A Receiver-Coordinated Approach for Throughput Aggregation in High Bandwidth Multicast

Abstract

In application-level high bandwidth multicast (HBM), physical links can be shared by multiple long-lived unicast flows. We identify several data transfer patterns which can cause suboptimal bandwidth usage of narrow links and which have not been clearly identified in previous solutions for application-level HBM.We propose a distributed solution to avoid these problematic patterns, with which end systems are coordinated and each is responsible to forward a bounded amount of data. Consequently, the outgoing traffic of each end system is balanced and limited. It avoids congestion due to merging unicast flows, which increases the utilization of the narrow links. Receivers that are close by topologically request their data in a disjoint and coordinated fashion, which leads to much reduced duplicated data at the narrow links. Simulation results show that our solution can achieve higher throughputs at the receivers, which is due to more efficient utilization of the narrow links' bandwidth, than meshbased or multiple-tree approaches.

Dynamic Multicast in Overlay Networks with Linear Capacity Constraints

“…Recent work by Tsang et al [16] identified a number of data transfer patterns that cause suboptimal bandwidth usage of bottleneck links. They proposed a solution which involves topologically close receivers to request data in a coordinated way.…”

Adaptive Multicast Tree Construction for Elastic Data Streams

“…Another difficulty in multicasting is the issue of cross-traffic where different multicast sessions interfere with one another due to commonly shared physical network. Tsang et al [9] applies a receiver-coodinated approach to overcome the issue of cross-traffic in multicasting using multicast trees.…”