[
https://issues.apache.org/jira/browse/HBASE-16421?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15780313#comment-15780313
]
Anastasia Braginsky commented on HBASE-16421:
---------------------------------------------
Hi [~ram_krish],
First of all, thank you very much for running the benchmark.
So you had two YCSB clients running for about 16 minutes, one doing sequential
writes and another one running the scans.
In addition, you used the CellChunkMaps which were pointing to the *on-heap*
MSLAB chunks (as you measure the off-heap separately).
Were you using the chunk pool with CellChunkMaps? Am I correct in my
understanding of your benchmark?
We assumed we will have more garbage with CellChunkMaps upon the scans (whether
the scans are coming from the flush or not). This can be seen from the results.
Although, the number of the GC pauses are also related to whether you are
running with the MSLAB/Chunk Pool. I assume that in all experiments (except
DefaultMemStore) you had MSLAB and Chunk Pool on, am I right? What is the
"total time" that you are presenting? Total time of the experiment? Isn’t it 16
minutes? Is it any accumulative latency? You can get specifically the scan
latency from YCSB.
Anyway, we can say that CellChunkMap is not decreasing the performance too
much, and it worth it at least as off-heaping worth it. Do you agree?
> Introducing the CellChunkMap as a new additional index variant in the MemStore
> ------------------------------------------------------------------------------
>
> Key: HBASE-16421
> URL: https://issues.apache.org/jira/browse/HBASE-16421
> Project: HBase
> Issue Type: Umbrella
> Reporter: Anastasia Braginsky
> Attachments: CellChunkMapRevived.pdf, ChunkCell_creation.png,
> IntroductiontoNewFlatandCompactMemStore.pdf
>
>
> Follow up for HBASE-14921. This is going to be the umbrella JIRA to include
> all the parts of integration of the CellChunkMap to the MemStore.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
