mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Guangya Liu <gyliu...@gmail.com>
Subject Re: Review Request 50387: Updated allocated number of range to 100 for sorter benchmark test.
Date Mon, 25 Jul 2016 09:21:50 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/50387/
-----------------------------------------------------------

(Updated 七月 25, 2016, 9:21 a.m.)


Review request for mesos and Benjamin Mahler.


Summary (updated)
-----------------

Updated allocated number of range to 100 for sorter benchmark test.


Repository: mesos


Description
-------

If setting port range fragments as 16, then the performance of
`AgentAndClientCount/Sorter_BENCHMARK_Test.FullSort/35` does not
improve much after the resource validation was removed when doing
port range `+=` operations.

We should increase the port range fragments to increase the workload
of the benchmark test for sorter port range fragments calculation.

After the port range fragments was increased to 100, we can see that
the performance of this benchmark test was improved by 10x when adding
port resources with and without resource validation.


Diffs
-----

  src/tests/sorter_tests.cpp ade356cbbba5b93a6d3e5c9de30eefd3982d15c1 

Diff: https://reviews.apache.org/r/50387/diff/


Testing
-------

make
make check

Without resource validation:
```
[==========] Running 1 test from 1 test case.
[----------] Global test environment set-up.
[----------] 1 test from AgentAndClientCount/Sorter_BENCHMARK_Test
[ RUN      ] AgentAndClientCount/Sorter_BENCHMARK_Test.FullSort/35
Using 50000 agents and 1000 clients
Added 1000 clients in 20019us
Added 50000 agents in 1.112101secs
Added allocations for 50000 agents in 4.532103secs
Full sort of 1000 clients took 41520us
No-op sort of 1000 clients took 395us
[       OK ] AgentAndClientCount/Sorter_BENCHMARK_Test.FullSort/35 (6886 ms)
[----------] 1 test from AgentAndClientCount/Sorter_BENCHMARK_Test (6886 ms total)

[----------] Global test environment tear-down
[==========] 1 test from 1 test case ran. (6907 ms total)
[  PASSED  ] 1 test.
```

With resource validation:
```
[==========] Running 1 test from 1 test case.
[----------] Global test environment set-up.
[----------] 1 test from AgentAndClientCount/Sorter_BENCHMARK_Test
[ RUN      ] AgentAndClientCount/Sorter_BENCHMARK_Test.FullSort/35
Using 50000 agents and 1000 clients
Added 1000 clients in 22417us
Added 50000 agents in 1.273512secs
Added allocations for 50000 agents in 43.449178secs
Full sort of 1000 clients took 31946us
No-op sort of 1000 clients took 307us
[       OK ] AgentAndClientCount/Sorter_BENCHMARK_Test.FullSort/35 (45996 ms)
[----------] 1 test from AgentAndClientCount/Sorter_BENCHMARK_Test (45996 ms total)

[----------] Global test environment tear-down
[==========] 1 test from 1 test case ran. (46019 ms total)
[  PASSED  ] 1 test.
```


Thanks,

Guangya Liu


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message