You are viewing a plain text version of this content. The canonical link for it is here.
Posted to oak-issues@jackrabbit.apache.org by "Francesco Mari (JIRA)" <ji...@apache.org> on 2016/09/27 13:30:21 UTC
[jira] [Commented] (OAK-4659) Address records by logic identifiers
instead of offsets
[ https://issues.apache.org/jira/browse/OAK-4659?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15526124#comment-15526124 ]
Francesco Mari commented on OAK-4659:
-------------------------------------
I run the same benchmarks on the third version of the patch.
Results on trunk:
{noformat}
Apache Jackrabbit Oak 1.6-SNAPSHOT
# ConcurrentReadTest C min 10% 50% 90% max N
Oak-Segment-Tar 1 44 95 116 138 232 515
# ConcurrentReadWriteTest C min 10% 50% 90% max N
Oak-Segment-Tar 1 43 107 125 148 239 477
# ConcurrentWriteReadTest C min 10% 50% 90% max N
Oak-Segment-Tar 1 3 6 19 33 158 3293
# ConcurrentWriteTest C min 10% 50% 90% max N
Oak-Segment-Tar 1 75 83 94 177 306 523
{noformat}
Results of the patch:
{noformat}
Apache Jackrabbit Oak 1.6-SNAPSHOT
# ConcurrentReadTest C min 10% 50% 90% max N
Oak-Segment-Tar 1 55 119 140 161 276 429
# ConcurrentReadWriteTest C min 10% 50% 90% max N
Oak-Segment-Tar 1 75 133 155 179 282 386
# ConcurrentWriteReadTest C min 10% 50% 90% max N
Oak-Segment-Tar 1 4 8 23 35 164 2723
# ConcurrentWriteTest C min 10% 50% 90% max N
Oak-Segment-Tar 1 15 95 125 366 767 318
{noformat}
The results show a constant decrease in performance in every operation, with no particular penalty on some specific operation.
> Address records by logic identifiers instead of offsets
> -------------------------------------------------------
>
> Key: OAK-4659
> URL: https://issues.apache.org/jira/browse/OAK-4659
> Project: Jackrabbit Oak
> Issue Type: Improvement
> Components: segment-tar
> Reporter: Francesco Mari
> Assignee: Francesco Mari
> Fix For: Segment Tar 0.0.14
>
> Attachments: OAK-4659-01.patch, OAK-4659-02.patch, OAK-4659-03.patch
>
>
> Records are currently identified by their offsets inside the segments, This prevents records from being relocatable inside a segment and, as a consequence, this prevents segments to be shrunk down. This limitation forced us into a very limited solution space when working on garbage collection. Logic record IDs and their impact on the current implementation should be investigated.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)