| performance.ms (15230) | performance.ms (19736) |
|---|---|
| 1.\" 2.\" ---------------------------------------------------------------------------- 3.\" "THE BEER-WARE LICENSE" (Revision 42): 4.\" <phk@login.dknet.dk> wrote this file. As long as you retain this notice you 5.\" can do whatever you want with this stuff. If we meet some day, and you think 6.\" this stuff is worth it, you can buy me a beer in return. Poul-Henning Kamp 7.\" ---------------------------------------------------------------------------- 8.\" | 1.\" 2.\" ---------------------------------------------------------------------------- 3.\" "THE BEER-WARE LICENSE" (Revision 42): 4.\" <phk@login.dknet.dk> wrote this file. As long as you retain this notice you 5.\" can do whatever you want with this stuff. If we meet some day, and you think 6.\" this stuff is worth it, you can buy me a beer in return. Poul-Henning Kamp 7.\" ---------------------------------------------------------------------------- 8.\" |
| 9.\" $Id$ | 9.\" $Id: performance.ms,v 1.1 1996/04/13 08:30:19 phk Exp $ |
| 10.\" 11.ds RH Performance 12.NH 13Performance 14.PP 15Performance for a malloc(3) implementation comes as two variables: 16.IP 17A: How much time does it use for searching and manipulating data structures. --- 40 unchanged lines hidden (view full) --- 58One of the names for this vague concept is ``current working set''. 59It has been defined many different ways over the years, mostly to 60satisfy and support claims in marketing or benchmark contexts. 61.PP 62For now we can simply say that it is the number of pages the process 63needs in order to run at a sufficiently low paging rate in a congested 64primary storage. 65(If primary storage isn't congested, this is not really important | 10.\" 11.ds RH Performance 12.NH 13Performance 14.PP 15Performance for a malloc(3) implementation comes as two variables: 16.IP 17A: How much time does it use for searching and manipulating data structures. --- 40 unchanged lines hidden (view full) --- 58One of the names for this vague concept is ``current working set''. 59It has been defined many different ways over the years, mostly to 60satisfy and support claims in marketing or benchmark contexts. 61.PP 62For now we can simply say that it is the number of pages the process 63needs in order to run at a sufficiently low paging rate in a congested 64primary storage. 65(If primary storage isn't congested, this is not really important |
| 66of course, but most systems would be better of using the pages for | 66of course, but most systems would be better off using the pages for |
| 67disk-cache or similar functions, so from that perspective it will 68always be congested.) | 67disk-cache or similar functions, so from that perspective it will 68always be congested.) |
| 69If this number of pages is to small, the process will wait for the its | 69If the number of pages is too small, the process will wait for its |
| 70pages to be read from secondary storage much of the time, if it's too 71big, the space could be used better for something else. 72.PP 73From the view of any single process, this number of pages is | 70pages to be read from secondary storage much of the time, if it's too 71big, the space could be used better for something else. 72.PP 73From the view of any single process, this number of pages is |
| 74"all of my pages", but from the point of view of the OS is should | 74"all of my pages", but from the point of view of the OS it should |
| 75be tuned to maximise the total throughput of all the processes on 76the machine at the time. 77This is usually done using various kinds of least-recently-used 78replacement algorithms to select page candidates for replacement. 79.PP 80With this knowledge, can we decide what the performance goal is for 81a modern malloc(3) ? 82Well, it's almost as simple as it used to be: --- 31 unchanged lines hidden --- | 75be tuned to maximise the total throughput of all the processes on 76the machine at the time. 77This is usually done using various kinds of least-recently-used 78replacement algorithms to select page candidates for replacement. 79.PP 80With this knowledge, can we decide what the performance goal is for 81a modern malloc(3) ? 82Well, it's almost as simple as it used to be: --- 31 unchanged lines hidden --- |