[
https://issues.apache.org/jira/browse/HBASE-20259?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16421812#comment-16421812
]
Anastasia Braginsky commented on HBASE-20259:
---------------------------------------------
Thanks for publishing the patch on RB!
I completely understand your need to cut the release. It is really taking us
too much time :(. However, the performance evaluation that was done in
HBASE-20188 so far, is not comprehensive yet, and for now only scenarios that
are disadvantageous for CompactingMemStore were run. This is why:
# Too much heap space was given (and just about 10 regions), which makes it
easier for GC, thus we don't see the advantage for "GC-friendly"
data-structures used by CompactingMemStore. In our tests we used 50-100 regions
and wrote much more data, so we can see how the different MemStores behave when
RAM is fully utilized, which should usually be the case.
# The tests were short (20 min), so we do not see the performance effect of
having less on-disk compaction provided by CompactingMemStore.
CompactingMemStore saves heap space and requires less flushes-to-disk in the
long term.
We really appreciate your effort in running the tests and we fully understand
that the above configuration was chosen in order to save time. It is clear we
are already overdue. On the other hand, the above configuration had shown "that
the benefit if any was too small", which is not quite objective.
We just have Passover holiday now, but still try to make progress with the
alternative evaluation. We are now running the tests which are close to your
configuration, but with little changes to add more stress and to show the
advantage of CompactingMemStore. We were not aware that CMS GC is going to be
default for HBase2.0, so we are considering this now.
Please reconsider delaying the decision about CompactingMemStore's default.
Thank you! :)
> hbase.systemtables.compacting.memstore.type=NONE does not do as expected
> ------------------------------------------------------------------------
>
> Key: HBASE-20259
> URL: https://issues.apache.org/jira/browse/HBASE-20259
> Project: HBase
> Issue Type: Bug
> Reporter: stack
> Assignee: stack
> Priority: Critical
> Fix For: 2.0.0
>
> Attachments: HBASE-20259.master.001.patch,
> HBASE-20259.master.002.patch
>
>
> I set {{hbase.systemtables.compacting.memstore.type}} to NONE but it seems
> like in-memory is still on. My table looks like this:
> {code}
> Table ycsb is ENABLED
> ycsb
> COLUMN FAMILIES DESCRIPTION
> {NAME => 'family', VERSIONS => '1', EVICT_BLOCKS_ON_CLOSE => 'false',
> NEW_VERSION_BEHAVIOR => 'false', KEEP_DELETED_CELLS => 'FALSE',
> CACHE_DATA_ON_WRITE => 'false', DATA_BLOCK_ENCODING => 'NONE', TTL =>
> 'FOREVER', MIN_VERSIONS => '0', REPLICATION_SCOPE => '0', BLOOMFILTER =
> > 'ROW', CACHE_INDEX_ON_WRITE => 'false', IN_MEMORY => 'false',
> > CACHE_BLOOMS_ON_WRITE => 'false', PREFETCH_BLOCKS_ON_OPEN => 'false',
> > COMPRESSION => 'NONE', BLOCKCACHE => 'true', BLOCKSIZE => '65536'}
> {code}
> Looks like table doesn't have it on either (IN_MEMORY_COMPACTION doesn't show
> in the above).
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
