Operation and Service Oriented Strategies in Resilience of Intelligent Optical Networks

(1)

Operation and Service Oriented Strategies in Resilience of Intelligent Optical Networks

Zsolt Lakatos, Tivadar Jakab

Budapest University of Technology and Economics, Department of Telecommunications Magyar Tudósok krt 2, Budapest, Hungary H-1117

E-mail: lakatos@hit.bme.hu, jakab@hit.bme.hu

Abstract

The introduction of switching capabilities and signalling based distributed intelligence enable to implement advanced and effective resilience schemes in intelligent optical networks. Assuming an optical transport layer with intelligent flexibility, resilience schemes are to be implemented in a distributed environment, and several operational oriented considerations should be taken into account to specify the proper recovery strategy. The paper presents a brief overview and some numerical results to analyse the potential solutions.

The new networking capabilities enable the differentiation of transport services from resilience points of view, as well. Some transported services may survive capacity degradations, however others require full recovery. In case of a failure, besides the amount of recovered capacity, requirements originated from different client services can be distinguished according to the time to recovery, as well.

Fundamental recovery schemes like 1+1 dedicated path protection and optimal capacity sharing based path restoration do not enable the tailoring of the recovery times. Resilience with tailored recovery time can be achieved by the application of pre-set switching units in certain cross-connects along the recovery path. This solution reduces as well as the processing load of the cross-connects and the number of active cross-connects along the recovery path. On the other hand pre-set switches reduce the network flexibility, and the resilience related capacity sharing, as well, which results in increased resource needs. The impact of tailored recovery time transport services on the resource needs is illustrated with small numerical examples.

Intelligent optical networks are dynamic networks to serve permanent optical channel requests spread in time and space. Thus, simple and effective distributed real time algorithms are required to operate these networks. Current paper gives some reference LP formulations for different resilience strategies and some illustrative numerical comparisons to support these efforts.

Introduction and Motivations

Despite of the difficult market and economical conditions in the telecommunication the traffic in networks is increasing based on the increasing multimedia content and the new data services like SAN (Storage Area Network) and NAS (Network Attached Storage). However, the rate of the traffic increase is significantly higher than the increase rate of the related revenues, thus, the decreasing profitability of the telecommunication services a key problem for the service providers. To change the negative tendency, there are two main approaches. The most important aspiration is to decrease operational expenses, which implies improved manageability and fast automatic provisioning capabilities. A significant additional issue is the implementation of services with differentiated, tailored characteristics on the available technological basis. It may help to meet user requirements on a more realistic economical basis without resource consuming over- provisioning. The improved switching flexibility and the signalling based enhanced network intelligence implemented in Automatically Switched Optical Networks (ASON) provide a good platform to realise a more cost-effective networking environment.

Quality of service (QoS) provisioning in optical networks is an issue of increasing importance in network design and management. One of the important performance metrics in optical networks is the availability.

Based on the differentiated reliability (DiR) concept [3], different resilience classes according to the recovered capacity (guaranteed entire, partial, best effort, not protected, pre-empted) can be specified.

Significant savings in network resources can be achieved by the joint capacity optimisation of these classes.

In case of guaranteed recovery the reaction time and the time to recovery is an important metric, since the amount of lost information and the performance of the carried application strongly depends on the time of service interruption. Available resilience schemes offer to basic solutions 1+1 dedicated protection based fast recover and shared capacity based re-routing oriented restoration, which recovers the interrupted connection quite slowly. Based on intelligent switching capabilities the recovery time achieved by shared capacity oriented restoration scheme can be tailored according to the client needs.

(2)

One of the main driving forces motivating the efforts to specify and realise ION are the effective support of advanced resilience techniques. To realise efficient solutions to recover optical channels in case of network element failures the optical networking technology should enable complex node and management functions.

The current OTN technology supports dedicated protection based optical resilience mainly, which is fast, but less effective than spare capacity sharing oriented techniques. Based on advanced management functions and automatic switching flexibility more complex resilience techniques like shared protection and restoration can be implemented.

Intelligent Optical Networks

Extended deployment of wavelength division multiplexing (WDM) based optical transport technology went on in recent years to meet the increased capacity demands. Since the WDM technology developments have been focused mainly on huge capacity (tens of wavelengths) and extra long haul (hundreds of kilometres) transmission system the lack of complex network functions limits the advanced applications of optical transport networks (OTNs). The intelligent optical network (ION) concept and technology is being developed to add distributed intelligence and automatic switching flexibility to OTN.

There are three main driving forces motivating the efforts to specify and realise ION:

• Advanced network flexibility: Due to the sharp competition and new services dynamically changing traffic should be served by transport networks. Together with the difficulties in modelling and forecasting traffic demands (IP traffic mainly) these problems result in uncertain traffic conditions, and flexible and re-configurable network elements are required to support efficient networking under these uncertain conditions.

• Fast provisioning of optical channels: Based on advanced management functions and switching flexibility the configuration and reconfiguration of network resources can be speed up significantly, and will become more secure.

• Advanced resilience techniques: To realise efficient solutions to recover optical channels in case of network element failures the optical networking technology should enable complex node and management functions. The current OTN technology supports dedicated protection based optical resilience mainly, which is fast, but less effective than spare capacity sharing oriented techniques. Based on advanced management functions and automatic switching flexibility more complex resilience techniques like shared protection and restoration can be implemented.

The ION is intended to allow switching of (optical) network connections within OTN under control of a proper signalling network, and the concept implies the existence of three separate planes in the network:

• The optical transport plane which provides the functionality required for the transport of the client signals; in particular, it provides the capability to cross-connect the characteristic information of the optical channels;

• The control plane which provides the functionality required for establishing end-to-end connections of client signals with the properties (in terms of protections applied, duration and time scheduling of the connection, etc.) that are specified by the customer himself during connection set-up phase;

• The management plane, which performs management functionality, related both to the transport plane and to the control plane.

The new capabilities implemented in ION enable different kinds of permanent optical transport services:

• Permanent OCh service (P-OCh service): provides the customer with an end-to-end OCh between two end-points. The service is provided by the network operator (NO) on the basis of an agreement between NO and customer. Provisioning of a P-OCh service is the responsibility of the NO, who dedicates multiple network resources to the path. It can be realised via manual equipment configuration or TMN-based equipment configuration.

• Soft-permanent OCh service (SP-OCh service): provides the customer with an end-to-end OCh between two end-points. The service is provided by the NO on the basis of an agreement between NO and customer. Provisioning of a SP-OCh service is the responsibility of the NO, who dedicates multiple network resources to the path. It is realised via distributed signalling-based TMN- activated equipment configuration resulting in provisioning speeds faster than for P-OChs. Like for P-OCh, a “natural” understanding of SP-OCh service is that of a “stable in time” service, as

(3)

Based on these point to point related services more complex ones like optical virtual private network service and lambda trunking service can be implemented, as well. Optical virtual private network service provides the customer with dedicated network resources (OChs) among two or more customer nodes, however, lambda trunking service offers a bundle of P-OChs/SP-OChs between two endpoints with same transmission performances. The EURESCOM P1012 FASHION Project gives a good summary on the ASON based implementation of the intelligent optical network concept [1, 2].

Permanent optical channel service resilience

Both permanent and soft-permanent optical channel based services are “stable in time” leased line services, and full service recovery in case of any single failure is required for them. Both protection and restoration oriented solutions can be applied to recover the transport services from any single failure. The applicable solutions can be evaluated according to their complexity, reaction time and extra resource needs for resilience. The basic resilience schemes can be fined according to operational considerations covering simplified implementation and configuration aspects.

Recent and current progress of optical technology enables the differentiation of transport services from resilience points of view. Some transported services may survive capacity degradations, however others require full recovery [3]. In case of a failure, besides the amount of recovered capacity, requirements originated from different client services can be distinguished according to the time to recovery, as well.

The time to recovery of a single transmission route can be considered proportional with the number of active cross-connects involved in the reconfiguration process, and with the reconfiguration related processing load of each active cross-connect. Fundamental recovery schemes like 1+1 dedicated path protection and optimal capacity sharing based path restoration do not enable the tailoring of the recovery times. Resilience with tailored recovery time can be achieved by the application of pre-set switching units in certain cross-connects along the recovery path. This solution reduces as well as the processing load of the cross-connects and the number of active cross-connects along the recovery path. On the other hand pre-set switches reduce the network flexibility, and the resilience related capacity sharing, as well.

Operational considerations based strategies

Protection based resilience schemes due to their point-to-point oriented structure can be applied both in the core and access part of the network. In dedicated 1+1 protection scheme the transmitted signal is split and permanently bridged to both working and protection systems. The decision on which signal to use is made by the receiver end analysing the signals at the receive terminal. A non-revertive single-ended protection switching is performed on the receiver end. No transfer of extra information is required simplifying the procedure considerably.

Shared protection schemes (n:m, 1:1) are applied if the protected network entities affected by independent failures. In case of a single failure only one protected entity is failed and based on a dual-ended revertive switching mechanism one of the shared protection resources is applied to recover the failure. One of the major application of the scheme in the practice the protection of terminal equipment failures. The scheme is applicable to protect optical channels, if the routing mechanism of the network results (able to set-up) disjoint routes between the same source-destination pairs.

Protection like solutions provide fast recovery time, in order of protection switching, and require high extra resource needs for resilience. 1+1 dedicated path protection practically duplicates the network load, however shared path protection, where two (or more) disjoint paths between the same source-destination pair share the same protection path may decrease the extra load according to the sharing efficiency, which depends on the topology.

Applying restoration with failure-state dependent recovery routes the further decrease of resilience related extra resources could be achieved. The penalty for the higher level of capacity sharing is the increased complexity of recovery processes and the longer recovery time in order of connection set-up time.

According to the different networking considerations restoration with different re-routing strategies based on simple routing algorithms supported by ION can be realised:

• Recovery via optimal capacity sharing related routes: restoration routes optimised to require minimum extra capacity. To achieve this goal the reuse of restoration related extra capacity in different failure cases should be maximise, and even the working paths and re-routing paths can be optimised together.

(4)

• Recovery via minimal paths: minimum weight (e.g. length) paths are used in each failure case to recover failed optical channels. (Suppose that a simple distributed routing protocol is supported in ASON, and applied to implement restoration.)

• Recovery via disjoint minimal paths: restoration routes disjoint to the corresponding working ones (supports simple processes to return to the original working path after clearing the failure), are calculated to minimise the weight (e.g. length).

• Recovery capacity optimal via disjoint minimal paths: like in the previous case the restoration routes disjoint to the corresponding working ones, are calculated to minimise the extra capacity for resilience taking into account the capacity sharing among recovery routes in different failure cases.

Service oriented considerations based solutions

End to end path based resilience schemes are appropriate to achieve service differentiation from resilience point of view. Assuming full network flexibility, i.e. each transport capacity unit terminates on a switched port, the shared capacity oriented restoration is the most effective solution to implement network resilience.

Since the switching flexibility is present in the network due to the advanced network flexibility requirements, the application of this capability to gain from shared capacity related resilience schemes is reasonable.

It is realistic to assume, that the time to recovery of a single transmission route is proportional with the number of active cross-connects involved in the recovery process, and with the reconfiguration related processing load of the active cross-connects originated from the parallel re-routing requests of different failed routes [4].

Fundamental recovery schemes like 1+1 dedicated path protection and optimal capacity sharing based path restoration do not enable the tailoring of the recovery times. 1+1 dedicated path protection realises a fast protection switching scheme with poor resource utilisation characteristics due to the dedicated capacity oriented approach. As far as the shared capacity based path restoration is concerned, the main objective of the design is the optimal capacity sharing without routing constraints. Thus, longer re-routing is applied in numerous cases to achieve high capacity sharing for the recovery from different failure cases, and longer routes imply slower recovery processes.

A minimal path based restoration may improve the recovery times. Applying this scheme failed channels re- routed end to end via their minimal path in the given failure case. The minimal path based re-routing results shorter re-routing path (faster recovery) and lower capacity sharing efficiency in general. However, the topology limits the achievable recovery time solutions.

Further decrease of the recovery time can be achieved by the application of pre-set switching units in certain XCs along the recovery path. This solution reduces as well as the processing load of the cross-connects and the number of active cross-connects along the recovery path.

According to this view the 1+1 dedicated path protection can be considered as recovery without capacity sharing. Since each switching unit is pre-set along the protection path, thus, only receiving end protection switching is applied in case of a failure, i.e. 1+1 dedicated path protection is a recovery mechanism based on a single logical hop recovery route.

Problem descriptions and modeling approaches

Based on the analysis of the technical aspects of the problems in scope a detailed problem description and Integer Linear Programming formulations are given. However, the formulation restricted to the basic problem and the recovery via optimal capacity sharing related disjoint routes, since the extension of the given general formulation to 1+1 dedicated protection and minimal path restoration is well known. The resilience with tailored recovery times requires a graph model extension only, and then almost the same formulation can be applied. The ILP framework presented in this paper an be a good reference developing, analyzing and evaluating provisioning related, real timer routing algorithms.

The general ILP formalisation of the basic resource assignment (network dimensioning) problem is given as follows: (Since fixed network topology is assumed, the graph model of the network can be constructed easily.)

(5)

Given:

the graph model G=(V,E) of the network,

the demands between the networks nodes in a matrix B,

the set and description of the considered failure cases

{

f ∈F

}

,

and additional constraints derived from the resilience strategies.

The task is:

To serve each capacity demand (s,d)s,d∈V, between nodes s and d given in matrix B.

A given demand (s,d) is routed via path p_s^*_,_d =(s,d);s,d∈V , described by the series of the covered edges

{ }

e_i ,(e_i∈E).

The network should serve each capacity demand in case of each specified failure case f , i.e. there is enough capacity in case of G'=(V ,'E');V'=V −V_f ésE'= E−E_f , where V_fis the set of failed nodes and E_f is the set of failed edges.

The capacity required on link e_i can be derived as follows:

⁽ ⁾ , | { }⁽^*, ⁾

*

∑

,

∈

∀

=

d s

i p

e d s

d s

i p

e

C where

{ }

p^*s,d is the set of edges via the active routes.

The target function to be minimised:

on the graph a G=(V,E) the proper path selection

{ }

ps^*,d for each demand ∀s,dshould be perform to minimise

∑

( )

i

ei

C .

Further constraints can be specified to perform the path selection according to the different resilience strategies.

In case of the traditional dimensioning of restoration the recovery routes requiring minimal extra capacity in the specified failure cases are designed. In order to have comparable results fixed working routes are assumed, and the recovery routes are find under this constraint. (Depending on the network size and topology the joint optimisation of the working and recovery routes may result in significant savings in comparison with the basic solution, however the problems to be solved are much more complex.)

A flow based formalisation of the problem can be given as follows:

Let denote A=

{ }

ai_,j the node-edge adjacency matrix, where  1 if node i is the termination of edge j i={1,…,n}

j

ai_, =  -1 if node i is the origin of edge j j={1,…,2m}

 0 otherwise.

Let denote B=

{ }

b_k_,_i the demand matrix, where

 rk if the destination of demand k is node i i={1...n}

i

bk_, =  -rk if the source of demand k is node i k={1...d}

 0 otherwise.

Let denote P=

{ }

pk_,j the path description matrix, where

 1 if edge j is via path of demand k j={1...m}

j

pk_, =  k={1...d}

 0 otherwise.

(6)

Let denote F =

{ }

fu_,j the failure configuration description matrix, where  1 if in failure case u edge j is failed j = {1...m}

fu,j =  u={1...e}

 0 otherwise.

Further notations:

- d denotes the number of capacity demands, -r_kgives the size of demand k,

- e denotes the number of failure cases,

- f_u gives the number of transmission routes failed in failure case u.

Let denote B=

{ }

bu_,l_,j the recovery matrix, where

i k i l

u b

b_,_, = _, if the failed route l in failure case u belongs to demand k (k = {1...d}; l = {1... f_u }, i = {1...n}; u = {1...e}), otherwise 0.

Let denote Y =

{ }

yu_,l_,j the re-routing matrix, where

 1 if in failure case u the failed route l is via edge j j={1...m(^*2m)}, u={1...e}

yu,l,j = 

 0 otherwise. l= {1...fu} Let denote:

-s_j the amount of spare capacities on edge j, and -c_j its cost.

The constraints are:

1. _u_l_i

m j

j l u j u j

i f y b

a _,_,

2 1

* , , ,

, )

( − ⋅ =

∑

=

i = {1...na}; u = {1...e}; l = {1...fu} The flow conservation low should be met in each node in each failure case.

2. _j

f l

j l

u s

y

u ≤

∑

=1 ,, u = {1...e}; j = {1...m}

In each failure case enough capacity is required on each edge to re-route the failed demands.

3. y_u_,_l_,_j,s_j∈Z⁺u = {1...e}; l = {1...fu}; j = {1...m}

Each variable is a non-negative integer.

The target function is:

Min:

∑

= m ⋅

j cj sj

1

An additional constraint is needed to extend the formulation to recovery via optimal capacity sharing related disjoint routes:

(7)

4. P_k_,_j+y_u_,_l_,_j≤1 for ∀k,∀j,∀u, if working route k is identical to recovery route l in failure case u.

This additional constraint restricts the recovery routes for those that node disjoint to the corresponding working path.

Modeling and ILP formulation for recovery time tailoring

Clients with full recovery requirements, but different time to recovery limitations are to support by resilient transport services on common technological, architectural and operational basis. A joint design (optimisation) of the different classes under the objective of high capacity utilisation, i.e. optimal capacity sharing, despite of the reduced flexibility (subset of switching elements are pre-set).

A proper network model is needed to study the problem. The basic idea of the model is to enable the selection of restoration paths with required number of switched hops in order to meet recovery time requirement. To achieve this kind of solutions paths bypassing the switching units via pre-set switches should be modelled. Therefore, additional logical links are introduced into the model with specified physical mapping.

Theoretically, the graph model of the network should have been upgraded to a full mesh with multiplied links (where the multiplicity of each link is the number of routes between the endpoints in the network) according to this approach.

However, simple technical and networking constraints limit the number of additional links significantly. A further reduction can be achieved taking into account that resilience schemes are designed to protect single network failures, in practice. Thus, instead of a full mesh, the initial graph model of the network can be adopted to the problem under study adding one more edge between neighbouring nodes and two extra edges between not neighbouring nodes. The additional edges represent the disjoint minimal routes between the nodes. Thus, at least one route is available in the model in case of any single failure. (Extra edges of the adopted model representing routes are interrupted in case of the failure of any physical link carrying in the original route in the basic model.)

Applying the above model extension, the size of the model is increased. The original network model is a G(N,E) graph, with nodes |N|=n and edges |E|=e. The extended model includes:

e more edges (one extra edge between neighbouring nodes)

2(n-e) more edges (two extra edges between not neighbouring nodes).

The number of nodes remains unchanged.

Thus, the extended model G*(N, E*) consists |N|=n nodes, and |E*|=e+e+2(n-e)=2n edges.

Based on the network model the well known ILP path based formulation is applied to elaborate case studies.

(There are (1+m)e+e conditions and (1+p)e+e variables in path formulations, where m represents the number directly connected pairs, and p represents the number of path to be re-routed in case of failures.)

The above described problem formulation can be applied to the extended graph model with a modified target function, where the proper weighting of the logical edges are included.

Each formulated problem is solved by ILOG CPLEX Solver to achieve the presented results.

Illustrative Numerical Results

Based on the developed models and methodology the applicability of different ION based resilience schemes to a small size network (hypothetical Hungarian core network with 7 nodes and 13 edges) is studied. Based on the illustrative numerical results the different re-routing strategies based on operational considerations can be compared, and the impact of the tailored recovery time approach in case of differentiated transport service classes can be analyzed.

Impact of Operational Considerations on Resource Needs of the Network

Figures 1a and 1b illustrate the efficiency of spare capacity related restoration. Concerning the line capacities (Figure 1a), as it is expected the extra capacity for resilience is the lowest for capacity-optimal restoration (less then 50% of the working) and the highest for 1+1 dedicated protection (about 150%). The minimal path restoration requires 25% more in comparison with the capacity-optimal one (it is 75% extra in total), and disjoint minimal path restoration is resulted in 75% more in comparison with the capacity-optimal one (it is 100% extra in total). However, when the disjoint paths are designed to achieve maximum capacity

(8)

sharing, the disjoint path restoration requires the same amount of extra capacity then the capacity optimal restoration (without path selection constraints).

0%

50%

100%

150%

200%

250%

300%

not prot. optimal path rest.

min.

path rest.

disjoint min.

path rest.

disjoint optimal path rest.

1+1

Resilience Cases

Relative Hop*OCh Extra for Resilience Working

Figure 1a – Total link capacities in different resilience cases

0%

50%

100%

150%

200%

250%

300%

not prot.

full flex.

1+1 end switch.

1+1 full flex.

optimal path rest.

full flex.

Resilience Cases

Relative #Switch Ports

Extra for Resilience Working

Figure 1b – Total switch capacities (#ports) capacities in different resilience cases

Despite of the small size of the network example (9 nodes, 16 links, more than 350 demand units) the restoration performs well, however, in larger size networks the sharing efficiency could be better. Instead of path based restoration the partitioning of the paths into commonly recovered sections further saving can be gained, and the reaction times can be shortened. The amount of savings in switching capacity (Figure 1b) achieved by the restoration is very similar to that for the line capacity. A full flexible network is assumed, i.e. each line capacity unit terminates on switches, and the switching capacity is expressed in number of ports. (As a reference, the amount of switches for protection switching only is given, as well.)

Impact of Tailored Recovery Time Approach on Resource Needs of the Network The trade-off between recovery speed and resilience related capacity sharing efficiency is illustrated on Figures 2 and 3. Figure 2 gives the average logical hop count of recovery routes.

Recovery path statistics for different resilience options

0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50

1+1 path prot.

single hop path rest.

min. path rest.

double hop path

rest.

double and triple

hop path rest.

capac.

opt. path rest.

Resilience options

Average logical hop count

Figure 2 – Recovery path logical hop statistics for different resilience options

Resource needs for networks with different resilience options

0%

20%

40%

60%

80%

100%

120%

1+1 path prot.

single hop path rest.

min. path rest.

double hop path

rest.

double and triple hop path rest.

capac.

opt. path rest.

Resilience options

OCh*hop

Figure 3 – Resource needs for networks with different resilience options (35% for working capacities)

Figure 3 depicts the link capacity needs for different resilience options, where the total link capacity of the network with 1+1 dedicated path protections selected as the reference (100%), (The amount of working connections related link capacity is 41%.)

The resilience options under study are as follows:

1+1 dedicated path protection: realises the fastest recovery scheme due to the receiving end single point protection switching. It can be considered as a single switched hop based solution. If there are several physical hops along the recovery path, the switches in the nodes adjacent to these links are pre-set to establish the recovery path in advance, during the provisioning phase. The information is launched both to the working and recovery path on the source end and forwarded to the receiving end. The receiving and selects the failure free information (if any) from the received streams and performs protection switching action if needed.

(9)

Single switched hop path restoration: It is a 1:1 like solution with preliminary fixed minimal working path.

There is no capacity sharing among the recovery routes of different failure states because of the single switched hop based solution. Both end switching is required in case of a failure, therefore, communication is required between the termination nodes to co-ordinate the switching action. Thus, the solution is slower then 1+1 dedicated protection, and may require more resources (due to the preliminary fixed minimal working paths). The potential advantage of the scheme is the empty channels via the recovery path, which may transport low priority traffic.

Minimal path based restoration: realises resilience via constrained recovery routes. Minimal paths in each failure state are applied to recover failed transmission routes. Capacity sharing among the recovery routes of different failure states can be achieved; however, the minimal paths may be not optimal from capacity sharing point of view. The number of switched hops via the recovery path (the number of active intermediate XC nodes via the recovery path) depends on the network topology.

Double switched hop path restoration: one active intermediate XC node is involved in the recovery re- routing. The intermediate flexibility point enables the capacity sharing among the partially or entirely adjacent recovery routes of different failure states.

Traditional restoration without routing constraints: is designed to achieve maximal capacity sharing among the recovery routes of different failure states. Thus, relatively long routes can be applied to maximise the capacity sharing, which results in slow recovery processes.

Besides the above described uniform cases a mixed case with two recovery time classes is analysed, as well.

The traffic demands are equally split for two recovery time classes, the first class for traffic with fastest, the second one with lowest recovery requirements, i.e. with double and tripe logical hop recovery paths, respectively.

As it can be depicted on Figures 2 and 3 schemes with increasing recovery requires less extra resources for resilience. On the other hand, recovery time decreasing restrictions result in minor resource need increases.

The capacity optimal restoration requires 3.3 switched hops via recovery path in average. Specifying a double switched hop constraint for each recovery path to achieve shorter recovery times, the penalty on resource needs is minimal (67% instead of the optimal 59%). Due to the joint optimisation the result is similar applying the above described two recovery time classes (64% instead of the optimal 59%).

Summary and Conclusions

Based on automatic switching capabilities and distributed intelligence in automatic switched optical networks restoration can be apply to recover permanent optical channel based services from different network failures. In comparison with dedicated protection significant resource savings can be achieved due to the effective spare capacity sharing (Table 1), however, the recovery time is increased from order of protection switching to order of connection set-up time.

Table 1 Summary of soft-permanent service resilience related capacity needs with different recovery strategies (not protected case =100%)

Resilience scheme Extra total link capacity Dedicated 1+1 path protection 152%

Shared path protection 131%

Disjoint optimal path restoration 39%

Disjoint minimal path restoration 103%

Minimal path restoration 72%

Optimal sharing path restoration 39%

Table 2 Summary of soft-permanent service resilience related recovery time vs. capacity needs results (1+1 dedicated path protection case=100%)

Resilience scheme Average number of switched hops via the recovery route

Relative total link capacity Single hop path restoration (1:1 like

solution)

1 103%

Minimal path restoration 2.5 82%

Double hop path restoration 2 65%

Capacity optimal path restoration 3.25 60%

(10)

Different simple restoration routing strategies may offer benefits in networking (supported implementation, simple return to working path after clearing a failure, etc.), however, increased resources needs are resulted from these simple routing strategies in comparison with the capacity optimal path based restoration.

A practical solution to implement resilient transport services with tailored recovery time has been proposed.

The intelligent automatic configuration and management based solution enables to refine the resilient transport service portfolio with services of different recovery time (Table 2).

The proposed scalable solution provide a promising approach to realise shared spare capacity oriented resilient transport services with faster recovery than traditional restoration and with more effective capacity utilisation than 1+1 dedicated protection.

Since in intelligent optical networks the permanent optical channel requests are assumed to arrive spread in time and space efficient distributed real-time routing algorithms are required to operate these networks. The analysis and ILP formulation of the different resilience strategies presented in the current paper provide a good reference frameworks for these works.

Acknowledgement

This work was partially supported by the research grant OTKA T30685.

References

[1] FASHION - Flexible, Automatically SwitcHed, client Independent Optical Networks, EURESCOM P1012 Project, Homepage: http: //www.eurescom.de/public/ projects/P1000-series/p1012/default.asp [2] B. Craignou, R. Clemente, J. Robeday, L. Jereb, Z. Ioannidis: Network Operators Perspectives on

Optical Networks - Evolution towards ASON, Proceedings of 10th International Network Strategy and Planning Symposium, NETWORKS’2002, Munich, Germany, June 2002, pp. 547-556

[3] A. Fumagalli, M. Tacca, F. Unghvary, A. Farago, Shared Path Protection with Differentiated Reliability, in Proc. IEEE ICC 2002 New York City, NY, April-May 2002

[4] B. X. Weiss: Ultra-fast protection and restoration in mesh networks, Proceedings of 10th International Network Strategy and Planning Symposium, NETWORKS’2002, Munich, Germany, June 2002, pp.

469-476