Date: Thu, 24 Jun 1999 12:29:32 -0700 (PDT) From: schuerge@cs.uni-sb.de To: freebsd-gnats-submit@freebsd.org Subject: kern/12379: Serious performance with CPU-intensive processes in the background Message-ID: <19990624192932.C135914C2F@hub.freebsd.org>
next in thread | raw e-mail | index | archive | help
>Number: 12379 >Category: kern >Synopsis: Serious performance with CPU-intensive processes in the background >Confidential: no >Severity: serious >Priority: medium >Responsible: freebsd-bugs >State: open >Quarter: >Keywords: >Date-Required: >Class: sw-bug >Submitter-Id: current-users >Arrival-Date: Thu Jun 24 12:30:01 PDT 1999 >Closed-Date: >Last-Modified: >Originator: Thomas Schürger >Release: FreeBSD 4.0-CURRENT >Organization: University of Saarbrücken, Germany >Environment: FreeBSD starfire.heim-d.uni-sb.de 4.0-CURRENT FreeBSD 4.0-CURRENT #0: Tue Jun 15 13:36:03 CEST 1999 schuerge@starfire.heim-d.uni-sb.de:/usr/src/sys/compile/STARFIRE i386 Running on dual PII/400 with 256 MB RAM, UW-SCSI and UDMA drive. >Description: FreeBSD's system performance is greatly affected when running a very CPU-intensive (not I/O-intensive) process in the background. I have tested it with RC5DES (misc/rc5des) and seti@home (astro/setiathome), but it seems to be a general problem. Note that these two programs nearly don't do any I/O at all (just receiving/sending blocks after some hours or days). When such a process is running in the background (niced to +20), other I/O-intensive processes in the foreground suffer a lot from it. Transfers over the LAN become a lot slower, hardisk accesses are slower and also floppy disk transfers become slower. I have a directory on my machine that is exported via NFS. When nothing's in the background, an "ls" in the directory(containing hundreds of files) on the remote machine takes about 0.5 seconds. With a CPU-intensive process in the background on my machine, it takes 3 to 4 seconds. Network transfers drop from e.g 500 KB/sec to 250-300 KB/sec and updating the ports takes about twice as long when such process is present. I haven't done more tests, but it seems to me that FreeBSD's performance noticably drops down to a level that is not really acceptable. I haven't heard of such problems on other Unix variants. >How-To-Repeat: Start some long-runner in the background and compare network I/O speeds to the system not running the process. >Fix: >Release-Note: >Audit-Trail: >Unformatted: To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-bugs" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19990624192932.C135914C2F>