Tightest constraint first: An efficient delay sensitive multicast routing algorithm

Abstract

As a key issue in multicast routing with quality of service (QoS) support, constrained minimum Steiner tree (CMST) problem has been a research focus for more than a decade, and tens of heuristics have been developed to solve this NP-complete problem. Among all the previously proposed algorithms, the bounded shortest path algorithm (BSMA) (IEEE INFOCOM'95 1995; 1:377–385) have been proved to be capable of producing a multicast tree that has on average the lowest cost. However, such an excellent cost performance is accompanied with an extremely high time complexity. Recently, Feng et al. presented an alternative implementation of BSMA, which makes use of the latest research results on the delay-constrained least cost (DCLC) routing problem. Simulations indicate that, in comparison with the original implementation, the alternative implementation has a much lower time complexity with virtually identical cost performance, and it also runs much faster than many renowned heuristics such as KPP (IEEE/ACM Trans. Networking 1993; 1(3):286–292) and CAO (The design and evaluation of routing algorithms for real-time channels. Technical Report ICSI TR-94-024, International Computer Science Institute, University of California at Berkeley, June 1994). In this paper, we propose a brand new heuristic TCF, which is based on an idea called ‘tightest constraint first.’ TCF runs a DCLC heuristic only once for each destination and therefore has a provably low time complexity. We further propose an iterative heuristic ITCF, which uses TCF to obtain an initial tree and then gradually refines it. Extensive simulations demonstrate that, in the average sense, TCF can achieve a cost performance comparable to or better than that of BSMA, the cost performance of ITCF is even better than that of TCF, TCF runs approximately twice as fast as ITCF, and ITCF runs 2–4 times as fast as the best implementation of BSMA. Copyright © 2005 John Wiley & Sons, Ltd.