Enhanced pool sharing: a constraint-based routing algorithm for shared mesh restoration networks [Invited] Hassan Naser and Hussein T. Mouftah School of Information Technology and Engineering, University of Ottawa, Ottawa, Ontario K1N 6N5, Canada hnaser@site.uottawa.ca; mouftah@site.uottawa.ca RECEIVED 27 OCTOBER 2003; REVISED 27 MARCH 2004; ACCEPTED 29 MARCH 2004; PUBLISHED 15 APRIL 2004 We investigate the availability performance of networks with shared mesh restoration and demonstrate that these networks cannot provide highly available protection services. A major factor in the poor performance of shared mesh restoration is that the resources at backup links are shared among demands. If multiple service-affecting failures occur in the network a multitude of these demands will rush to utilize the spare resources on backup links. These resources are not adequate to serve all of these demands simultaneously. We propose a heuristic routing algorithm that attempts to improve the availability performance of shared mesh restoration. We measure the likelihood that a backup link will not be available to restore a newly arrived demand if or when more than one failure occurs in the network. We adjust the backup bandwidth on that link if the measured likelihood exceeds a preset threshold. As a typical representative, we show that the downtime improves by 7%, 11%, and 18% when the total backup bandwidth in the network is increased by 5%, 10%, and 20%, respectively. These values are obtained through a series of fitting experiments. © 2004 Optical Society of America OCIS codes: 060.4250, 060.0060. 1. Introduction Shared mesh restoration attempts to minimize the spare capacity that is required for com- plete recovery of traffic from any single service-affecting network failure (e.g., fiber cut). Many studies have explored the capacity saving of various shared mesh restoration schemes over the alternative protection and restoration schemes [1–3]. However, not many studies have been reported on the evaluation of the quality of service (QoS) performance of the shared mesh restoration schemes. In this article, QoS represents the amount of time that a connection is down or not available. The connection is defined by two paths—one for working (service) and one for backup (restoration). Such a connection is computed and signaled when an arriving demand requires a protected service. The working and backup paths that make up the connection must be failure disjoint, meaning that they shall not become simultaneously unavailable when a single network element fails. In shared mesh restoration, a demand may share resources (e.g., bandwidth) on links along its backup path with many other demands as long as their working paths are mutually disjoint. Only one of these demands can claim (use) the resources on any of these mutual links at any time. These resources are not available to all other demands while they are in use. It is not clear what level of QoS shared mesh restoration schemes can deliver to a typical demand—called tagged demand in this paper. Reliability methods—such as block © 2004 Optical Society of America JON 3217 May 2004 / Vol. 3, No. 5 / JOURNAL OF OPTICAL NETWORKING 303