Bug 376448 - mysql poor performance
Summary: mysql poor performance
Status: RESOLVED NORESPONSE
Alias: None
Product: openSUSE 10.3
Classification: openSUSE
Component: Kernel (show other bugs)
Version: Final
Hardware: i586 openSUSE 10.3
: P5 - None : Major with 5 votes (vote)
Target Milestone: ---
Assignee: Nick Piggin
QA Contact: E-mail List
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2008-04-02 18:04 UTC by Samuel Kroslak
Modified: 2008-09-10 10:26 UTC (History)
2 users (show)

See Also:
Found By: Customer
Services Priority:
Business Priority:
Blocker: ---
Marketing QA Status: ---
IT Deployment: ---

Attachments
sql-bench (16.42 KB, application/x-zip-compressed)
2008-04-02 18:04 UTC, Samuel Kroslak 		Details
View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Samuel Kroslak 2008-04-02 18:04:06 UTC 
Created attachment 205787 [details]
sql-bench

I want to continue with bug id=335584 which was marked as resolved because no answer. I can see very similar behavior on my OpenSUSE 10.3 box with default mysql  5.0.45 that can be found in repository. I also have regular installation of OpenSUSE with no special kernel or kernel customizations. I tried to run ./run-all-tests --server=mysql --log --user=root --password=*** command from mysql-bench package and I attached results. I'm not very keen in DB stuff, but I certainly know that the databases I'm running on my box are X-times slower than in Windows or other linux OS on same HW configuration.
Comment 1 Nick Piggin 2008-05-23 01:59:07 UTC 
Thanks for keeping at it.

A couple of things spring to mind. Firstly, I believe SUSE kernels try to do the right thing WRT disk writeback caching, wheras some other distros and I think even the upstream kernel doesn't do by default. Also, we default to data=ordered on journalling filesystems.

Could you try ext3 mounted with -o barrier=0 data=writeback ?

The other thing is that we've had a few scalability problems in the Linux kernel in general with a funny interaction between the kernel, mysql, and glibc's malloc/free. Although this only seems to cause a problem on very CPU bound workloads and servers with about 8 or more cores.
Comment 2 Nick Piggin 2008-07-31 03:19:14 UTC 
Any updates on this issue?
Comment 3 Samuel Kroslak 2008-07-31 06:09:22 UTC 
sorry guys, but I no longer use OpenSUSE, so I can't help right now, maybe I'll return to it in 1-2 months and then I can try to play with it and see what I can find out
Comment 4 Nick Piggin 2008-09-10 10:26:29 UTC 
Well I'll close this bug now. Thanks anyway for reporting, and sorry I couldn't help this time.