Off-campus UMass Amherst users: To download dissertations, please use the following link to log into our proxy server with your UMass Amherst user name and password.
Non-UMass Amherst users, please click the view more button below to purchase a copy of this dissertation from Proquest.
(Some titles may also be available free of charge in our Open Access Dissertation Collection, so please check there first.)
Scheduling network processing on multimedia and multiprocessor servers
Abstract
Techniques for the efficient management of network communication are of increasing importance as computer communication networks become more widespread and distributed applications move into the mainstream. The focus of this dissertation is on multimedia and multiprocessor computer end-systems, and the design of the network software on these machines. In particular, we explore scheduling techniques for improving performance, motivated by practical considerations arising in a wide variety of real-world communication scenarios. In the first part of the dissertation, we address the problem of designing end-system network software in a manner that avoids the high memory overheads found on many modern multiprocessors. We explore the application of processor-cache affinity scheduling, a technique for managing processing and resources for higher cache hit rates, in the parallelized network software on a commercial shared-memory multiprocessor. We find that affinity-based scheduling can significantly reduce the communication delay associated with network protocol processing, and in some cases substantially increase protocol processing throughput. In addition, we compare the performance of the two protocol parallelization alternatives which perform best on shared-memory multiprocessors, yet have fundamentally different caching behaviors. Our conclusions are shown to hold over a wide range of workload characteristics and system configurations. In the second part of the dissertation, we address the problem of designing end-system network software in a manner that reduces the high transmission rate variability of distributed applications sending stored video. We develop an optimal smoothing algorithm for achieving the greatest possible reduction in rate variability when transmitting stored video across a network to a client with given buffer size. We show a formal proof of optimality, demonstrate the performance of the algorithm on a set of long MPEG-1 video traces, and evaluate the impact of optimal smoothing under two well-known network service models specifically designed for the transmission of real-time video: Deterministic Guaranteed service and Renegotiated Constant Bit Rate (RCBR) service. Under both models, we find the impact of optimal smoothing to be dramatic.
Subject Area
Computer science
Recommended Citation
Salehi, James Darius, "Scheduling network processing on multimedia and multiprocessor servers" (1996). Doctoral Dissertations Available from Proquest. AAI9709646.
https://scholarworks.umass.edu/dissertations/AAI9709646