Published on April 16, 2008
Design and Analysis of Mesh Networks with Multiple Quality of Protection (QoP) Classes: Design and Analysis of Mesh Networks with Multiple Quality of Protection (QoP) Classes Matthieu Clouqueur, Wayne D. Grover [email protected], [email protected] TRLabs and University of Alberta Edmonton, Alberta, Canada Motivations and Objectives of Study: Motivations and Objectives of Study Most work to date considers “single failure restorability” as the only protection option Increase of data traffic Best-effort restoration or no restoration at all in the optical layer may be sufficient for some customers This study investigates the properties of networks with multiple Quality of Protection classes: From a capacity planning viewpoint. and From the viewpoint of the effects of failures on the different types of service Definitions: Definitions Quality of Protection (QoP) Characterizes the level of protection against various failure types (e.g. single span failures, dual span failures, etc.) Different from QoS (usually characterizes throughput and delay in packet-switched networks) R1 service assured single span failure restorability R2 service assured dual span failure restorability Spare capacity shared but idle standby capacity of a mesh network that is used to protect services from different failure scenarios QoP Classes: QoP Classes Multiple Quality of Protection (QoP) Classes: Possible additional QoP class Gold and Silver may both preempt Economy service capacity but Silver only does so after all of Gold-class requirements are met. Demand Scenarios: Demand Scenarios Considers four multi-service demand scenarios for testing : Designs will be “ jointly optimized” routing of gold and economy paths are synergistic decisions Scenario A: Mostly Gold and Economy Scenario B: Mostly Economy and Silver Scenario C: Mostly Silver and Gold Scenario D: Mostly Gold and Silver Study Methodology: Study Methodology Capacity Design AMPL®/CPLEX® Restorability Analysis C Program Demand Scenario Network Topology Network Design Multi QoP design IP Formulation Restorability Analysis Results Conclusions: Capacity properties of multi QoP network environment Conclusions: Restorability properties of multi QoP network environment Experimental Data: Experimental Data Three test network topologies: Net-A (20 nodes, 40 spans) Net-B (25 nodes, 50 spans) Net-C (30 nodes, 60 spans) Test demand: 20 lightpath demands between each node pair (55% = 11 lightpaths, 30% = 6 lightpaths, 15% = 3 lightpaths) Design Formulation: Design Formulation Minimize: total cost of capacity installed subject to: (a) All Gold, Silver, Bronze and Economy service demands are routed and assigned working capacity. (b) Restorability of Gold services to single failures (c) Sufficient spare and/or Economy-class capacity to support the largest restoration flows needed for (b) (d) (optionally) working and spare capacities allocated on each span fit under a limited set of available modular capacities. Solved using AMPL® and CPLEX ® Sample Capacity Design Results: Sample Capacity Design Results Capacity Costs: (Results for network Net-B) Observation: In the first three demand scenarios: No spare capacity is required ! Main Findings of Capacity Design: Main Findings of Capacity Design No spare capacity needed in most cases: Gold-class restoration requirements fully met by preemption of Economy-class services. Significance: Such network designs are “fully restorable” in the usual sense (for the Gold-class customers) but there is no unused standby capacity. All capacity is earning revenue at some level or other. Such flexible, highly efficient “multi-QoP” capabilities may turn out to be the most commercially significant advantage (the “killer app”) for mesh-based transport networking. Restorability Analysis: Main Research Questions: Restorability Analysis: Main Research Questions What is the restorability of Silver-class services, R(Silv.)? If they are not allowed to preempt the Economy class? If they are allowed to preempt the Economy class? How bad is life for the Economy-class services? What is the probability of preemption P(preempt.) of an Economy-class service during a failure that did not directly affect it? Is there a trade-off? Can we control it? Economy path P(preempt.)? Restorability Models: Restorability Models Two options for restoration model: Option 1: Silver cannot preempt Economy Silver can preempt Economy Option 2: Use “true spare capacity” or Economy-class cap. indifferently when searching restoration paths Always use “true spare capacity” first when possible before preempting Economy Sample Restorability Results: Sample Restorability Results There is indeed a trade-off Sample result R(Silv.) P(preempt.) for Economy-class However: Econ-class P(preempt.) remains fairly low, especially for demand mix B (Mostly Econ. and Silver) (Results for network Net-B) Main Findings of Restorability Analyses: Main Findings of Restorability Analyses The restorability of silver-class services R(Silv.) highly depends on Option 1 (allowing it to preempt Economy or not): If Silver cannot preempt Economy: Silver-class restorability can be very low because very little spare capacity is allocated If Silver can preempt Economy: Silver-class restorability depends on the demand scenario (almost identical to Gold-class restorability in scenario B) In both cases P(preempt.) remains reasonably low (always below 15%) Choice of Option 2 has much less impact on R(Silv.) and P(preempt.). Summary of Study: Summary of Study A model for optimal multi QoP capacity design was developed In many cases no spare capacity was required at all Mesh networks with multi-QoP = higher capacity efficiency R(Silv.) and P(preempt.) for Econ. class were studied using four restoration models R(Silv.) can be fairly high if Silver allowed to preempt Economy For Economy, P(preempt.) remains fairly low Some demand scenarios (A and B) offer all desirable characteristics: High silver-class restorability and low frequency of preemption of Economy. A pricing strategy can be used to create these demand mixes. Mesh restoration appears to be a perfect candidate for providing services with various protection options. Publications (available upon request): Publications (available upon request) W. D. Grover, M. Clouqueur, “Span-restorable mesh networks with multiple quality of protection (QoP) service-classes,” in Proceedings of the International Conference on Optical Communications and Networks, ICOCN 2002, Singapore, November 2002, pp. 321-323. Extended version of above paper to appear this year in Photonic Network Communications (Kluwer).