[
https://issues.apache.org/jira/browse/HBASE-68?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12728878#action_12728878
]
Jonathan Gray commented on HBASE-68:
------------------------------------
Locality groups are something we should do. But agree we can treat it separate
from this issue as we are not doing it now.
KVs can not be reconstituted without making every read copy all data twice. We
have to read blocks in from hdfs. At that point, we can just pass the entire
buffer along and make KV "views" against the big block. Or, we could rewrite
the block again reconstituting larger KVs. That could be done when reading in,
or when building results. In either case, we are no longer zero-copy read.
I don't see any way to do this without going backwards towards how things used
to work... all the massive improvements we see are because of this consistent,
explicit, and immutable KV.
> [hbase] HStoreFiles needlessly store the column family name in every entry
> --------------------------------------------------------------------------
>
> Key: HBASE-68
> URL: https://issues.apache.org/jira/browse/HBASE-68
> Project: Hadoop HBase
> Issue Type: Improvement
> Components: regionserver
> Reporter: Bryan Duxbury
> Priority: Minor
> Fix For: 0.20.0
>
>
> Today, HStoreFiles keep the entire serialized HStoreKey objects around for
> every cell in the HStore. Since HStores are 1-1 with column families, this is
> really unnecessary - you can always surmise the column family by looking at
> the HStore it belongs to. (This information would ostensibly come from the
> file name or a header section.) This means that we could remove the column
> family part of the HStoreKeys we put into the HStoreFile, reducing the size
> of data stored. This would be a space-saving benefit, removing redundant
> data, and could be a speed benefit, as you have to scan over less data in
> memory and transfer less data over the network.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
