[
https://issues.apache.org/jira/browse/AURORA-1854?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15806175#comment-15806175
]
Reza Motamedi commented on AURORA-1854:
---------------------------------------
reviewable at: https://reviews.apache.org/r/55284/
sample benchmark for SlaCalculation
{code}
Benchmark (numTasks) Mode Cnt
Score Error Units
SlaAlgorithmBenchmark.MTTXBenchmark.measureMTTA 100 thrpt 5
156249.415 ± 52660.004 ops/s
SlaAlgorithmBenchmark.MTTXBenchmark.measureMTTA 1000 thrpt 5
3539.248 ± 1757.523 ops/s
SlaAlgorithmBenchmark.MTTXBenchmark.measureMTTA 10000 thrpt 5
54.701 ± 19.743 ops/s
SlaAlgorithmBenchmark.MTTXBenchmark.measureMTTR 100 thrpt 5
124326.484 ± 45198.021 ops/s
SlaAlgorithmBenchmark.MTTXBenchmark.measureMTTR 1000 thrpt 5
3973.739 ± 1216.296 ops/s
SlaAlgorithmBenchmark.MTTXBenchmark.measureMTTR 10000 thrpt 5
57.102 ± 5.567 ops/s
SlaAlgorithmBenchmark.MTTXBenchmark.measureMTTS 100 thrpt 5
129655.298 ± 38446.901 ops/s
SlaAlgorithmBenchmark.MTTXBenchmark.measureMTTS 1000 thrpt 5
3475.401 ± 2283.158 ops/s
SlaAlgorithmBenchmark.MTTXBenchmark.measureMTTS 10000 thrpt 5
56.215 ± 11.524 ops/s
{/code}
> Create JMH benchmarks for SLA metrics calculation
> -------------------------------------------------
>
> Key: AURORA-1854
> URL: https://issues.apache.org/jira/browse/AURORA-1854
> Project: Aurora
> Issue Type: Story
> Components: Efficiency, Scheduler
> Reporter: Mehrdad Nurolahzade
> Assignee: Reza Motamedi
> Priority: Minor
> Labels: newbie
>
> SLA metrics calculation logic in {{MetricCalculator}} and {{SlaAlgorithm}}
> causes great churn in scheduler. Profiling shows the burden of calculation
> can grow as much as 50% of scheduler CPU.
> Create JMH benchmarks for this logic to enable analysis and improvement of
> these calculators.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
