Sequantial IO Performance improvement in 6.1-STABLE

Mark Kirkwood markir at
Wed May 17 19:02:48 PDT 2006

Mike Jakubik wrote:
> Mark Kirkwood wrote:
>> Mark Kirkwood wrote:
>>> I recently upgraded my FreeBSD machines from 6.0R to 6.1-STABLE
>>> I noticed a pleasant surprise - improved disk IO!
>>> In 6.0R best sequential throughput I ever saw was 170 MB/s (32k block 
>>> size reads on an 4G file). In 6.1-STABLE I'm seeing 190 Mb/s (32k 
>>> block size reads on an 4G file).
>> Would have been even better if I'd spelled "sequential" correctly in 
>> the subject :-(...
>> I'm also seeing sequential write speed increasing from 120Mb/s to 
>> 140Mb/s.
>> I did wonder if this increased IO throughput is what is triggering the 
>> posts about data corruption in 6.1 - i.e. more throughput means more 
>> heat to dissipate!
> Or, maybe the file was in a different location on the disk. Outer and 
> inner tracks perform differently. Did you test with the same exact file 
> and the same file system?

Good point, but yes - I made sure that a pre-existing file was used for 
the *read* test.

About the *write* test - I am mistaken unfortunately, as I found some 
old test results showing 137Mb/s, I conclude that 140Mb/s is not 
significantly different from that!



More information about the freebsd-stable mailing list