Shared Backup Path Optimization in Telecommunication Networks Bal´azsG´aborJ´ ozsa * and D´ aniel Orincsay * Abstract Thispaperpresentsnewapproachesforoff-lineglobalpathoptimizationintelecommunicationnetworks wherebackuppathsusesharedbandwidthreservation.Thisproblemariseswhenaserviceproviderwantsto findoptimalpathsforinterruption-criticaltrafficflowswithinitsnetworkinordertoimprovetheresource utilization.Inthispaperwefocusonprotectedtrafficflows,whereeachtrafficdemandhasanactiveworking pathandabackuppathforrestoration. Withthehelpofanoff-lineglobalpathoptimizationprocessthe network load can be reduced, even if on-line routing algorithms optimize the placement of traffic flows, because the sequential optimal local routing decisions may result in a globally suboptimal routing of the flows. Weintroduceanumberofdifferentapproachesforsolvingtheaboveproblem,whichcanbeusedin several types of networks (e.g., Asynchronous Transfer Mode, Multiprotocol Label Switching, Wavelength Division Multiplexing) that allows such traffic engineering features as bandwidth guarantees and backup paths. We investigate the algorithms through empirical tests based on practical size randomly generated networksaswellasonreal-worldbackbonenetworktopologies. Theresultsshowthatusingtheproposed methodstheresourceutilizationcanbeimprovedsignificantly. Keywords: global path optimization, backup path, shared protection, spare capacity allocation, IP and MPLS 1 Introduction Nowadays, network customers need guaranteed bandwidth from Internet Service Providers (ISPs) besides the typical best-effort services, moreover, protection of such traffic that should not be interrupted by failures is also necessary. During normal operation, traffic flows use the so-called active paths that are protected by pre- established disjoint backup paths where the traffic flows can be switched quickly when a link failure is detected. This technique can be used in several types of networks (e.g., Asynchronous Transfer Mode, Multiprotocol Label Switching, Wavelength Division Multiplexing) that support the above-mentioned traffic engineering feature [1]. In the literature various terms are used for the above two kinds of paths; in this paper we use terms active and working path interchangeably as well as terms backup, recovery, and restoration path. More disjoint backup paths can be used for one active path, with the result that we can tolerate more failures at one time. Due to the rare failures, we assume that only one link fails at one time (single failure), therefore we consider one backup path per active path. To guarantee the same quality for traffic flows when a network component fails, the reserved bandwidth for a backup path must be equal to the bandwidth of the active path. There are different approaches of protection and restoration techniques [2]. “1+1” protection means that data is sent over both (active and backup) paths simultaneously, which is the fastest and most secure method but it doubles the network load. In “1:1” protection each active path has a hot-standby backup path whose resource reservations are independent from the other paths. This type of protection is simple and fast but it requires high amount of resources (double amount compared to the case without protection). The resource reservation can be reduced significantly—without reliability degradation—in the following way: in case of a single failure, among two disjoint active paths at most one can fail, thus they can share their bandwidth reservation on the common links of their backup paths (see Figure 1). This is called shared backup reservation (also called shared protection) that has been studied extensively in the literature [3, 4, 5, 6, 7, 8, 9] and it seems to be a good compromise between protection and bandwidth consumption. ISPs are not only interested in providing reliable services with quality guarantees but on the other hand, they aspire to improve the network performance, therefore optimal placement of traffic flows within their networks is also very important. There are two quite different ways of path optimization. Using on-line optimization (e.g., [10]) we establish the paths immediately for the arriving traffic demands based on optimal local decision * Budapest University of Technology and Economics, Department of Telecommunication and Telematics, High Speed Networks Laboratory, H-1117 Magyar tud´osok k¨or´ utja 2, Budapest, Hungary; Ericsson Research, Traffic Analysis and Network Performance Laboratory, POB 107, H-1300 Budapest, Hungary. E-mail: {jozsa, orincsay} @ttt-atm.ttt.bme.hu Fax: +36-1-437-7767 Phone: +36-1-437-7179