Bug #42819 High concurrent load performs badly on multicore Solaris
Submitted: 13 Feb 2009 9:33 Modified: 5 Sep 2009 0:20
Reporter: Bernt Marius Johnsen Email Updates:
Status: No Feedback Impact on me:
None 
Category:MySQL Server Severity:S2 (Serious)
Version:6.0 (and 5.1?) OS:Solaris
Assigned to: CPU Architecture:Any

[13 Feb 2009 9:33] Bernt Marius Johnsen 
Description:
The attached program runs pretty fast on Linux computers (<10 seconds on my desktop) but uses 5-10 minutes on a 8-core/64-thread T2 Niagara with Solaris.

Applies to both Falcon and InnoDB storage engines. Seems that an unreasonable time is used LOCK_open. See the attached plockstat output.

How to repeat:
Run the attached Java program.
[13 Feb 2009 9:33] Bernt Marius Johnsen 
Java program

Attachment: NiagaraTest.java (text/x-java), 2.25 KiB.
[13 Feb 2009 9:34] Bernt Marius Johnsen 
plockstat output

Attachment: plockstat.txt (text/plain), 12.54 KiB.
[13 Feb 2009 10:18] Bernt Marius Johnsen 
More readable plockstat output (made with c++filt)

Attachment: plockstat2.txt (text/plain), 12.55 KiB.
[5 Sep 2009 23:00] Bugs System 
No feedback was provided for this bug for over a month, so it is
being suspended automatically. If you are able to provide the
information that was originally requested, please do so and change
the status of the bug back to "Open".