Characterizing achievable multicast rates in multi-hop wireless networks
Proceedings of the 6th ACM international symposium on Mobile ad hoc networking and computing - MobiHoc '05
In this paper, we consider the multicast throughput optimization problem in multi-hop wireless networks. Given a source, and a set of receivers, we would like to find the set of multicast trees and a schedule such that the rate that the source can multicast to the receivers is maximized. We consider two transmission models: broadcast and unicast. In the broadcast model, a transmission is received by multiple downstream nodes in a multicast tree. In the unicast model, a separate transmission has
... to be sent to each downstream node. We consider the fundamental constraint that a node can not be involved in multiple communications at the same time. We consider two multicast models: a single multicast tree per session and multiple multicast tree per session. In the single multicast tree case, (1) for the unicast model, we show that the problem is NP-hard and it is not approximable to a factor better than 1.5; we then give a 1.5-approximation algorithm if all links have the same data rate, a 5-approximation algorithm if all nodes have the same transmission power and a 24-approximation algorithm for a realistic heterogeneous ad hoc network where nodes can have different transmission power. (2) for the broadcast model, we show that the problem is NP-hard and it is not approximable to a factor better than 2; we then give a simple 2-approximation algorithm to find the multicast tree and the transmission schedule. In the multiple multicast tree case, (1) for the unicast model, we show that the problem is APX-hard, and give a 1.5ρ-approximation where ρ is the best approximation ratio of the minimal cost Steiner tree problem; (2) for the broadcast model, our results indicate that the problem is hard, may not be approximable within a factor better than log(n) where n is the number of multicast receivers. Our evaluation shows that the throughput achieved by our algorithms is much better than both the throughput achieved by using pruned shortest path tree and by using optimal unicast.