[
https://issues.apache.org/jira/browse/HBASE-16436?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15607391#comment-15607391
]
stack commented on HBASE-16436:
-------------------------------
Hard question by [~anoop.hbase] ported from HBASE-16421
bq. The CellChunkMap variant needs cell data to be available in MSLAB chunks.
So when MSLAB is OFF, we will not be using this. What abt the size restriction?
In MSLAB there is a max size for Cells and above which cells will not be
cloned. We might nor be able to handle such a case. Thoughts?
MSLAB is on always in master branch, right? If not, lets make it so by fiat in
2.0.
On size restrtiction, current default for mslab is 2MB. Seems small. For
smallest possible Cell (32bytes?), thats 1M Cells. Not that much, especially
when we are not looking them up via CSLS but via a nice flat index. Go to 8MB
slabs? 16M?
If someone wants to write a 64M Cell into HBase, should we even let them? It is
going to mess us (GC, etc.) up. We have a bound on max size currently.
Reexamine?
What is the lifecycle of a SLAB? Can we set aside negative chunkids to mean
onheap and these chunks would have one object in them only? (Might have to have
a globally accessible means of mapping negative chunkid to an object).
> Add the CellChunkMap variant
> ----------------------------
>
> Key: HBASE-16436
> URL: https://issues.apache.org/jira/browse/HBASE-16436
> Project: HBase
> Issue Type: Sub-task
> Reporter: ramkrishna.s.vasudevan
>
> This sub-task is specifically to add the CellChunkMap created by [~anastas]
> and [~eshcar] with specific tests and integrate it with the in memory
> flush/compaction flow.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
