You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@geode.apache.org by "Ernest Burghardt (Jira)" <ji...@apache.org> on 2020/03/30 23:34:07 UTC
[jira] [Closed] (GEODE-7434) Unnecessary object allocation and
locking in GMSHealthMonitor.contactedBy
[ https://issues.apache.org/jira/browse/GEODE-7434?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Ernest Burghardt closed GEODE-7434.
-----------------------------------
Assignee: Ernest Burghardt (was: Jacob Barrett)
> Unnecessary object allocation and locking in GMSHealthMonitor.contactedBy
> -------------------------------------------------------------------------
>
> Key: GEODE-7434
> URL: https://issues.apache.org/jira/browse/GEODE-7434
> Project: Geode
> Issue Type: Bug
> Components: benchmarks, membership
> Reporter: Jacob Barrett
> Assignee: Ernest Burghardt
> Priority: Major
> Labels: performance
> Fix For: 1.12.0
>
> Time Spent: 20m
> Remaining Estimate: 0h
>
> GMSHealthMonitor.contactedBy allocates a {{TimeStamp}} for every call but throws it away almost all of the time in the call to {{putIfAbsent}}. Additionally, {{putIfAbsent}} always brags a monitor on the hash segment resulting in unnecessary synchronization.
> All with use of {{JavaWorkarounds.computeIfAbsent}} to avoid the locking and object allocation for the hot path where the entry already exists.
> Benchmark profiling shows 77% of P2P reader threads spent waiting for this lock.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)