Year: 1980
Type: conference-paper
Abstract: The performance of a process executing in a virtual memory environment is largely determined by the interaction between the process's memory referencing behavior and the system's memory management policies. The structure of the memory reference string, reflected in such quantities as the page fault rate, the stack depth distribution, and the working set size, has traditionally been expensive to measure because of the overhead in capturing every memory reference. This paper reports on the development and testing of a sampling technique designed to extract accurate measurements of reference string characteristics while recording only a part of the complete reference string. The cost of the measurements is controlled by the sampling rate. Results in this paper are based on experiments using synthetic reference strings from an LRU generative model.
Source: ACM Annual Conference
Authors: James Wittneben, Dennis Kafura
Institution: Iowa State University
Language: English
Subfield: Hardware and Architecture
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W2092044393