[ovirt-devel] Introducing engine micro-benchmarks

Oved Ourfali oourfali at redhat.com
Thu Mar 23 19:56:28 UTC 2017


On Mar 23, 2017 12:16 PM, "Juan Hernández" <jhernand at redhat.com> wrote:

On 03/23/2017 11:05 AM, Roy Golan wrote:
> Lately we came across an interesting case where multi-host+mult-networks
> resulted editing a host to conclude in minutes. One assumption that was
> raised which we wanted to eliminate was that the decryption we perform
> on a fence agent password might be taking too long.
>
> So these days it's an easy task thanks to JMH[1], supplied by the jdk
> itself. I kickstarted [2] and added a 'DecryptionBenchmark', see the
> output as an example[3]
>
> Although The JMH project recommends to create a separate project I find
> it would be less trivial to people to contribute benchmarks let alone
> just playing around with current code they want to test.
>
> - So, (when it will be merged) you add your benchmark under
>  backend/manager/modules/benchmarks/MyBenchmark.java
>
> - run it from intellij using the jmh plugin exactly like a unit-test
> OR
> - mvn test -P benchmarks -pl org.ovirt.engine:benchmarks
> OR
> - java -jar benchmarks.jar
>
> I hope this would serve all of us well, please review and add your
> benchmarks.
>
> PS - this will not run in the CI atm.
>
> [1] http://openjdk.java.net/projects/code-tools/jmh/
> [2] https://gerrit.ovirt.org/74537 microbenchmarks: Introduce
> microbenchmarks using JMH
> [3] DecryptionBenchmark output (short version):
>
> # Run complete. Total time: 00:09:06
>
> Benchmark                             Mode  Samples    Score  Score
error  Units
> b.DecryptionBenchmark.decryption     thrpt       50  101.258
1.270  ops/s
> b.DecryptionBenchmark.encryption     thrpt       50  238.587
4.667  ops/s
> b.DecryptionBenchmark.decryption      avgt       50    0.010
0.000   s/op
> b.DecryptionBenchmark.encryption      avgt       50    0.004
0.000   s/op
> b.DecryptionBenchmark.decryption    sample     5544    0.010
0.000   s/op
> b.DecryptionBenchmark.encryption    sample    13067    0.004
0.000   s/op
> b.DecryptionBenchmark.decryption        ss       50    0.014
0.001      s
> b.DecryptionBenchmark.encryption        ss       50    0.009
0.001      s
>
> Process finished with exit code 0
>

Very interesting Roy, thanks.


Indeed!


What was the root cause of the host edit taking minutes?


No.
Juan - your original thought that the encrypt / decrypt are there due to
way too many queries was right. In the env there were around 1800 calls to
get all the hosts due to network checks that ran there.

We just wanted to check whether the encryption and decryption has any
impact in addition, and Roy looked into it and saw it doesn't impact much.


_______________________________________________
Devel mailing list
Devel at ovirt.org
http://lists.ovirt.org/mailman/listinfo/devel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ovirt.org/pipermail/devel/attachments/20170323/97f20f87/attachment-0001.html>


More information about the Devel mailing list