Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!yoyo.aarnet.edu.au!fang.dsto.defence.gov.au!foxhound.dsto.gov.au!munnari.oz.au!spool.mu.edu!usenet.eel.ufl.edu!news.mathworks.com!tank.news.pipex.net!pipex!news.sprintlink.net!in1.uu.net!news.acadia.net!bhb21.acadia.net!user From: gary@first.acadia.net (Gary Robinson) Newsgroups: comp.unix.bsd.freebsd.misc Subject: Re: "An HTTP software server can pummel a CPU..." Date: Sun, 17 Sep 1995 09:12:31 -0500 Organization: First Software, Inc. Lines: 37 Message-ID: <gary-1709950912310001@bhb21.acadia.net> References: <gary-1309951409030001@bhb17.acadia.net> <439ooa$fdc@bonnie.tcd-dresden.de> NNTP-Posting-Host: bhb21.acadia.net Joerg, Thanks for the excellent explanation about priorities in Unix. Gary In article <439ooa$fdc@bonnie.tcd-dresden.de>, joerg_wunsch@uriah.heep.sax.de wrote: > Gary Robinson <gary@first.acadia.net> wrote: > > >[ Now that I think of it, I maybe I do remember hearing before that Unix > >didn't allow priorities ]. > > Unix is considered a time-sharing o/s (as opposed to a real-time o/s) > and as such has no static priority handling (and hence no guaranteed > response time). However, there is a mechanism to modify the initial > priority of a process, called "nice" (see nice(1) and nice(3)). > > Further, the basics of the scheduler are arranged so that it would > lower the priority of processes that did consume much CPU, while it > will raise the priority of processes that used to have long sleep > times (i.e., were waiting for I/O to complete), in order to give them > a better interactive response behaviour. > > You can also limit the amount of CPU a process will be allowed to get > _in total_, it will be killed after exceeding this amount (but you can > arrange it to send the process a SIGXCPU first). See [sg]etrlimit(2) > and the csh-builtin command limit. > -- > cheers, J"org private: joerg_wunsch@uriah.heep.sax.de > http://www.sax.de/~joerg/ > > Never trust an operating system you don't have sources for. ;-)