Title: A model-based scalable reliable multicast transport protocol for satellite networks
Authors:
Prawit Chumchu, Roksana Boreli, Aruna Seneviratne
Abstract:
In this Paper, we design a new scalable reliable multicast transport protocol for satellite networks (RMT). This paper is the extensions of paper in [16]. The proposed reliable multicast transport protocol does not require inspection and/or interception of packets at intermediate nodes. The protocol would not require any modification of satellites, which could be bent-pipe satellites or onboard processing satellites. We divide the proposed protocol in 2 parts: error control part and congestion control part. In error control part, we intend to solve feed back implosion and improve scalability by using a hybrid of ARQ (Auto Repeat Request) and adaptive forward error correction (AFEC). The AFEC algorithm adapts proactive redundancy levels following the number of receivers and packet loss rate. This leads to the number of transmissions and the number of feed back signals are virtually independent of the number of receivers. Therefore, wireless link utilization used by proposed protocol is virtually independent of number of multicast receivers. In congestion control part, the proposed protocol employs a window-based congestion control scheme, which is optimized for satellite networks. To be fair to other traffics, the congestion control mimics congestion control in the well-known Transmission Control Protocol (TCP) which relies on “packet conservation” principle. To reduce feedback implosion, only a few receivers, ACKers, are selected to report receiving status. In addition, in order to avoid “drop-to-zero” problem, we use a simple wireless loss filter algorithm. This loss filter algorithm significantly reduces the probability of the congestion window size unnecessarily reduced because of wireless spatially uncorrelated losses. Furthermore, to improve achievable throughput, we employ slow start threshold adaptation based on estimated bandwidth. Our congestion control also deals with variations in network conditions by dynamically electing ACKers