[GRLUG] Disk benchmarking (WAS: Solid State Drives)

Adam Tauno Williams awilliam at whitemice.org
Thu Dec 4 11:21:03 EST 2008


> >> I happened to run some tests tonight while rebuilding my desktop
> >> computer. (Assembling a new box out of new and old components, adding
> >> a new drive and moving data from the old drives around)
> >> The command I ran tested sequential read performance for the first
> >> gigabyte of the disk:
> >> dd if=(device) of=/dev/null bs=1M count=1024

This test will mean more if you disable the drives read-ahead feature
(-A0 in hdparm).  This number will usually be much lower, but is more
correct if your data access turns out not to actually be sequential.

> >> My old 40GB IDE IBM Deskstar clocked in at 47 MB/s average, while my
> >> somewhat newer 7200RPM SATA2 500GB Seagate Barracuda clocked in at
> >> 67MB/s.  I didn't test my new 1TB SAMSUNG drive, as it was in the
> >> middle of an offline SMART test, and I wasn't sure how much the self
> >> test would impact the benchmark.

If your interesting in testing WRITE performance it is imperative to
disable the evil write cache prior to testing (-W0 in hdparm).
<http://www.jasonbrome.com/blog/archives/2004/04/03/writecache_enabled.html>

> >> I don't have a good way of testing seek times, though.  Does anyone
> >> know of a good, free hard drive benchmark tool I can run from an
> >> Ubuntu live CD?  I'm not going to build a Bart PE disk for the
> >> purpose, nor can I put the drives in a Windows box.

hdparm numbers are interesting, but don't tell you much about how
something is going to perform in real-life.



More information about the grlug mailing list