[
https://issues.apache.org/jira/browse/HADOOP-10172?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Daryn Sharp updated HADOOP-10172:
---------------------------------
Attachment: HADOOP-10172.patch
Add a factory which caches the mechanism to standard sasl factory lookup.
> Cache SASL server factories
> ---------------------------
>
> Key: HADOOP-10172
> URL: https://issues.apache.org/jira/browse/HADOOP-10172
> Project: Hadoop Common
> Issue Type: Improvement
> Components: ipc
> Affects Versions: 0.23.0, 2.0.0-alpha, 3.0.0
> Reporter: Daryn Sharp
> Assignee: Daryn Sharp
> Priority: Critical
> Attachments: HADOOP-10172.patch
>
>
> Performance for SASL server creation is _atrocious_.
> {{Sasl.createSaslServer}} does not cache the provider resolution for the
> factories. Factory resolution and server instantiation has 3 major
> contention points. During bursts of connections, one reader accepting a
> connection stalls other readers accepting connections, in turn stalling all
> existing connections handled by those readers.
> I benched 5 threads at 187 instances/s - total, not per thread. With this
> and another change, I've boosted it to 33K instances/s.
--
This message was sent by Atlassian JIRA
(v6.1.4#6159)
