couchdb-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Damien Katz (JIRA)" <j...@apache.org>
Subject [jira] Commented: (COUCHDB-568) When delayed_commits = true, keep updated btree nodes in memory until the commit
Date Wed, 11 Nov 2009 20:44:39 GMT

    [ https://issues.apache.org/jira/browse/COUCHDB-568?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12776646#action_12776646
] 

Damien Katz commented on COUCHDB-568:
-------------------------------------

I know this isn't what you are trying to achieve with the batching, but I just remembered
something.

Any easy way to parallelized actions and use more CPUs is to spawn_link a process to query/modify
a sub node with a matching batch of key/actions, then move on to the next sub node and matching
key/actions and do the same, etc, etc. Then you wait for your sub-pids to return back their
result. The maximal # of recursive sub-processes active at any time would K*Log(N), where
K is the number of keys we are query/modifying and N is the total number of entries in the
index. In practice it wouldn't be that many processes, but it might want to limit the number
of processses spawned anyway, depending on tuning.

I think that actually would be a pretty good way to use more of the CPUs while updating an
index, assuming you are CPU bound on writes. For reads of multiple keys, it likely would also
be a performance win.

> When delayed_commits = true, keep updated btree nodes in memory until the commit
> --------------------------------------------------------------------------------
>
>                 Key: COUCHDB-568
>                 URL: https://issues.apache.org/jira/browse/COUCHDB-568
>             Project: CouchDB
>          Issue Type: Improvement
>    Affects Versions: 0.10
>            Reporter: Adam Kocoloski
>
> rnewson reported on IRC that the new batch=ok implementation results in significantly
larger overhead in the .couch files.  This makes sense; the old batch mode waited 1 second
before saving, but the new implementation just updates the doc asynchronously.  With fast
hardware and moderate write rates it's likely that each document is being written separately.
> The overhead presumably arises from frequently updated btree inner nodes being written
to disk many times over.  I'm interested in exploring a modification of the delayed_commits
mode whereby the updated btree nodes are not actually written to disk immediately, but are
instead held in memory until the commit.  I'd like to think that this will result in more
compact files without any decrease in durability.  New read requests would still be able to
access these in-memory nodes.
> I realize the notion that updates go directly to disk is baked pretty deeply into couch_btree,
but I still thought this was worth bringing up to a wider audience.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message