cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Benedict (JIRA)" <>
Subject [jira] [Commented] (CASSANDRA-8897) Remove FileCacheService, instead pooling the buffers
Date Mon, 06 Apr 2015 13:39:12 GMT


Benedict commented on CASSANDRA-8897:

bq. Does the buffer pool need to support metrics just like the FileCacheService did?
Perhaps. At most, though, we need to indicate how often we fail a cache lookup, since it shouldn't
really ever happen in the new scheme on a healthy system.
bq. What about the maximum memory size beyond which it cannot grow?
The file cache size in specifies how big we should keep any cached portion to,
but we will have to still produce a buffer if this cache is exhausted.
bq. Should we create an initial number of buffers or should we just grow the pool dynamically?
I don't think it matters terribly. I would personally opt for a small core pool of a few megabytes
allocated upfront, with the remainder grown dynamically. 
bq. Should the pool be only a replacement of FileCacheService, therefore used only by PoolingSegmentedFile
and derived classes when calling createPooledReader(), or should the scope be extended to
all RARs created by SegmentedFile instances?
I think PoolingSegmentedFile and PoolingBufferedFile can simply be removed entirely, and these
pooled buffers can be shared between all SegmentedFile instances

> Remove FileCacheService, instead pooling the buffers
> ----------------------------------------------------
>                 Key: CASSANDRA-8897
>                 URL:
>             Project: Cassandra
>          Issue Type: Improvement
>          Components: Core
>            Reporter: Benedict
>            Assignee: Stefania
>             Fix For: 3.0
> After CASSANDRA-8893, a RAR will be a very lightweight object and will not need caching,
so we can eliminate this cache entirely. Instead we should have a pool of buffers that are

This message was sent by Atlassian JIRA

View raw message