Exploring gate-limited analytical models for high-performance network storage servers

Mapp, Glenford E. ORCID: https://orcid.org/0000-0002-0539-5852, Thakker, Dhawal and Gemikonakli, Orhan ORCID: https://orcid.org/0000-0002-0513-1128 (2011) Exploring gate-limited analytical models for high-performance network storage servers. Journal of Computer and System Sciences, 77 (5) . ISSN 0022-0000 [Article] (doi:10.1016/j.jcss.2010.08.002)

Preview

PDF (In press version) - Final accepted version (with author's formatting) Download (1MB) | Preview

Abstract

Gate-limited service is a type of service discipline found in queueing theory and can be used to describe a number of operational environments, for example, large transport
systems such as buses, trains or taxis, etc. Recently, there has been the observation that such systems can also be used to describe interactive Internet Services which use a Client/Server interaction. In addition, new services of this genre are being developed for the local area. One such service is a Network Memory Server (NMS) being developed here at Middlesex University. Though there are several examples of real systems that can be modelled using gate-limited service, it is fair to say that the analytical models which have been developed for gate-limited systems have been difficult to use, requiring many iterations before practical results can be generated. In this paper, a detailed gate-limited bulk service queueing model based on Markov chains is explored and a numerical solution
is demonstrated for simple scenarios. Quantitative results are presented and compared with a mathematical simulation. The analysis is used to develop an algorithm based on
the concept of optimum operational points. The algorithm is then employed to build a high-performance server which is capable of balancing the need to prefetch for streaming
applications while promptly satisfying demand misses. The algorithm is further tested using a systems simulation and then incorporated into an Experimental File System (EFS)
which showed that the algorithm can be used in a real networking environment.

Actions (login required)

View Item

Statistics

DownloadsShow export options

Export as CSVXMLJSON

Activity Overview

6 month trend

414Downloads

6 month trend

490Hits

Additional statistics are available via IRStats2.