Predictive scalable and fair protocols for distributed systems
Ming-Hsing Chiu, Mostafa A. Bassiouni · Journal of International Crisis and Risk Communication Research · 1999
Efficient protocols that can successfully predict or adapt to the dynamic demands of system loads are becoming increasingly important to the design and implementation of modern distributed systems. This dissertation examines such protocols in the context of three distributed computation problems. The first problem is related to handoff blocking in cellular networks. Specifically, we present a new protocol that dynamically reserves channels for handoff requests based on predicting the motion of each mobile station. The scheme is motivated by the advancement of GPS/GSM-based positioning technology. Performance evaluation of our scheme has been obtained by a detailed simulation model as well as a semi-Markov analytical model. The second protocol deals with the scalability issue of large-scale distributed interactive simulation systems. We present a detailed analysis on the performance and reliability of prediction-based relevance filtering, a promising technique to improve the scalability of distributed simulation. The third protocol deals with interconnecting legacy LANs via an ATM backbone. The potential traffic mismatch between the source LANs and the destination LAN may cause congestion at the destination interworking-unit (IWU). We present an end-to-end credit-based feedback congestion control scheme that allows multiple traffic sources and provides fair allocation to the source traffic demands in a max-min fashion.