130 likes | 143 Views
This paper discusses the impacts of Shared-Risk Link Groups (SRLGs) on survivable network capacity and restorability. It examines the cost of protecting against SRLGs, the effects of co-incident SRLGs on restorability, and strategies for identifying troublesome SRLGs. The experimental approach involves design formulation, testing with varying numbers and positions of random co-incident SRLGs, and analysis of their impact on network capacity. The results show that co-incident SRLGs are more troublesome than typical dual-span failures and require more spare capacity.
E N D
Capacity Design Studies of Span-Restorable Mesh Transport Networks With Shared-Risk Link Group (SRLG) Effects John Doucette, Wayne D. Grover doucette@trlabs.ca, grover@trlabs.ca TRLabs and University of Alberta Edmonton, AB, Canada OptiComm 2002 Boston, MA, USA 30/July/2002
Outline • What is a Shared-Risk Link Group (SRLG)? • Research Questions • Experimental Approach • Design Formulation • Results • Cost of Protecting Against SLRGs • Co-Incident SRLG Effects on Restorability • Effects of Co-Incident SRLG Location • Identifying Troublesome SRLGs • Concluding Remarks
What is a Shared-Risk Link Group (SRLG)? • Here a Shared-Risk Link Group (SRLG) is a set of spans with a common cause of failure. • A Co-Incident SRLG is an SRLG involving multiple (nominally disjoint) spans, incident upon a common node. • Most common type of multi-span SRLG.
Research Questions • What impacts do SRLGs have on survivable network capacity requirements? • How costly is it to protect against SRLGs? • How do SRLGs affect a network’s restorability? • Which types of SRLGs are most costly to protect against? • How many SRLGs can be sustained before the capacity penalty becomes too severe? • Can we identify which SRLGs are most troublesome? • When is it worthwhile to take physical measures to eliminate an SRLG?
Experimental Approach • Define new design model: • Spare capacity design for 100% restorability in the face of all single span failures and specified SRLGs. • Test with varying numbers and position of random co-incident SRLGs. • What is the cost of spare capacity for full restorability? • Test network has 40 nodes, 70 spans, 780 O-D demands of 1-10 wavelength paths each. • Analyze and rank individual co-incident SRLG impact based on nodal degree and location. • Which individual SRLGs cost more to make fully restorable?
SRLG-Tolerant Design Formulation Minimize: Total Spare Capacity Subject To: (1) Restoration flow for all single-span failure scenarios (2) Restoration flows for all specified SRLG scenarios (3) No restoration flow on co-failed spans in SRLG scenarios (4) Spare capacity to support (1) and (2)
Cost of Protecting Against SLRGs It costs nearly as much to make a given percentage of co-incident SRLGs restorable as it does to make the same percentage of all possible dual span failures restorable.
Co-Incident SRLG Effects on Restorability Making a given percentage of all co-incident SRLGs restorable provides restorability/availability nearly as good as making the same percentage of all dual span pairs restorable. Co-Incident SRLGs Making even a small number of co-incident SRLGs restorable will make a much greater number of dual failures fully restorable. Random Dual Failure SRLGs % Fully restorable dual failures
Classifying Node Locations “interior” “edge” “near-edge”
Effects of Co-Incident SRLG Location Co-incident SRLGs at nodes in the interiorof the network are more costly than those closer to the edge.
Effects of Co-Incident SRLG Location (2) Co-incident SRLGs at degree-3 nodes are more costly than those at degree-4 nodes. Degree-3 Degree-4
Removing the 4 worst ranked SRLGs gives a 55% reduction in capacity penalty. Remove the most costly SRLGs and re-solve the capacity design with remaining SRLGs. Identifying The Most Troublesome SRLGs A random set of 18 co-incident SRLGs and evaluated their costs as a set and individually.
Concluding Remarks • Co-incident SRLGs are more troublesome than typical dual span failures. • Believed to be most common type of multi-span SRLG. • Requires more spare capacity on average than other SRLGs or dual failures. • Dual failure restorability (hence availability) responds best to investment for coverage of the co-incident SRLGs. • Co-incident SRLG effects are generally more troublesome (require more capacity) if located at degree-3 nodes in the network interior. • Identification and removal of the individually most troublesome co-incident SRLGs gets most of the benefit of removal of all SRLGs.