[
https://issues.apache.org/jira/browse/OAK-12069?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Julian Sedding resolved OAK-12069.
----------------------------------
Fix Version/s: 1.92.0
Resolution: Fixed
> reduce SegmentPreloader memory usage
> ------------------------------------
>
> Key: OAK-12069
> URL: https://issues.apache.org/jira/browse/OAK-12069
> Project: Jackrabbit Oak
> Issue Type: Improvement
> Components: segment-tar
> Reporter: Julian Sedding
> Assignee: Julian Sedding
> Priority: Major
> Fix For: 1.92.0
>
>
> The {{SegmentPreloader}} builds a cache containing the segment graph of each
> archive. For large segmentstores, this can consume a lot of memory.
> An attempt to solve this by deduplicating in-memory UUIDs (OAK-12005),
> leveraging the fact that all UUIDs are already in-memory due to the archive's
> indexes being cached, proved too slow on large segmentstores to be
> practically useful. Therefore, this change is reverted with OAK-12068.
> Instead, the usage of the segment graphs in the {{SegmentPreloader}} will be
> replaced by reading segment references directly from the segment-data. This
> will cause more I/O in the form of additional reads from the
> {{{}PersistentCache{}}}, however, these reads are not in the critical path
> and should be fast (e.g. local disk). In comparison to remote segment loads,
> these should have no significant impact.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
