Constructing delay-bounded multicast tree with optimal delay variation

Strict QoS requirements on delay and delay variation are required for many time-sensitive applications. For real-time multicast communications, a multicast tree satisfying both end-to-end delay and delay variation constraints often needs to be determined. Such a problem is called the DVBMT problem, and proved to be NP-complete [1]. Several algorithms have been developed to solve the problem based on a simple heuristic, the 'CBT+SP' heuristic, which uses the CBT (Core Based Tree) with the SP (Shortest Path). But this heuristic has several drawbacks, which can bring serious problems to algorithms based on it. Although attempt [2] has been made to overcome the problems, it actually has fatal drawbacks. In this paper, we first analyze the drawbacks of the heuristic, and explain why the improvement attempt does not work. Then we propose an efficient algorithm, also based on the 'CBT+SP' heuristic but with drawbacks completely removed. Experimental results show that, with the same time complexity, our algorithm outperforms the most efficient existing algorithm for the DVBMT problem in terms of multicast delay variation under multicast end-to-end delay constraint.