Browse Prior Art Database

Gating Techniques for Improved Input/Output Queuing

IP.com Disclosure Number: IPCOM000062348D
Original Publication Date: 1986-Nov-01
Included in the Prior Art Database: 2005-Mar-09
Document File: 1 page(s) / 12K

Publishing Venue

IBM

Related People

Duvall, KE: AUTHOR [+4]

Abstract

By varying the queued I/O quantum as a function of file size and memory constraints, and by using the Virtual Resource Manager (VRM) Device Driver queue depth and the process size/time paradigm, system load variance and the opportunity for memory exhaustion are reduced. In a UNIX* operating system which uses paging techniques to resource manage, the memory and paging functions can quickly become saturated with excessive I/O requests when working in a multiple simultaneous task environment. Further, it is necessary to minimize or eliminate the possibility of prequeued I/O operations which cause thrashing of memory between the paging disk and RAM (random-access memory). Two techniques have been developed which reduce these problems.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 73% of the total text.

Page 1 of 1

Gating Techniques for Improved Input/Output Queuing

By varying the queued I/O quantum as a function of file size and memory constraints, and by using the Virtual Resource Manager (VRM) Device Driver queue depth and the process size/time paradigm, system load variance and the opportunity for memory exhaustion are reduced. In a UNIX* operating system which uses paging techniques to resource manage, the memory and paging functions can quickly become saturated with excessive I/O requests when working in a multiple simultaneous task environment. Further, it is necessary to minimize or eliminate the possibility of prequeued I/O operations which cause thrashing of memory between the paging disk and RAM (random-access memory). Two techniques have been developed which reduce these problems. In the first technique, a means of self-regulation is employed which is based upon using file size aggregate (total demand for memory) and current memory constraints (cast out/page demand rate) to regulate the threshold of the queued I/O quantum. This "ranging" can vary based upon storage technology and geometry, but would typically range from 4 to 32 2K pages which are I/O queued for a given file. The kickoff threshold would require 2 pages to remain buffered, so that individual processing queues always remain primed. With the second technique, the VRM Disk Driver queue depth and the process size/time paradigm are used to minimize or eliminate the possibility of memory exhaustion and ...