Enhancing the reliability of medium access control level wireless multicast
Increasing proliferation of wireless networks and their group-based applications such as multiplayer gaming and smart classroom has motivated research in reliable medium access (MAC) level multicast protocols. Existing Automatic Repeat Request (ARQ) based reliable MAC multicast protocols address different characteristics of wireless communication and differ in their throughput-efficiency and scalability. In order to understand the limits of throughput (delay) efficiency attainable by such protocols, this dissertation presents results of a comprehensive theoretical, analytical, and simulative study which explores the fundamental tradeoff between reliability and throughput (delay) taking into account media characteristics---including its broadcast nature, distance-dependent error, and multicast hidden terminal problem (MHTP). Additionally, reliable multicast protocols must deal with Feedback Implosion Problem (FIP) and the problem of increase in probability of transmission error with group size.
This research first theoretically analyzes common solutions to these problems under the assumption that only collisions cause packet corruption. It is proven that MHTP is prevented throughput-optimally only when each member of a group blocks its one-hop neighbor. A busy tone mechanism to implement optimal-blocking is proposed. Further, it is shown that some throughput-efficient FIP solutions are not reliable. Finally, retransmission error probability is reduced by prohibiting members that receive a packet correctly from successive retransmissions of that packet.
Tradeoffs between reliability, throughput and delay are investigated through analysis and modeling. The error assumption is relaxed to include packet corruption due to channel noise. It is shown that a single protocol configuration is not suitable for both delay- and loss-tolerant applications. The models allow the prediction of throughput efficiency and delay for a desired level of reliability.
Three protocols to meet various application requirements of throughput and reliability are proposed. The Improved Leader Based Protocol supports high-throughput traffic. The Tone Based Protocol is more reliable; its novel method of using channel state information for feedback makes it scalable. Multi-channel Multicast Feedback Protocol confirms data delivery to individual members.
Results from simulations using an error model that includes distance between stations, interference, capture, and mobility are provided to characterize the performance benefits of the proposed protocols in comparison to existing protocols.