Improved survivability analysis for SONET SHRs

Recently, a special type of Markov model called parametric state reward Mar kov model (SRMM/p) [H.C. Cankaya, V.S.S. Nair, in: IEEE Proceedings of GLOB ECOM 97, vol. i,IEEE Computer Sec. Press, Silver Spring, MD, 1997, pp. 252- 256] and a set of survivability metrics comprising reliability, availabilit y, and restorability have been proposed for the evaluation of self-healing SONET mesh networks [H.C. Cankaya, V.S.S, Nair, in: Proceedings of ISCIS 'X II, vol. i, Bogazici University, Bogazici University Press, 1997, pp. 269-2 76]. The SRMM/p, accommodates multiple consecutive link failures and uses t opology-free approximation in order to calculate the average performance lo ss due to a failure. The SRMM/p is equally applicable to the analysis of se lf-healing SONET rings by considering a ring as a special case of a mesh to pology [H.C. Cankaya, V.S.S. Nair, in: IEEE Proceedings of GLBBECOM '98, vo l. 4, IEEE Computer Sec. Press, Silver Spring, MD, 1998, pp. 2276-2281]. Fu rther, the topological uniformity and simplicity of rings allow one to incl ude more detailed features of the network in the model so that the analysis will be more accurate. For this purpose, we propose an improved approach t o the survivability analysis of self-healing SONET rings which employs a pr obability-tree based evaluation of the probability of various system states resulting from all possible combinations of node and link failures. The co rresponding tree-construction and traversal algorithms are presented. Survi vability of rings with distinctive demand patterns are studied with the imp roved analysis and compared experimentally. One limitation to the model is the high run-time complexity caused mainly by the disparity between transit ion rates amongst various states in the SRMM/p [H.C. Cankaya, V.S.S. Nair, ACM Comput. Commun. Rev. 28 (4) (1998) 268-277]. In this paper, we also pre sent an approach to circumvent this problem by state aggregation method and compare the results in terms of runtime complexity and accuracy by conduct ing an experimental study. (C) 1999 Elsevier Science B.V. All rights reserv ed.