[
https://issues.apache.org/jira/browse/IGNITE-28493?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Konstantin Orlov updated IGNITE-28493:
--------------------------------------
Ignite Flags: (was: Docs Required,Release Notes Required)
> Reduce contention in RemotelyTriggeredResourceRegistry
> ------------------------------------------------------
>
> Key: IGNITE-28493
> URL: https://issues.apache.org/jira/browse/IGNITE-28493
> Project: Ignite
> Issue Type: Improvement
> Components: transactions ai3
> Reporter: Konstantin Orlov
> Assignee: Konstantin Orlov
> Priority: Major
> Labels: ignite-3
>
> As for now, all resource tracking on per-host basis is done inside `compute`
> methods of hash map. For example:
> {code}
> private void addRemoteHostResource(UUID remoteHostId,
> FullyQualifiedResourceId resourceId) {
> remoteHostsToResources.compute(remoteHostId, (k, v) -> {
> if (v == null) {
> v = ConcurrentHashMap.newKeySet();
> }
> v.add(resourceId);
> return v;
> });
> }
> {code}
> Given that number of nodes is relatively small, and we use concurrent
> collections as values anyway, such an approach creates unnecessary contention
> on hot path.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
