aurora-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Aurora ReviewBot <wfar...@apache.org>
Subject Re: Review Request 63454: Use a pair of fields for caching offer resources rather than a Cache
Date Wed, 01 Nov 2017 15:47:04 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/63454/#review189804
-----------------------------------------------------------


Ship it!




Master (d106b4e) is green with this patch.
  ./build-support/jenkins/build.sh

I will refresh this build result if you post a review containing "@ReviewBot retry"

- Aurora ReviewBot


On Oct. 31, 2017, 10:49 p.m., Bill Farner wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/63454/
> -----------------------------------------------------------
> 
> (Updated Oct. 31, 2017, 10:49 p.m.)
> 
> 
> Review request for Aurora, David McLaughlin and Jordan Ly.
> 
> 
> Repository: aurora
> 
> 
> Description
> -------
> 
> Was surprised to see `TierInfo#hashCode()` show up in a profiler, and ended up in this
rabbit hole.  A `Cache` is overkill for caching revocable and non-revocable resources in an
offer.  I then pulled a thread on `TierInfo` being passed around when we only need the `revocable`
flag.
> 
> I find the resulting code easier to comprehend as well.
> 
> 
> Diffs
> -----
> 
>   src/jmh/java/org/apache/aurora/benchmark/SchedulingBenchmarks.java e0ec793cad05674fb4b65246a6d153521b28b914

>   src/main/java/org/apache/aurora/scheduler/HostOffer.java acc637a5f695741e8ee527385bad9dec17ea4f15

>   src/main/java/org/apache/aurora/scheduler/mesos/MesosTaskFactory.java 34ed8207c6bc09cde2a82b5867f4fd47a047a54e

>   src/main/java/org/apache/aurora/scheduler/resources/AcceptedOffer.java 291d5c95916915afc48a7143759e523fccd52feb

>   src/main/java/org/apache/aurora/scheduler/resources/ResourceManager.java 9aa263a9cfae03a9a0c5bc7fe3a1405397d3009c

>   src/main/java/org/apache/aurora/scheduler/state/TaskAssigner.java 549fd2988569d37cfff199d058010259750938fc

>   src/test/java/org/apache/aurora/scheduler/mesos/MesosTaskFactoryImplTest.java d91b54047a611bce5c89a0b77b07fa88ff40eae9

>   src/test/java/org/apache/aurora/scheduler/resources/AcceptedOfferTest.java c47f7c9fec61399f09a2b2a7eccfd1615dde8230

>   src/test/java/org/apache/aurora/scheduler/resources/ResourceManagerTest.java 37dfcbc073e3c0b6c3126dba31fdc4422629b335

>   src/test/java/org/apache/aurora/scheduler/state/FirstFitTaskAssignerTest.java 5837250407cace3fa614f83a4992d430b70cdee3

> 
> 
> Diff: https://reviews.apache.org/r/63454/diff/2/
> 
> 
> Testing
> -------
> 
> No discernable performance changes in benchmarks (within error margins).
> 
> Master:
> ```
> Benchmark                                                                     (numPendingTasks)
  Mode  Cnt       Score      Error  Units
> SchedulingBenchmarks.ClusterFullUtilizationBenchmark.runBenchmark                   
       N/A  thrpt   10  178037.291 ± 6295.285  ops/s
> SchedulingBenchmarks.FillClusterBenchmark.runBenchmark                              
       N/A  thrpt   10    1465.314 ± 1538.509  ops/s
> SchedulingBenchmarks.InsufficientResourcesSchedulingBenchmark.runBenchmark          
       N/A  thrpt   10   13993.963 ±  747.815  ops/s
> SchedulingBenchmarks.LimitConstraintMismatchSchedulingBenchmark.runBenchmark        
       N/A  thrpt   10    1296.855 ±   61.355  ops/s
> SchedulingBenchmarks.PreemptorSlotSearchBenchmark.runBenchmark                      
         1  thrpt   10      87.170 ±    1.377  ops/s
> SchedulingBenchmarks.PreemptorSlotSearchBenchmark.runBenchmark                      
        10  thrpt   10      85.035 ±    7.125  ops/s
> SchedulingBenchmarks.PreemptorSlotSearchBenchmark.runBenchmark                      
       100  thrpt   10      86.725 ±    4.409  ops/s
> SchedulingBenchmarks.PreemptorSlotSearchBenchmark.runBenchmark                      
      1000  thrpt   10      75.962 ±    0.975  ops/s
> SchedulingBenchmarks.ValueConstraintMismatchSchedulingBenchmark.runBenchmark        
       N/A  thrpt   10   16358.086 ±  139.123  ops/s
> ```
> 
> With this patch:
> ```
> Benchmark                                                                     (numPendingTasks)
  Mode  Cnt       Score      Error  Units
> SchedulingBenchmarks.ClusterFullUtilizationBenchmark.runBenchmark                   
       N/A  thrpt   10  174874.109 ± 8002.217  ops/s
> SchedulingBenchmarks.FillClusterBenchmark.runBenchmark                              
       N/A  thrpt   10    1194.696 ± 1400.142  ops/s
> SchedulingBenchmarks.InsufficientResourcesSchedulingBenchmark.runBenchmark          
       N/A  thrpt   10   15787.903 ±  837.119  ops/s
> SchedulingBenchmarks.LimitConstraintMismatchSchedulingBenchmark.runBenchmark        
       N/A  thrpt   10    1341.742 ±   49.151  ops/s
> SchedulingBenchmarks.PreemptorSlotSearchBenchmark.runBenchmark                      
         1  thrpt   10      86.147 ±    2.213  ops/s
> SchedulingBenchmarks.PreemptorSlotSearchBenchmark.runBenchmark                      
        10  thrpt   10      86.729 ±    1.586  ops/s
> SchedulingBenchmarks.PreemptorSlotSearchBenchmark.runBenchmark                      
       100  thrpt   10      84.574 ±    2.526  ops/s
> SchedulingBenchmarks.PreemptorSlotSearchBenchmark.runBenchmark                      
      1000  thrpt   10      71.776 ±    3.214  ops/s
> SchedulingBenchmarks.ValueConstraintMismatchSchedulingBenchmark.runBenchmark        
       N/A  thrpt   10   14757.428 ±  402.668  ops/s
> ```
> 
> 
> Thanks,
> 
> Bill Farner
> 
>


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