Monday Apr 22, 2013

Running 'perf top' on VirtualBox Oracle Linux guest

I was intrigued by this very informative blog and decided to run the same on my Virtualbox.

The linux ‘perf’ utility with EL 6 | Frits Hoogland Weblog

‘perf top’ sub-utility shows the kernel and userland functions which are active

In my first "select count(*) from dbaextents, dbaextents, dba_extents" run, I got different stack.

   PerfTop:    1846 irqs/sec  kernel:51.6%  exact:  0.0% [1000Hz cpu-clock],  (all, 2 CPUs)
             samples  pcnt function                     DSO
             _______ _____ ____________________________ __________________________________________________
             7085.00 42.8% kxhrHash                     /u01/app/oracle/product/11.2.0/dbhome_1/bin/oracle
             1812.00 10.9% qerhjSplitProbe              /u01/app/oracle/product/11.2.0/dbhome_1/bin/oracle
             1332.00  8.0% acpi_pm_read                 [kernel.kallsyms]
             1307.00  7.9% rworofprVeryFastUnpackRow    /u01/app/oracle/product/11.2.0/dbhome_1/bin/oracle
              778.00  4.7% sorgetqbf                    /u01/app/oracle/product/11.2.0/dbhome_1/bin/oracle

I thought high kernel CPU was caused by "acpi_pm_read". I googled and got some hints and
shut down the VM and changed the setting.

$ vboxmanage modifyvm Node01 --hpet on
$ vboxmanage modifyvm Node01 --acpi off

Now, I got the same stack but kernel is still 49.5%

   PerfTop:    1994 irqs/sec  kernel:49.5%  exact:  0.0% [1000Hz cpu-clock],  (all, 2 CPUs)
-----------------------------------------------------------------------------------------------------------------------

             samples  pcnt function                    DSO
             _______ _____ ___________________________ __________________________________________________

             9069.00 57.0% kxhrHash                    /u01/app/oracle/product/11.2.0/dbhome_1/bin/oracle
             2309.00 14.5% qerhjSplitProbe             /u01/app/oracle/product/11.2.0/dbhome_1/bin/oracle
             1532.00  9.6% rworofprVeryFastUnpackRow   /u01/app/oracle/product/11.2.0/dbhome_1/bin/oracle
             1011.00  6.4% smbget                      /u01/app/oracle/product/11.2.0/dbhome_1/bin/oracle
              811.00  5.1% sorgetqbf                   /u01/app/oracle/product/11.2.0/dbhome_1/bin/oracle
              695.00  4.4% qersoFetch                  /u01/app/oracle/product/11.2.0/dbhome_1/bin/oracle
              316.00  2.0% qerhjInnerProbeHashTable    /u01/app/oracle/product/11.2.0/dbhome_1/bin/oracle

I think the question is now whether the clock source change made the query finish quicker, to which I don't have the answer yet.

About

The views expressed on this blog are my own and do not necessarily reflect the views of Oracle.

Search

Categories
Archives
« April 2013 »
SunMonTueWedThuFriSat
 
1
2
3
4
5
6
7
9
10
11
12
13
14
20
21
23
24
27
28
29
    
       
Today