Dynamic Cache Allocation Algorithm and Replacement Policy for Reliable Multicast Network

Abstract

In the field of reliable multicast, the design of most current router cache management protocols is limited by network environment with single multicast group. With such protocols, expected targets can not be achieved in Internet environment in which a router is shared by a plurality of multicast groups. ACP is a router cache management protocol suitable for the environment shared by multiple groups. In Internet in which there exists frequently emerged bursty data flow, however, ACP allocation algorithm has such drawbacks as slow convergence and instable allocation results when the number of multicast groups or group members changes. Based on the design and implementation of a dynamic cache allocation algorithm as well as time-out algorithm and packet discard policy, a novel Adaptive and Active Reliable Multicast Protocol (AARM for short) suitable for large-scale multicast network was proposed. Simulation results show that AARM will achieve significant enhancement of performance for multicast network in terms of delayed restoration, bandwidth consumption and network throughput, etc. compared with ACP.

A Dynamic Cache Allocation Mechanism (DCAM) for Reliable Multicast in Information-Centric Networking

“…Reference [18] proposed an adaptive and active reliable multicast (AARM) protocol suitable for large-scale multicast networks, in which the cache allocation algorithm is divided into two steps. Step 1 is to estimate the average packet loss rate of the multicast group according to the exponentially weighted moving average of NACK, and pre-allocate cache space according to the ratio of the number of NACK of the multicast group to the total number of NACK.…”

“…Table 1 shows the parameter settings of our simulation experiments. We compared DCAM with AARM [18] and Equal [19] with regard to loss recovery delay, cache hit ratio, transmission completion time, and overhead under different cache strategies (LCE [9], CAPC [20], ProbCache [21], and Prob [22]). The parameter settings of CAPC, ProbCache, and Prob refer to [20].…”

“…In order to optimize the overall loss recovery performance of all multicast groups and improve the utilization of cache resources in the network, this paper comprehensively considers the packet loss probability, node depth, and transmission rate to design a dynamic cache allocation mechanism (DCAM). We also combine the three cache allocation mechanisms (DCAM, AARM [18], Equal [19]) with the four cache strategies (LCE [9], CAPC [20], ProbCache [21], Prob [22]), and conduct experiments to compare their performance. Experiments show that, among all combinations of cache allocation mechanisms and cache strategies, the combination of DCAM and CAPC performs best in terms of loss recovery delay, cache hit ratio, transmission completion time, and overhead.…”