[
https://issues.apache.org/jira/browse/HBASE-13082?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14993245#comment-14993245
]
ramkrishna.s.vasudevan commented on HBASE-13082:
------------------------------------------------
bq.The locks are not the problem, it's the memory barriers on every call to
next and pretty that cause the performance issue. The locks are (almost) never
contended
Yes Lars. I too agree with this. I am trying to profile before I could come to
any conclusion on the memory barrier that this volatile introduces.
Ideally for the compaction cases we don't even need the notifying mechansim
and can always be going on without any locks or resets of the heap but for
flushes we may need it because if we don't reset then the flushed snapshot
cannot be GCed and if this is going to be bigger scans then may lead to GC
issues. Other than that as you always say Scanner is >99% single threaded.
> Coarsen StoreScanner locks to RegionScanner
> -------------------------------------------
>
> Key: HBASE-13082
> URL: https://issues.apache.org/jira/browse/HBASE-13082
> Project: HBase
> Issue Type: Bug
> Reporter: Lars Hofhansl
> Assignee: ramkrishna.s.vasudevan
> Attachments: 13082-test.txt, 13082-v2.txt, 13082-v3.txt,
> 13082-v4.txt, 13082.txt, 13082.txt, HBASE-13082.pdf, HBASE-13082_1.pdf,
> HBASE-13082_1_WIP.patch, HBASE-13082_2_WIP.patch, HBASE-13082_3.patch,
> HBASE-13082_4.patch, HBASE-13082_9.patch, HBASE-13082_9.patch, gc.png,
> gc.png, gc.png, hits.png, next.png, next.png
>
>
> Continuing where HBASE-10015 left of.
> We can avoid locking (and memory fencing) inside StoreScanner by deferring to
> the lock already held by the RegionScanner.
> In tests this shows quite a scan improvement and reduced CPU (the fences make
> the cores wait for memory fetches).
> There are some drawbacks too:
> * All calls to RegionScanner need to be remain synchronized
> * Implementors of coprocessors need to be diligent in following the locking
> contract. For example Phoenix does not lock RegionScanner.nextRaw() and
> required in the documentation (not picking on Phoenix, this one is my fault
> as I told them it's OK)
> * possible starving of flushes and compaction with heavy read load.
> RegionScanner operations would keep getting the locks and the
> flushes/compactions would not be able finalize the set of files.
> I'll have a patch soon.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
