Monthly Archives: December 2005

Linux hard drive performance statistics

Seems like I manage to forget that hdparm exists for getting stats on disks in Linux. In any event, device and cache read timings are quite useful from time to time.

For example, this is from a celeron web server with an in-use 40gb ide:
hdparm -tT /dev/hda
Timing buffer-cache reads: 128 MB in 0.35 seconds =365.71 MB/sec
Timing buffered disk reads: 64 MB in 1.17 seconds = 54.70 MB/sec

This is from a dual proc Xeon web server with two 80gb IDE, this test is on the drive that’s not in use, though the in-use drive was similar (2080MB on buffer-cache instead:
hdparm -tT /dev/hdd
Timing buffer-cache reads: 2100 MB in 2.00 seconds = 1050.00 MB/sec
Timing buffered disk reads: 174 MB in 3.00 seconds = 58.00 MB/sec

This is not so good. It’s a drive that I thought I was maybe having problems with sitting in a PIII fileserver:
hdparm -tT /dev/hdc
Timing buffer-cache reads: 788 MB in 2.00 seconds = 394.00 MB/sec
Timing buffered disk reads: 16 MB in 3.24 seconds = 4.94 MB/sec

The operating system drive in this box is SCSI and here are its results:
Timing buffer-cache reads: 784 MB in 2.00 seconds = 392.00 MB/sec
Timing buffered disk reads: 90 MB in 3.03 seconds = 29.70 MB/sec

Still looking for writers

I’m still actively looking for writers for LinuxWorld Magazine. If you’d like to write for LWM, please contact me (steve.suehring (at) linuxworld (dot) com). The editorial calendar, roughly speaking and subject to change, is as follows for 2005:

Jan – Best of the Best (best products, best practices, and so on over
the past year)
Feb – Systems Administration/Management
Mar – Transitioning to Linux
Apr – The Wonderful World of LAMP
May – Security and Identity Management
Jun – Linux in Government and Non-Profits
Jul – Open Source: It Ain’t Just Linux
Aug – Linux in the Enterprise
Sep – Desktop and Gaming
Oct – Clustering
Nov – Mobile Linux
Dec – Linux in Telecom

More fun with dates


my @monthnames = qw/Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec/;
my @weekdays = qw/Sunday Monday Tuesday Wednesday Thursday Friday Saturday/;

my $nextweek = time+604800;

my ($sec,$min,$hour,$mday,$mon,$year,$dayname,$dayofyear) = gmtime($nextweek);
$year += 1900;

print “Content-type: text/html\n”;
print “Set-Cookie: testcookie=testcookievalue;”;
printf (“expires=%s, %02d-%s-%d %02d:%02d:%02d GMT”,$weekdays[$dayname],$mday,$m
print “\n\n”;
print “You’ve received a cookie

printf (“%s, %02d-%s-%d %02d:%02d:%02d GMT”,$weekdays[$dayname],$mday,$monthnamess[$mon],$year,$hour,$min,$sec);


The editorial group of LinuxWorld Magazine (the print version) are still working on a proposal to enable us to exert some influence over the content on The editors of have gone to great, great lengths to separate the content between and other Sys-Con properties. It’s all just too bad, imo, because we did leverage some good content from their other magazines and web sites.

I went back and re-read my comments over the last few days. I found it funny that one day I posted something asking people to relax but then the next day sent a flame myself. It’s a little difficult for us editors to relax about this since many of us have been involved in LinuxWorld Magazine since the beginning. Seeing someone tear down all of our hard work with poorly-researched stories is not so much fun. For the record, the story that I’m referring to is the Compuware v. IBM case that was corrected on Groklaw. That, together with the recent story from Maureen O’Gara detailing IBM’s alleged inability to find some other code is enough proof of a pattern to me. There seems to be a disconnect between the true facts and what shows up in MOG’s reports. Don’t believe me, read the Compuware story linked above.