> To stress test the CPU, RAM and disks (well, mostly disk access), my personal > > favorite is what a co-worker, UNIX guru showed me. Set up the following shell > > scripts and run 5-20 (or more) copies of each of them depending on the size > of your hardware. > > Script 1 > ----------------------- > #!/bin/sh > > # Comment in and out according to your HW > # For SCSI > cp /dev/sda /dev/null > > #For IDE > cp /dev/hda /dev/null > -----------------------
I don't know that sequential copies actually exercise very much. Even 20 in parallel - the data from the first should be cached and I'd expect them to keep in step for however long it takes. The first to read a new sector blocks and waits for the others to catch up. > > Script 2 > ----------------------- > #!/bin/sh > > find / | cpio -o | compress > /dev/null > # replace "compress" with "gzip" or "bzip2" if you like > ----------------------- > > In the past, I usually run more of script2 then script1 so that memory and > CPU get eaten up fairly well. Throwing in a linux kernel source compile > wouldn't hurt either. Again, they're reading the same data. A couple of years ago I had 64 Mbytes RAM on my Pentium II. Now it has 256 Mbytes "because it's cheap." The probability of data being kept in cache long enough for all processes to use it is high. > > Note: you may have to tune up your system to have these scripts run > well; see http://linuxperf.nl.linux.org or other Linux performance tuning sit > e > > For network you can add some copying to or from NFS or samba filesystems or > use something like iperf or ttcp (correct name?). Whether this will stress the network depends on how good your drives are. I get around 5 Mbytes/sec of my Pentium systems disks, much less than the 100 Mbit network can deliver. The caveats I mentioned before remain - it's not scientific, it's useful as a rough measure of whether it works, but not a good guide to how well. For that you need well-designed benchmarks, preferably tailored to your own usage of the computer equipment. -- Cheers John Summerfield Microsoft's most solid OS: http://www.geocities.com/rcwoolley/ Note: mail delivered to me is deemed to be intended for me, for my disposition. _______________________________________________ Redhat-devel-list mailing list [EMAIL PROTECTED] https://listman.redhat.com/mailman/listinfo/redhat-devel-list
