mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Mesos Reviewbot Windows <revi...@mesos.apache.org>
Subject Re: Review Request 70409: Parameterized some allocator benchmarks for DRF and random sorters.
Date Tue, 09 Apr 2019 02:13:18 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/70409/#review214482
-----------------------------------------------------------



FAIL: Failed to apply the current review.

Failed command: `python.exe .\support\apply-reviews.py -n -r 70409`

All the build artifacts available at: http://dcos-win.westus2.cloudapp.azure.com/artifacts/mesos-reviewbot-testing/3120/mesos-review-70409

Relevant logs:

- [apply-review-70409.log](http://dcos-win.westus2.cloudapp.azure.com/artifacts/mesos-reviewbot-testing/3120/mesos-review-70409/logs/apply-review-70409.log):

```
´╗┐error: patch failed: src/tests/hierarchical_allocator_benchmarks.cpp:502
error: src/tests/hierarchical_allocator_benchmarks.cpp: patch does not apply
```

- Mesos Reviewbot Windows


On April 5, 2019, 3:02 p.m., Meng Zhu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/70409/
> -----------------------------------------------------------
> 
> (Updated April 5, 2019, 3:02 p.m.)
> 
> 
> Review request for mesos and Benjamin Mahler.
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Parameterized some allocator benchmarks for DRF and random sorters.
> 
> 
> Diffs
> -----
> 
>   src/tests/hierarchical_allocator_benchmarks.cpp a240a2bd51be5bda7f230c37c86431a0aff9173c

> 
> 
> Diff: https://reviews.apache.org/r/70409/diff/2/
> 
> 
> Testing
> -------
> 
> make check
> 
> Benchmark result with optimized build
> 
> TL;DR: Currently, the random sorter is much slower than DRF. In the steady state, where
no allocations are made (and thus on fair ground), random sorter can be twice as slower as
DRF. And the Geo-mean of all the results below (if it means anything) is: **Random is 56%
slower than DRF**.
> 
> ## BENCHMARK_HierarchicalAllocator_WithQuotaParam.LargeAndSmallQuota
> 
> ### DRF
> 
> Added 30 agents in 1.392422ms
> Added 30 frameworks in 7.724456ms
> Benchmark setup: 30 agents, 30 roles, 30 frameworks, with drf sorter
> Made 36 allocations in 7.851884ms
> Made 0 allocation in 3.102969ms
> 
> Added 300 agents in 10.252311ms
> Added 300 frameworks in 307.372697ms
> Benchmark setup: 300 agents, 300 roles, 300 frameworks, with drf sorter
> Made 350 allocations in 179.977552ms
> Made 0 allocation in 119.256302ms
> 
> Added 3000 agents in 84.523515ms
> Added 3000 frameworks in 20.026577907secs
> Benchmark setup: 3000 agents, 3000 roles, 3000 frameworks, with drf sorter
> Made 3500 allocations in 14.054427554secs
> Made 0 allocation in 9.833201145secs
> 
> ### Random
> 
> Added 30 agents in 44.185761ms
> Added 30 frameworks in 9.911124ms
> Benchmark setup: 30 agents, 30 roles, 30 frameworks, with random sorter
> Made 40 allocations in 6.776869ms
> Made 0 allocation in 3.893328ms
> 
> Added 300 agents in 8.416072ms
> Added 300 frameworks in 283.036985ms
> Benchmark setup: 300 agents, 300 roles, 300 frameworks, with random sorter
> Made 383 allocations in 230.376393ms
> Made 0 allocation in 178.068443ms
> 
> Added 3000 agents in 98.593695ms
> Added 3000 frameworks in 19.533776725secs
> Benchmark setup: 3000 agents, 3000 roles, 3000 frameworks, with random sorter
> Made 3865 allocations in 15.980519031secs
> Made 0 allocation in 16.991084682secs
> 
> ## BENCHMARK_HierarchicalAllocator_WithNonQuotaVsQuotaParam.NonQuotaVsQuota
> 
> ### DRF
> 
> Added 20 agents in 1.325004ms
> Added 20 frameworks in 3.299965ms
> Nonquota run setup: 20 agents, 10 roles, 20 frameworks, with drf sorter
> Made 20 allocations in 2.622706ms
> Made 0 allocation in 1.431925ms
> 
> Added 20 agents in 993319ns
> Added 20 frameworks in 2.705544ms
> Quota run setup: 20 agents, 10 roles, 20 frameworks, with drf sorter
> Made 20 allocations in 2.769981ms
> Made 0 allocation in 1.546198ms
> 
> Added 200 agents in 6.465386ms
> Added 200 frameworks in 92.599204ms
> Nonquota run setup: 200 agents, 100 roles, 200 frameworks, with drf sorter
> Made 200 allocations in 16.849375ms
> Made 0 allocation in 7.103731ms
> 
> Added 200 agents in 7.213524ms
> Added 200 frameworks in 97.207735ms
> Quota run setup: 200 agents, 100 roles, 200 frameworks, with drf sorter
> Made 200 allocations in 31.746068ms
> Made 0 allocation in 13.463503ms
> 
> Added 2000 agents in 63.492793ms
> Added 2000 frameworks in 5.497572945secs
> Nonquota run setup: 2000 agents, 1000 roles, 2000 frameworks, with drf sorter
> Made 2000 allocations in 724.310139ms
> Made 0 allocation in 579.387343ms
> 
> Added 2000 agents in 64.445124ms
> Added 2000 frameworks in 5.140273309secs
> Quota run setup: 2000 agents, 1000 roles, 2000 frameworks, with drf sorter
> Made 2000 allocations in 1.076597059secs
> Made 0 allocation in 886.381921ms
> 
> ### Random
> 
> Added 20 agents in 15.791371ms
> Added 20 frameworks in 3.631405ms
> Nonquota run setup: 20 agents, 10 roles, 20 frameworks, with random sorter
> Made 20 allocations in 2.615553ms
> Made 0 allocation in 1.575244ms
> 
> Added 20 agents in 990663ns
> Added 20 frameworks in 2.512379ms
> Quota run setup: 20 agents, 10 roles, 20 frameworks, with random sorter
> Made 20 allocations in 3.564547ms
> Made 0 allocation in 1.691739ms
> 
> Added 200 agents in 7.420338ms
> Added 200 frameworks in 77.987593ms
> Nonquota run setup: 200 agents, 100 roles, 200 frameworks, with random sorter
> Made 200 allocations in 25.26035ms
> Made 0 allocation in 15.715339ms
> 
> Added 200 agents in 8.107327ms
> Added 200 frameworks in 90.610819ms
> Quota run setup: 200 agents, 100 roles, 200 frameworks, with random sorter
> Made 200 allocations in 47.804156ms
> Made 0 allocation in 24.579104ms
> 
> Added 2000 agents in 60.38782ms
> Added 2000 frameworks in 5.169120074secs
> Nonquota run setup: 2000 agents, 1000 roles, 2000 frameworks, with random sorter
> Made 2000 allocations in 1.048805121secs
> Made 0 allocation in 959.389901ms
> 
> Added 2000 agents in 62.258101ms
> Added 2000 frameworks in 5.762369459secs
> Quota run setup: 2000 agents, 1000 roles, 2000 frameworks, with random sorter
> Made 2000 allocations in 1.825850875secs
> Made 0 allocation in 1.873957753secs
> 
> 
> Thanks,
> 
> Meng Zhu
> 
>


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