[ 
https://issues.apache.org/jira/browse/CASSANDRA-7443?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14169265#comment-14169265
 ] 

Sylvain Lebresne commented on CASSANDRA-7443:
---------------------------------------------

bq.  can "hide" those settings but internally we still need access to them as 
developers

Why? We've always transitioned sstables format by continuing to read old 
sstables and only writing new ones, and I'm fan of sticking to that. If we do 
so, the actual format used is hardcoded for write (for a given C* version) and 
based on the sstable for read. No reason to have a Config setting, even hidden. 
Note that if we ever change our mind on how we want to proceed, adding a hidden 
setting is pretty easy, but I'd rather not anticipate on stuffs we haven't 
agreed on yet.

bq. I'm imagining (perhaps prematurely) that some sstable formats could be 
compacted in a much more efficient way if we have a way

I do think that's premature. If we come up with a format that support that, we 
can always find the proper abstraction at that time.

> SSTable Pluggability v2
> -----------------------
>
>                 Key: CASSANDRA-7443
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-7443
>             Project: Cassandra
>          Issue Type: Improvement
>          Components: Core
>            Reporter: T Jake Luciani
>            Assignee: T Jake Luciani
>             Fix For: 3.0
>
>         Attachments: 7443-refactor-v1.txt, 7443-testformat-v1.txt
>
>
> As part of a wider effort to improve the performance of our storage engine we 
> will need to support basic pluggability of the SSTable reader/writer. We 
> primarily need this to support the current SSTable format and new SSTable 
> format in the same version.  This will also let us encapsulate the changes in 
> a single layer vs forcing the whole engine to change at once.
> We previously discussed how to accomplish this in CASSANDRA-3067
>   



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to