[Bioclusters] Tool to benchmark disk IO?

[Joe Landman]

Ok ....

Dan Bolser wrote:

>On Fri, 24 Sep 2004, Joe Landman wrote:
>
>  
>
>>Hi Dan:
>>
>> What size memory do you have on the test machine?
>>    
>>
>
>1548036k
>  
>

Use at least 3 GB as your test size (you used 2GB).  Linux is agressive 
on buffer caching in 2.4 (almost to a fault).

>Here is the new result for scsi/ide/nfs
>
>     -------Sequential Output-------- ---Sequential Input-- --Random--
>     -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
>Machi MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
>scs 2047 15781 91.0 33870 20.5 14318  6.1 12880 66.7 33974  8.3 234.0  1.3
>
>
>Seeker 1...Seeker 2...Seeker 3...start 'em...done...done...done...
>         -------Sequential Output-------- ---Sequential Input-- --Random--
>         -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
>Machi MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
>ide 2047 17293 96.4 42104 28.2 18361  8.5 16149 83.9 60721 15.6 141.6  0.7
>  
>

So your IDE is about 9.6% faster on per char writes, 24% faster on block 
writes,  28% faster on rewrites,  25% faster on per char reads, and ~79% 
faster on block reads.  What technology/make/manufacturer are the disks?

>
>         -------Sequential Output-------- ---Sequential Input-- --Random--
>         -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
>Machi MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
>nfs 2047  4863 69.2 10959  8.0 11380  8.2  7000 100.0 1520816 100.1 2901.2 17.4
>  
>

I am having lots of trouble with your block reads.  It appears to be 
reading at 1.5 GB/s, while writing at 0.01 GB/s.   This is either rather 
asymmetric, or possibly wrong.  What are your mount options?  What type 
of NFS server and what type of network?  I would tremendously respect 
any NFS device that  can pump out 1.5 GB/s, though I would be hard 
pressed to find a single connection that could support that speed.  I 
would expect over fast ethernet that your sequential input would be 
close to 11000-13000 for a really well tuned server and client, and 
60000-80000 for a really well tuned gigabit connected system.  The 1.5M 
number is 20x  what I expect.  I am intrigued ...

>
>
>
>  
>
>> I usually use 2-4 GB for my tests on machines with 1 GB or less 
>>memory.  Some of the numbers look a little off.  The SCSI rewrite speed 
>>is 4x the IDE rewrite speed, and the IDE seems to be doing 133MB/s on 
>>sequential writes (cool, but I don't believe it unless you have a 
>>multiway RAID0, or an IDE raid card with a big honking cache ...,  I 
>>have hit a sustained 110 MB/s on 2 way IDE RAID0's properly tuned).
>>
>>Joe
>>
>>Dan Bolser wrote:
>>
>>    
>>
>>>FYI:
>>>
>>>Here is my local SCSI and IDE disk for comparison (size 104857600 again)
>>>
>>>
>>>
>>>          -------Sequential Output-------- ---Sequential Input-- --Random--
>>>          -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
>>>Machine MB K/sec %CPU K/sec %CPU  K/sec %CPU K/sec %CPU  K/sec %CPU    /sec %CPU
>>>scsi   100 10537 76.3 44958 32.5 108986 44.7 12314 74.6 418500 98.1 22027.8 88.1
>>>
>>>
>>>
>>>          -------Sequential Output-------- ---Sequential Input-- --Random--
>>>          -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
>>>Machine MB K/sec %CPU  K/sec %CPU K/sec %CPU K/sec %CPU  K/sec %CPU    /sec %CPU
>>>ide    100 13921 89.5 133969 81.1 27470 11.8 14553 87.6 430990 96.8 21952.2 93.3
>>>
>>>
>>>
>>>          -------Sequential Output-------- ---Sequential Input-- --Random--
>>>          -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
>>>Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU   K/sec %CPU   /sec %CPU
>>>nfs    100  4240 25.9  5437  2.1  4780  2.4 19595 100.1 566942 99.7 2023.0 10.1
>>>
>>>
>>>_______________________________________________
>>>Bioclusters maillist  -  Bioclusters@bioinformatics.org
>>>https://bioinformatics.org/mailman/listinfo/bioclusters
>>> 
>>>
>>>      
>>>
>>    
>>
>
>_______________________________________________
>Bioclusters maillist  -  Bioclusters@bioinformatics.org
>https://bioinformatics.org/mailman/listinfo/bioclusters
>  
>

-- 
Joseph Landman, Ph.D
Scalable Informatics LLC,
email: landman@scalableinformatics.com
web  : http://scalableinformatics.com
phone: +1 734 612 4615