.6, and violation of both conditions
C.5 and C.6. Fig. 5a and b shows the impact of the various
contribution to BP in the AGP and HTA approach, respectively, for increasing trafc load. Note that in AGP approach,
the inability to guarantee the SLA availability target to existing connections is the main causes of connection blocking
(violation of condition C.6). As show in Fig. 5b, HTA outperforms AGP, because it drastically reduces the blocking due to
violation of condition C.6.

804

6.2. SLA success ratio

805

814

In Fig. 6a we show the ratio of the connections which

violate the SLA target over the accepted connections to
the network. Our numerical results demonstrate that the
number of violations is much lower when HTA is used.
For the sake of conciseness, we report only the case for
the SLA Class 3, i.e., the class of connections which always
require a backup path. Note that this metric may not result
in a totally fair comparison because different schemes may
accommodate different requests, while this measure does
not differentiate one request from another.

815

6.3. Resource distribution

816

In order to investigate the fairness of these approaches,

we now evaluate the resource distribution among different
classes in AGP and HTA. Connections can be provisioned in
two different ways: unprotected or shared-protected. We

806
807
808
809
810
811
812
813

817
818
819

observe that all the connection requests in SLA Class 1

are unprotected, and all the connection requests in SLA
Class 3 are shared-path protected. The proportion of
shared-path-protected connection increases with the increase in trafc load. In Fig. 6b it can be seen that, with
AGP approach, increasing the network load the percentage
of unprotected connection for SLA Class 2 is incremented:
AGP encourages the routing of connection which does
not require a backup path and blocks the other connections, i.e. connections that cannot meet its SLA target with
only a provisioned working path. On the contrary, using
HTA approach the percentage of unprotected connection
is almost constant with different trafc loads. This comes
form the fact that HTA is able to better assign the backup
sharing during connection holding times and thus it protects a higher number of connections.

820

6.4. Availability gap

836

The relevant BP decrement shown in Table 1 can be

motivated also looking at the reduction of the gap between
the stipulated SLAs and the actual value of the availability
A provided to connections in the AGP and HTA scenario. In
Table 2, we compare the average stipulated SLA (which is
the NO objective and it would be reached if all the connections i in the network would be provided exactly with Ai = SLAi), with the actual average availability provided by HTA
(AHTA) and AGP (AAGP). It can be seen that values for AHTA are
much closer to the SLA target than those of AAGP.
This means that HTA is able to give connections a level
of service in terms of availability which is closer to that required by the customers, freeing backup capacity to be
used for other connections. A second important aspect is
that the degree of sharing of backup resources, for each
connection, varies during the holding time of the connection itself for both AGP and HTA. In the case of HTA, the
accurate redenition of SLA targets allows us to provide
to connections a fairer amount of availability.

837

6.5. The SLA Violation Risk for SPP-protected connections

856

As presented in the rest of the paper, in traditional SPP

schemes the provisioning of a working lightpath and a
backup lightpath for an incoming connection is constrained by the conditions C.1C.6. If the SLA Violation Risk
concept is utilized instead of long-term availability, then
conditions C.5bis and C.6bis have to be respected. Unfortunately, in an availability-guaranteed SPP scheme where an
availability-trading technique is not adopted (such as
AGP), the condition C.6bis may become very stringent, because the SLA-VR tends to grow signicantly during the
connection lifetime. In fact, intuitively, the shorter is the
residual holding time of an existing connection, the larger
becomes its SLA-VR. So the older connections (with a
short residual holding time) may lead to a blocking of
new incoming connections given that the NO is not be able
to guarantee, with a Prexed Risk Probability PRP, the stipulated target availability SLA of these older connections.
In other words an SLA-VR-guaranteed provisioning scheme
could be hardly applicable due to the increase of the value

857

Q1 Please cite this article in press as: D. Lucerna et al., Trading availability among shared-protected dynamic connections in WDM networks,
Comput. Netw. (2012), http://dx.doi.org/10.1016/j.comnet.2012.04.021

821
822
823
824
825
826
827
828
829
830
831
832
833
834
835

838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855

858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875

COMPNW 4771

No. of Pages 13, Model 3G

9 May 2012
Q1

10

D. Lucerna et al. / Computer Networks xxx (2012) xxxxxx

Fig. 5. Blocking probability for AGP (a) and HTA (b).

Fig. 6. Percentage of violated SLA targets for SLA Class 3 (a) and unprotected connections percentage for SLA Class 2 (b): AGP vs. HTA.

Table 2
Actual availability supplied by AGP and HTA, compared to the target SLA.
Arrival rate

Average stipulated
SLA
AGP
HTA

876
877
878
879
880

20

40

60

80

100

0.9963

0.9963

0.9963

0.9963

0.9963

0.99897
0.99882

0.99896
0.99870

0.99895
0.99862

0.99895
0.99854

0.99894
0.99848

of SLA-VR for connections, unless some mechanism for

availability trading is applied, as proposed in this paper.
To numerically support this consideration, in Fig. 7a
we consider the variation of the value of SLA-VR over
the entire holding time (tih 2 years, MTTRi = 12 h) of a

connection i having a long term availability Ai = 0.99897

and a stipulated availability AT i 0:9963 (values taken
from Table 2). Increasing the percentage of expired holding-time, the SLA-VRi tends to increase up to unacceptable
values: at the beginning of its holding time, the connection

Q1 Please cite this article in press as: D. Lucerna et al., Trading availability among shared-protected dynamic connections in WDM networks,
Comput. Netw. (2012), http://dx.doi.org/10.1016/j.comnet.2012.04.021

881
882
883
884
885

COMPNW 4771

No. of Pages 13, Model 3G

9 May 2012
Q1

11

D. Lucerna et al. / Computer Networks xxx (2012) xxxxxx

Fig. 7. SLA Violation Risk for increasing percentage of the expired holding time without (a) and with (b) availability trading and SLA target redenition.

909

AT i has a violation risk of about 0.45%; at about 80% of the

total holding time, the violation risk achieves about 24%.
The sawtooth prole of the SLA-VR is related to the number of allowed number of failures x in Eq. (9): every time
the value of x increases of one unit, the SLA-VR has a peak,
which becomes very high in proximity to the end of the
holding time.
Let us now see how the value of SLA-VR depends on the
percentage of expired holding time if we apply the availability-trading mechanism and we redene the target SLAi. In
Fig. 7b we consider the SLA-VRi evolution for the same connection i with a xed supplied availability Ai = 0.99897 and
a beginning stipulated availability SLAi = 0.9963. For sake of
simplicity, we assume that the connection i is not affected
by any failure during its holding time. The starting value
of SLA-VRi is equal to 0.45%, as in the previous case; by
applying the availability trading and the SLA target redenition, the SLA-VRi now decreases very rapidly. For this reason, the NO would able to reduce the provided availability
Ai, while still guaranteeing an acceptable SLA-VRi P PRP
during the entire connection lifetime. The SLA target redefinition in SLA-VR scenarios conduces to similar benets also
in the NO debit case. For sake of conciseness, we do not report here other results.

910

7. Conclusion

911

924

In this paper we have proposed a new methodology for

availability-guaranteed Shared Path Protection that allows
us to trade availability credits and debits among various connections in a dynamic environment, by increasing
or decreasing the shareability level of the shared backup
capacity. We have shown that our approach allows us to
obtain relevant improvements on blocking probability
and reduces SLA target violations, by appropriately updating the availability targets according to changes in the
sharing degree of the backup resources due to new connection arrivals or to connection departures. We have also
analytically demonstrated how our approach can also be
applied directly to the SLA Violation Risk metric, instead
of using the availability metric.

925

Acknowledgments

886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908

912
913
914
915
916
917
918
919
920
921
922
923

926
927
928

Q4

This work has been partially funded by the Italian

Education Ministry within the Project BESOS (Bandwidth
efciency and Energy Saving by sub-lambda Optical

Switching). Preliminary versions of the content of this paper

were presented at the conferences OFC08 and DRCN09.

929

Appendix A. Availability evaluation for an SPP

connection

931

We provide in this appendix the analytical formulation

to evaluate the availability of a an SPP connection.
In general, availability is the probability of a repairable
system to be in an operating state. Failures and down
states occur, but maintenance or repair action always return the system to an operating state. The basic equation
for the availability of a system with constant failure rate
k and repair rate l is:

933

MTTR
l

;
MTBF MTTR l k

932

A:1

where A is the availability, MTBF = 1/k is the mean time between two consecutive failures, and MTTR = 1/l is the
mean time to repair [19]. According to [30] we consider
failure-immune nodes and we focus only on link failures.
Then, let Ae denotes the availability of the link e.
The availability of the working and the backup paths
can be individually computed. Let lw and lb denote the set
of links used by working path and backup path, respectively. Then, the availabilities of the paths are given by
the following equations:

Al w

Ae ;

A:2

Ae :

A:3

934
935
936
937
938
939
940

941
943
944
945
946
947
948
949
950
951
952
953

954

e2lw

Al b

956

e2lb

Since SPP provides 100% restorability on single failures,

let us consider the effects of double failures on a connection to evaluate the availability [31]: the two additional
0 00
e0 ;e00
parameters ze
and Dee ;e will allow us to identify the
links that cause a resource conict with links owing to
the working path under the assumption that (i) two concurrent failures are affecting the network and (ii) one out
e0 ;e00
of these two failures is affecting the working path. ze
represents the number of working paths that cross the
two links e0 and e00 , and whose backup path contains e.
0 ;e00
The parameter De
denotes the number of wavelengths
e
that would be required on link e in order to fully restore
the trafc on link e even if e0 and e00 fail
e0 ;e00

De

e0

e00

e0 ;e00

v e v e  ze

A:4

Q1 Please cite this article in press as: D. Lucerna et al., Trading availability among shared-protected dynamic connections in WDM networks,
Comput. Netw. (2012), http://dx.doi.org/10.1016/j.comnet.2012.04.021

930

957
958
959
960
961
962
963
964
965
966
967
968
969

970
972

COMPNW 4771

No. of Pages 13, Model 3G

9 May 2012
Q1

D. Lucerna et al. / Computer Networks xxx (2012) xxxxxx

12
0

00

982

In other words, if Dee ;e is larger than me , and if e0 and e00

fail, then some demands cannot be restored on e, because
of insufcient backup bandwidth. Now, for each connec0 ;e00
tion, we can dene Sc fe00 j9fe0 2 lw ; e 2 lb g ^ De
> v e )}
e
00
as the set of links e whose failure causes a conict on link
e if two concurrent failures occur on e0 and e00 . Sc is formed
by a series of links which are all disjoint from the backup
and the working path of the connection.
Q
In summary, if ASc e2Sc Ae , the following equation can
be applied to evaluate the availability of a SPP connection:

983
985

A  Alw Alb ASc  Alw Alb ASc :

986

References

973
974
975
976
977
978
979
980
981

987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042

Q5

Q6

A:5

[1] C. Pinart, G.J. Giralt, On managing optical services in future controlplane-enabled IP/WDM networks, Journal of Lightwave Technology
23 (2005) 28682876.
[2] L. Song, J. Zhang, B. Mukherjee, Dynamic provisioning with
availability guarantee for differentiated services in survivable mesh
networks, IEEE Journal on Selected Areas in Communications 25
(2007) 3543.
[3] A. Iselt, A. KirstSdter, R. Chahine, The role of ASON and GMPLS for the
bandwidth trading market, in: Proc. 1st International Conference on
E-business and Telecomm. Networks (ICETE2004), 2004.
[4] Z. Pandi, M. Tacca, A. Fumagalli, L. Wosinska, Dynamic provisioning
of availability-constrained optical circuits in the presence of optical
node failures, IEEE/OSA Journal of Lightwave Technology 24 (2007)
32683279.
[5] Z. Ling, W.D. Grover, A Theory for Setting the Safety Margin on
Availability Guarantees in an SLA, DRCN 2005, October 2005.
[6] D. Mello, H. Waldman, G.S. Quiterio, Interval availability estimation
for protected connections in optical networks, Computer Networks
55 (2011) 193204.
[7] C.M.M. Xia, M. Tornatore, B. Mukherjee, Risk-aware provisioning for
optical WDM Mesh Networks, IEEE/ACM Transactions on
Networking 19 (2011) 921931.
[8] Canhui Ou, Jing Zang, Hui Zang, L.H. Sahasrabuddhe, Biswanath
Mukherjee, New and improved approaches for shared-path
protection in WDM mesh networks, IEEE Journal of Lightwave
Technology 22 (2004) 12231232.
[9] J.Y. Yen, Finding K shortest loopless paths in a network, Management
Science (1971) 712716.
[10] G. Li, D. Wang, C. Kalmanek, R. Doverspike, Efcient distributed path
selection for shared restoration connections, in: Proc. IEEE
INFOCOM, June 2002, pp. 140149.
[11] M. Tornatore, C. (Sam) Ou, J. Zhang, A. Pattavina, B. Mukherjee,
PHOTO: an efcient shared-path protection strategy based on
connection-holding-time
awareness,
IEEE/OSA
Journal
on
Lightwave Technology 23 (2005) 31383146.
[12] J. Zhang, K. Zhu, H. Zang, B. Mukherjee, A new provisioning
framework to provide availability-guaranteed service in WDM
mesh networks, in: Proc. IEEE ICC03, May 2003, pp. 14841488.
[13] D.A.A. Mello, J.U. Pelegrini, R.P. Ribeiro, D.A. Schupke, H. Waldman,
Dynamic provisioning of shared-backup path protected connections
with guaranteed availability requirements, International Workshop
on Guaranteed Optical Service Provisioning (GOSP), October 2005.
[14] Y. Li, Q. Qiu, L. Li, Availability-aware routing in optical networks with
primary-backup sharing, in: Proc. High Performance Switching and
Routing (HPSR08).
[15] Q. Guo, P.-H. Ho, A. Haque, H.T. Mouftah, Availability-constrained
shared backup path protection (SBPP) for GMPLS-based spare
capacity reconguration, in: Proc. ICC07.
[16] B. Kantarci, H.T. Mouftah, S. Oktug, Adaptive schemes for
differentiated availability-aware connection provisioning in optical
transport networks, OSA Journal of Optical Networking 27 (2009)
45954602.
[17] C. Cavdar, L. Song, M. Tornatore, B. Mukherjee, Holding-time-aware
and availability-guaranteed connection provisioning in optical WDM
mesh networks, in: International Symposium High Capacity Optical
Networks and Enabling Technologies, HONET, November 2007.

[18] J. Zhang, K. Zhu, H. Zang, N.S. Matloff, B. Mukherjee, Availabilityaware provisioning strategies for differentiated protection services
in wavelength-convertible WDM mesh networks, IEEE/ACM
Transactions on Networking 15 (2007) 11771190.
[19] M. Tornatore, G. Maier, A. Pattavina, Availability design of optical
transport network, IEEE Journal on Selected Areas in
Communications 23 (2005) 15201532.
[20] D.A.A. Mello, D.A. Schupke, H. Waldman, A matrix-based analytical
approach to connection unavailability estimation in shared backup
path protection, IEEE Communications Letters 9 (2005) 3543.
[21] L. Zhou, M. Held, U. Sennhauser, Connection availability analysis of
shared backup path-protected mesh networks, IEEE/OSA Journal of
Lightwave Technology 25 (2007) 11111119.
[22] J. Segovia, E. Calle, P. Villa, J. Marzo, J. Tapolcai, Topology-focused
availability analysis of basic protection schemes in optical transport
networks, OSA Journal of Optical Networking 7 (2008) 351364.
[23] M. Tornatore, D. Lucerna, B. Mukherjee, A. Pattavina, Multilayer
protection with availability guarantees in WDM optical networks,
Journal of System and Network Management (Springer) 20 (2012)
3455.
[24] R. Clemente, M. Bartoli, M. Bossi, G. DOrazio, G. Cosmo, Risk
management in availability SLA, in: Proc. of Conference on Design of
Reliable Communication Network (DRCN05). Ischia, Italy.
[25] M. Xia, J. Choi, T. Wang, Risk Assessment in SLA-Based WDM
Backbone Networks, OFC 2009, March 2009.
[26] P. Cholda, A. Mykkeltveit, B. Helvik, A. Jajszczyk, Continuity-based
resilient communication, in: Proc. of Conference on Design of
Reliable Communication Network (DRCN09). Washington, DC, USA.
[27] D. Mello, G.S. Quiterio, H. Waldman, D.A. Schupke, Specication of SLA
survivability requirements for optical path protected connections, in:
Proc. of Optical Fiber Communication Conference (OFC06).
[28] K. Jong, K. Shin, Performance evaluation of dependable real-time
communication with elastic QoS, in: Proc. of International
Conference on Dependable Systems and Networks, (DSN), pp. 295
303.
[29] M. Tornatore, D. Lucerna, L. Song, B. Mukherjee, A. Pattavina,
Dynamic SLA Redenition for Shared-Path-Protected Connections
with Known Duration, OFC 2008, February 2008.
[30] L. Jereb, T. Jakab, F. Unghvary, Availability analysis of multi-layer
optical networks, Optical Networks Magazine (2002) 8494.
[31] J. Doucette, M. Clouqueur, W.D. Grover, On the availability and
capacity requirements of shared backup path-protected mesh
networks, Optical Networks Magazine 4 (2003) 2944.

1086
Diego Lucerna received a Ph.D. in Information
Engineering in 2011 from Politecnico di
Milano. His research interests include
switching technologies, network equipments,
telematic applications and management of
public and private networks. He is currently
enrolled in Huawei Technologies Italia as
Customer Support Optical Engineer for
WDM, SDH and microwave systems.

1089
1090
1091
1092
1093
1094
1095
1096
1097
1098

1088
Massimo Tornatore (S03-M07) received the
Ph.D. degree in information engineering from
the Polytechnic University of Milan, Milan,
Italy, in May 2006. He is currently an Assistant
Professor with the Department of Electronics
and Information, Polytechnic University of
Milan. From 2007 to 2009, he was a PostDoctoral Researcher with the University of
California, Davis, where he is still collaborating as a Visiting Researcher. He is coauthor of
more than 80 conference and journal papers.
His research interests include design, energy
efciency, trafc grooming in optical networks, and group communication security.

Q1 Please cite this article in press as: D. Lucerna et al., Trading availability among shared-protected dynamic connections in WDM networks,
Comput. Netw. (2012), http://dx.doi.org/10.1016/j.comnet.2012.04.021

1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085

1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1100
1115

COMPNW 4771

No. of Pages 13, Model 3G

9 May 2012
Q1
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1117
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147

D. Lucerna et al. / Computer Networks xxx (2012) xxxxxx

Biswanath Mukherjee is a Distinguished
Professor at University of California, Davis,
where he has been since 1987, and served as
Chairman of the Department of Computer
Science during 1997 to 2000. He received the
B.Tech. (Hons) degree from Indian Institute of
Technology, Kharagpur, in 1980, and the Ph.D.
degree from University of Washington, Seattle, in 1987.
He served as Technical Program Co-Chair of
the Optical Fiber Communications (OFC)
Conference 2009, and General Co-Chair of OFC
2011. He served as the Technical Program Chair of the IEEE INFOCOM 96
conference. He is Editor of Springers Optical Networks Book Series. He
serves or has served on the editorial boards of eight journals, most
notably IEEE/ACM Transactions on Networking and IEEE Network. He was
the Founding Steering Committee Chair of the IEEE Advanced Networks
and Telecom Systems (ANTS) Conference, and served as General Co-Chair
of ANTS in 2007 and 2008.
He is co-winner of the Optical Networking Symposium Best Paper Awards
at the IEEE Globecom 2007 and IEEE Globecom 2008 conferences. He won
the 2004 UC Davis Distinguished Graduate Mentoring Award. He also
won the 2009 UC Davis College of Engineering Outstanding Senior Faculty
Award.
To date, he has supervised to completion the Ph.D. Dissertations of 48
students, and he is currently supervising approximately 15 Ph.D. students
and research scholars.
He is author of the textbook Optical WDM Networks published by
Springer in January 2006.
He served a 5-year term as a Founding Member of the Board of Directors

13

of IPLocks, Inc., a Silicon Valley startup company. He has served on the

Technical Advisory Board of a number of startup companies in networking, most recently Teknovus (acquired by Broadcom), Intelligent
Fiber Optic Systems, and LookAhead Decisions Inc. (LDI).
He is a Fellow of the IEEE.

1148
1149
1150
1151
1152
1153

Achille Pattavina received his Dr. Eng. degree

in Electronic Engineering from the University
La Sapienza of Rome, Rome, Italy, in 1977.
He was with the University La Sapienza of
Rome until 1991, when he moved to the
Politecnico di Milano, Milan, Italy, where he is
now a Full Professor. He has been the author
of more than 100 papers in the area of communications networks published in leading
international journals and conference proceedings. He has authored two books,
Switching Theory, Architectures and Performance in Broadband ATM Networks (Wiley, 1998) and Communication
Networks (McGraw-Hill, 1st edn. 2002, 2nd edn. 2007, in Italian). He has
been engaged in many research activities, including European Union
funded projects. His current research interests are in the areas of optical
switching and networking, trafc modeling and multi-layer network
design. Dr. Pattavina has been Guest or Co-Guest Editor of special issues
on switching architectures in IEEE and non-IEEE journals. He has been an
Editor for switching architecture performance for IEEE Transactions on
Communications since 1994, and Editor-in-Chief of European Transactions on Telecommunications since 2001.

1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1155
1170
1171
1172
1173
1174
1175
1176
1177
1178

Q1 Please cite this article in press as: D. Lucerna et al., Trading availability among shared-protected dynamic connections in WDM networks,
Comput. Netw. (2012), http://dx.doi.org/10.1016/j.comnet.2012.04.021