Quantcast

Linux - Odd results when compared to Windows 2003

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Linux - Odd results when compared to Windows 2003

Matthew Wilby
I've been running some basic disk benchmarking for local SCSI U320, FC & iSCSI attached storage, to compare performance under Windows & Linux.

All tests have been carried out using the same equipment, and same test setup. The only difference is the OS. The test bed server is dual boot Win2k3 and Linux (SuSE SLES 9 - 2.6.5 kernel). 

The results I get back under Linux are bizarre! Below are a few of the results......
 
 
100% rand. write

Windows 2003 - 27 439
Linux - 629 10071

100% seq. write

Windows 2003 - 67 1068
Linux - 627 100033


100% rand. read

Windows 2003 - 30 491
Linux - 1777 28431


100% seq. read

Windows 2003 - 67 1079
Linux - 1769 28312

20% read 80% write
(50% seq./write)

Windows 2003 - 32 517
Linux - 687 10996


50% read/write
(50% seq./write)

Windows 2003 - 26 419
Linux - 903 14452
 
The first figure for each test is MB/sec and the second in IOPS. These are tests from local SCSI U320 drives.
 
How can I accurately benchmark I/O between Windows & Linux with IOmeter? Is this a bug or can these figures be explained?
 
Thanks.
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

RE: Linux - Odd results when compared to Windows 2003

Coyne, Kate

Matthew,

 I looks like, on Linux, you have a good amount of cache on the system which would effect the iometer on linux numbers to be inflated.  Did you turn off the cache(system) so that it’s more a raw disk test.  I was having problems with Random Writes on Linux and have not been able to get good consistant numbers here…and I only have 512MB of RAM so that there is minimal cache hits plus I use a 2Gb file.

Hope this helps.

Kate

 

-----Original Message-----
From: [hidden email] [mailto:[hidden email]] On Behalf Of Matthew Wilby
Sent: Tuesday, September 27, 2005 12:27 PM
To: [hidden email]
Subject: [Iometer-user] Linux - Odd results when compared to Windows 2003

 

I've been running some basic disk benchmarking for local SCSI U320, FC & iSCSI attached storage, to compare performance under Windows & Linux.

All tests have been carried out using the same equipment, and same test setup. The only difference is the OS. The test bed server is dual boot Win2k3 and Linux (SuSE SLES 9 - 2.6.5 kernel). 

The results I get back under Linux are bizarre! Below are a few of the results......

 

 

100% rand. write

Windows 2003 - 27 439
Linux - 629 10071


100% seq. write

Windows 2003 - 67 1068
Linux - 627 100033


100% rand. read

Windows 2003 - 30 491
Linux - 1777 28431


100% seq. read

Windows 2003 - 67 1079
Linux - 1769 28312


20% read 80% write
(50% seq./write)

Windows 2003 - 32 517
Linux - 687 10996


50% read/write
(50% seq./write)

Windows 2003 - 26 419
Linux - 903 14452

 

The first figure for each test is MB/sec and the second in IOPS. These are tests from local SCSI U320 drives.

 

How can I accurately benchmark I/O between Windows & Linux with IOmeter? Is this a bug or can these figures be explained?

 

Thanks.

Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Linux - Odd results when compared to Windows 2003

Ming Zhang
In reply to this post by Matthew Wilby
On Tue, 2005-09-27 at 17:27 +0100, Matthew Wilby wrote:
> I've been running some basic disk benchmarking for local SCSI U320, FC
> & iSCSI attached storage, to compare performance under Windows &
> Linux.
>
> All tests have been carried out using the same equipment, and same
> test setup. The only difference is the OS. The test bed server is dual
> boot Win2k3 and Linux (SuSE SLES 9 - 2.6.5 kernel).  
>

* what is you HW spec?
* when test a U320 drive, you test raw disk or a disk with file system?
* if test on file system, what is the size of your testing file?
* what iometer version u are using?



> The results I get back under Linux are bizarre! Below are a few of the
> results......
>  
>  
> 100% rand. write
>
> Windows 2003 - 27 439
> Linux - 629 10071
>
>
> 100% seq. write
>
> Windows 2003 - 67 1068
> Linux - 627 100033
>
>
> 100% rand. read
>
> Windows 2003 - 30 491
> Linux - 1777 28431
>
>
> 100% seq. read
>
> Windows 2003 - 67 1079
> Linux - 1769 28312
>
>
> 20% read 80% write
> (50% seq./write)
>
> Windows 2003 - 32 517
> Linux - 687 10996
>
>
> 50% read/write
> (50% seq./write)
>
> Windows 2003 - 26 419
> Linux - 903 14452
>
>  
> The first figure for each test is MB/sec and the second in IOPS. These
> are tests from local SCSI U320 drives.
>  
> How can I accurately benchmark I/O between Windows & Linux with
> IOmeter? Is this a bug or can these figures be explained?
>  
> Thanks.



-------------------------------------------------------
This SF.Net email is sponsored by:
Power Architecture Resource Center: Free content, downloads, discussions,
and more. http://solutions.newsforge.com/ibmarch.tmpl
_______________________________________________
Iometer-user mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/iometer-user
Loading...