Efficient Heuristic for Minimum Cost Trees Construction in Multi-Groups Multicast

Abstract

Abstract. We propose an efficient and fast heuristic algorithm (MGST) to construct minimum cost tree for each multicast-group in multi-groups multicast. The idea is to construct just one spanning shared-tree that allows minimum cost tree of each multicast-group to be computed from this shared-tree. MGST is compared with the low overheads Core-Based Tree algorithm (CBT) and approximate optimal cost Minimum Cost Tree heuristic (MCT) through simulations. The results showed that MGST constructs trees with average cost close to the approximate optimal (7% difference). The overheads incurred from MGST are also comparable to CBT and much lesser than MCT. In addition, MGST has also shown to construct minimum cost trees much faster than MCT and comparable to CBT. To further improve the performance, "trace back" method is introduced to MGST and denoted as MGST-T. MGST-T yields average cost performance of only 2% difference from the optimal and with almost the same overheads and trees' construction time as MGST.

Study on performance efficient trees for multicast communication

“…To solve this, the "trace back" method is used to eliminate such redundant nodes. This method was first introduced in [YPC08] and is adopted here by making use of the GLAGM instead of the conventional graph. The concept of "trace back" is very simple.…”

“…However, it was shown in [YPC08] that the "trace back" method eliminates the "ill effect" of the distributed approach of having redundant nodes and, hence, is able to obtain near optimal result. Note that although the "trace back" procedure may appear to be a centralized procedure where the best replacement link will be found to improve the overall cost, there is actually no computation of extra paths needed since only the SF of the direct link and the corresponding existing tree path between two nodes are required to be computed and compared.…”