[ https://issues.apache.org/jira/browse/OAK-9116?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17138217#comment-17138217 ]
Thomas Mueller commented on OAK-9116: ------------------------------------- TODO: * a script (maybe groovy script, using oak-run) to inline very small binaries, or add them to this cache. * investigate if we could use the pre-extraction cache (the same key-value store and code might be used there). * allow deleting binaries from the cache, otherwise we hide DSGC misconfiguration problems. > Datastore: read-only cache that uses a key-value store > ------------------------------------------------------ > > Key: OAK-9116 > URL: https://issues.apache.org/jira/browse/OAK-9116 > Project: Jackrabbit Oak > Issue Type: Improvement > Components: blob > Reporter: Thomas Mueller > Assignee: Thomas Mueller > Priority: Major > > For data stores that are large and contain a lot of small entries, the > overhead of the current file cache is quite large: at least one file per > entry is needed. For most file systems, that means 4 KB per entry. If entries > are much smaller (e.g. 30 bytes), the overhead is large. > We could instead use a read-only key-value store for such entries. -- This message was sent by Atlassian Jira (v8.3.4#803005)