[
https://issues.apache.org/jira/browse/HBASE-16438?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15942943#comment-15942943
]
Anastasia Braginsky commented on HBASE-16438:
---------------------------------------------
bq. I dont think Soft ref can be used. Once the CSLM is converted to
CellChunkMap, we dont have any hard refs to Chunks from any where (Cells are
gone). Now the ChunkMap contain chunkId and offset , length info. Now assume a
GC picked and removed these chunks (Ya there are no hard refs to it), and after
a read comes and we want to read back that Cell, what to do? !!!!
We can "harden" the references in the map in the process of transferring to
CellChunkMap (flattening). The chunks that can not be reached will not be a
part of this process and their IDs won't be found so their references are not
going to be "harden", which is OK.
bq. So the Chunk that is put in the map will have a soft reference?
Yes. The chunk that is used for CSLM or CellArrayMap is going to have soft
reference the chunk that is used for CellChunkMap is going to have hard
reference.
bq. Can we say CellChunkMap to be used along with off heap MSLAB pool usage
only?
This is the first question to answer. I think we should not limit the
CellChunkMap to off-heap right now. As we have seen with CompactingMemStore
only when you implement it all and make a good performance testing you can see
what is benefit. As now we see the merge is a good thing to do, although in the
past we already thought to remove this code. So if we implement CellChunkMap
now without any possibility to use it on-heap, we are limiting ourselves
without any experimental evidence. I believe we should write the code generally
enough.
> Create a cell type so that chunk id is embedded in it
> -----------------------------------------------------
>
> Key: HBASE-16438
> URL: https://issues.apache.org/jira/browse/HBASE-16438
> Project: HBase
> Issue Type: Sub-task
> Affects Versions: 2.0.0
> Reporter: ramkrishna.s.vasudevan
> Assignee: ramkrishna.s.vasudevan
> Attachments: HBASE-16438_1.patch,
> HBASE-16438_3_ChunkCreatorwrappingChunkPool.patch,
> HBASE-16438_4_ChunkCreatorwrappingChunkPool.patch, HBASE-16438.patch,
> MemstoreChunkCell_memstoreChunkCreator_oldversion.patch,
> MemstoreChunkCell_trunk.patch
>
>
> For CellChunkMap we may need a cell such that the chunk out of which it was
> created, the id of the chunk be embedded in it so that when doing flattening
> we can use the chunk id as a meta data. More details will follow once the
> initial tasks are completed.
> Why we need to embed the chunkid in the Cell is described by [~anastas] in
> this remark over in parent issue
> https://issues.apache.org/jira/browse/HBASE-14921?focusedCommentId=15244119&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-15244119
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
