[
https://issues.apache.org/jira/browse/CASSANDRA-6689?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13942576#comment-13942576
]
Pavel Yaskevich commented on CASSANDRA-6689:
--------------------------------------------
Here is the quote from my previous comment:
_As we don't need Impl for this ticket before #3 (if we need it at all in that
way is the topic of the separate discussion, did we at least test how much do
we actually safe by doing it that way? I think Jonathan Ellis with me on this
one after reading CASSANDRA-6694) let's remove it, the less code we have to
read the faster review process would go. *So if the NativeCell is going to be
introduced in #3 that means all of that Impl stuff should go there too*._
So first of all it's not clear if we even want to move in Impl direction as
mentioned in CASSANDRA-6694, secondly we all agree that we are going to keep
this to the point with roadmap we have so each of the points is individually
deliverable/commitable, which essentially means less work for everybody and the
ticket moves on faster, thirdly it looks like 4 people are reaching to one
target reflected at least in Jonathan's roadmap where we want to get off-heap
memtables with on-heap read side copy and so on _incrementally_, and you are
just trying to do as such refactoring/changes as possible in all three of the
bullet points, it looks that way because iss-6689-{1,2,3} all have at least
*3K* additions and *2K* deletions, concerns about the size of those changes
have already been raised by multiple people multiple times and yet we are still
at the same spot with this.
> Partially Off Heap Memtables
> ----------------------------
>
> Key: CASSANDRA-6689
> URL: https://issues.apache.org/jira/browse/CASSANDRA-6689
> Project: Cassandra
> Issue Type: New Feature
> Components: Core
> Reporter: Benedict
> Assignee: Benedict
> Labels: performance
> Fix For: 2.1 beta2
>
> Attachments: CASSANDRA-6689-small-changes.patch
>
>
> Move the contents of ByteBuffers off-heap for records written to a memtable.
> (See comments for details)
--
This message was sent by Atlassian JIRA
(v6.2#6252)
