Abstract

This thesis presents the design, implementation and evaluation of a novel cooperative caching algorithm based on the bloom filter data structure. The new algorithm uses a decentralized approach to resolve the problems that prevent the existing solutions from being scalable. The problems consist of an overloaded manager, a communication overhead among clients, and a memory overhead on the global cache. The new solution reduces the manager load and the communication overhead by distributing the global cache information among cooperating clients. Thus, the manager no longer maintains the global cache. Furthermore, the memory overhead is decreased due to a bloom filter data structure. The bloom filter saves memory space in the global cache and makes the new algorithm scalable. The correctness of the research hypothesis is verified by running experiments on the caching algorithms. The experiment results demonstrate that the new caching algorithm maintains a low block access time as existing algorithms. In addition, the new algorithm decreases the manager load by the factor of nine. Moreover, the communication overhead is reduced by nearly a factor of six as a result of distributing the global cache to clients. Finally, the results show a significant reduction in the memory overhead which also contributes to the scalability of the new algorithm.

Library of Congress Subject Headings

Cache memory; Data transmission systems; Electronic data processing--Distributed processing; Computer algorithms--Evaluation

Publication Date

2011

Document Type

Thesis

Department, Program, or Center

Computer Science (GCCIS)

Advisor

Kwon, Minseok

Comments

Note: imported from RIT’s Digital Media Library running on DSpace to RIT Scholar Works. Physical copy available through RIT's The Wallace Library at: TK7895.M4 S43 2011

Campus

RIT – Main Campus

Download

Share

COinS