cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ariel Weisberg (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-9738) Migrate key-cache to be fully off-heap
Date Mon, 17 Aug 2015 22:55:47 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-9738?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14700387#comment-14700387
] 

Ariel Weisberg commented on CASSANDRA-9738:
-------------------------------------------

Good stuff. Coverage of the OHC key cache provider looks good.

* How are the 2i paths tested?
* The null case in makeVal isn't tested, maybe not that interesting
* SerializationHeader forKeyCache is racy and can result in an undersize array clobbering
a properly sized one. But... it doesn't retrieve the value it sets so odds are eventually
it will work out to be the longer one. It works it's just intentionally racy.
* In CacheService does that comment about singleton weigher even make sense anymore?
* NIODataInputStream has a derived class DataInputBuffer that exposes the constructor you
made public.
* The string encoding and decoding helpers you wrote seem like they should be factored out
somewhere else, maybe ByteBufferUtil? Also you don't specify a string encoding and there may
be some issues with serialized size of non-latin characters lurking as well.
* An enhancement we can file for later is to replace those strings with vints that reference
a map of possible table names. For persistence definitely fully qualify, but in memory we
can store more entries that way.

> Migrate key-cache to be fully off-heap
> --------------------------------------
>
>                 Key: CASSANDRA-9738
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-9738
>             Project: Cassandra
>          Issue Type: Sub-task
>            Reporter: Robert Stupp
>            Assignee: Robert Stupp
>             Fix For: 3.0 beta 2
>
>
> Key cache still uses a concurrent map on-heap. This could go to off-heap and feels doable
now after CASSANDRA-8099.
> Evaluation should be done in advance based on a POC to prove that pure off-heap counter
cache buys a performance and/or gc-pressure improvement.
> In theory, elimination of on-heap management of the map should buy us some benefit.



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

Mime
View raw message