Files in this item



application/pdf8422796.pdf (4MB)Restricted to U of Illinois
(no description provided)PDF


Title:Automatic File Migration: Reload Queueing Simulations and Markov Models
Author(s):Peterson, James Stuart
Department / Program:Computer Science
Discipline:Computer Science
Degree Granting Institution:University of Illinois at Urbana-Champaign
Subject(s):Computer Science
Abstract:Automatic migration of files to a mass storage device can save substantial amounts of disk space while causing minimal user inconvenience.
Measurements of the user inconvenience have generally been done in terms of demand curves or miss ratios, which are useful for measuring the overall performance of a migration system, but which convey little information about the real time delay a user suffers while waiting for a reload. For example, given a fixed number of files reloaded, user delay can vary greatly depending on whether his files are all on one media or on many, a condition strongly affected by the migration policy. In addition, delay is strongly influenced by the characteristics of the mass storage device, such as media size, mount time, etc., and by the queuing policy. Using data taken from a real system, we are able to simulate the effects of the media assignment policy, the migration policy, mass storage characteristics, and the queuing policy on user delay. Results will be presented for both the CDC 38500 and the proposed Phillips optical disk.
The user inconvenience can also be measured from the miss ratio, which depends strongly on the interaction between the migration policy and the file access patterns. We have developed a mathematical model of the migrate/access interactions and evaluated its accuracy using three file access trace databases varying in length from 233 to 384 days. To improve the accuracy of the model we have developed a method for measuring the locality of reference present in the file access traces. This model can be used either by itself or in conjunction with the user delay study to estimate system performance from a relatively small amount of new data.
Issue Date:1984
Description:128 p.
Thesis (Ph.D.)--University of Illinois at Urbana-Champaign, 1984.
Other Identifier(s):(UMI)AAI8422796
Date Available in IDEALS:2014-12-15
Date Deposited:1984

This item appears in the following Collection(s)

Item Statistics