[
https://issues.apache.org/jira/browse/MESOS-7876?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16127588#comment-16127588
]
Benno Evers commented on MESOS-7876:
------------------------------------
Spot-checking some of the mesos benchmarks using jemalloc vs. system malloc, I
can observe a small but consistent speedup from 1% to 6% using jemalloc over
glibc. There certainly is no indication that switching to jemalloc would lead
to performance regressions.
With jemalloc:
{code}
[ OK ]
SlaveAndFrameworkCount/HierarchicalAllocator_BENCHMARK_Test.DeclineOffers/1
(575213 ms)
[ OK ]
SlaveAndFrameworkCount/HierarchicalAllocator_BENCHMARK_Test.Metrics/1 (1963 ms)
[ OK ]
SlaveAndFrameworkCount/HierarchicalAllocator_BENCHMARK_Test.Metrics/10 (18756
ms)
[ OK ]
SlaveAndFrameworkCount/HierarchicalAllocator_BENCHMARK_Test.Metrics/11 (37044
ms)
[ OK ]
SlaveAndFrameworkCount/HierarchicalAllocator_BENCHMARK_Test.Metrics/12 (97298
ms)
[ OK ] Tasks/SchedulerReconcileTasks_BENCHMARK_Test.SchedulerLibrary/0
(302 ms)
[ OK ] Tasks/SchedulerReconcileTasks_BENCHMARK_Test.SchedulerLibrary/1
(2311 ms)
[ OK ] Tasks/SchedulerReconcileTasks_BENCHMARK_Test.SchedulerLibrary/2
(12104 ms)
{code}
With default malloc:
{code}
[ OK ]
SlaveAndFrameworkCount/HierarchicalAllocator_BENCHMARK_Test.DeclineOffers/1
(610002 ms)
[ OK ]
SlaveAndFrameworkCount/HierarchicalAllocator_BENCHMARK_Test.Metrics/1 (2065 ms)
[ OK ]
SlaveAndFrameworkCount/HierarchicalAllocator_BENCHMARK_Test.Metrics/10 (20207
ms)
[ OK ]
SlaveAndFrameworkCount/HierarchicalAllocator_BENCHMARK_Test.Metrics/11 (38086
ms)
[ OK ]
SlaveAndFrameworkCount/HierarchicalAllocator_BENCHMARK_Test.Metrics/12 (98475
ms)
[ OK ] Tasks/SchedulerReconcileTasks_BENCHMARK_Test.SchedulerLibrary/0
(281 ms)
[ OK ] Tasks/SchedulerReconcileTasks_BENCHMARK_Test.SchedulerLibrary/1
(2448 ms)
[ OK ] Tasks/SchedulerReconcileTasks_BENCHMARK_Test.SchedulerLibrary/2
(12673 ms)
{code}
> Investigate alternative malloc implementations for mesos
> --------------------------------------------------------
>
> Key: MESOS-7876
> URL: https://issues.apache.org/jira/browse/MESOS-7876
> Project: Mesos
> Issue Type: Improvement
> Reporter: Benno Evers
> Assignee: Benno Evers
> Attachments: jemalloc_benchmark_raw.txt, malloc.cpp, noprof.png,
> prof.png
>
>
> It is currently very hard to debug memory issues, in particular memory leaks,
> in mesos.
> An alluring way to improve the situation would be to change the default
> malloc to jemalloc, which has built-in heap-tracking capabilities.
> However, some care needs to be taken when considering to change such a
> fundamental part of mesos:
> * Would such a switch have any adverse impact on performance?
> * Is it available and will it compile on all our target platforms?
> * Is the jemalloc-licensing compatible with bundling as third-party library?
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
