[
https://issues.apache.org/jira/browse/CASSANDRA-801?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Stu Hood updated CASSANDRA-801:
-------------------------------
Attachment: trunk-0006-Remove-JMXAggregatingCache.patch
trunk-0005-SSTableTracker-sets-the-cache-for-tracked-Readers.patch
trunk-0004-Migrate-rowCache-into-SSTableTracker.patch
Patches 3 through 6 for trunk and 0.6 remove JMXAggregatingCache, and instead
use a merged cache for all SSTables. Using individual caches per table would
have lead to fragmentation, no matter how we sized them.
Both the row and key caches are moved into SSTableTracker, and they are resized
as the estimated number of keys for the CFS changes.
Patches were stress tested with trunk and 0.6
> The Key Cache should be configurable by absolute numbers
> --------------------------------------------------------
>
> Key: CASSANDRA-801
> URL: https://issues.apache.org/jira/browse/CASSANDRA-801
> Project: Cassandra
> Issue Type: Improvement
> Components: Core
> Affects Versions: 0.6
> Reporter: Ryan King
> Assignee: Stu Hood
> Fix For: 0.6, 0.7
>
> Attachments: 0.6-0001-Convert-KeysCachedFraction-to-KeysCached.patch,
> 0.6-0001-Convert-KeysCachedFraction-to-KeysCached.patch,
> 0.6-0002-Update-config-values-and-defaults.patch,
> 0.6-0002-Update-config-values-and-defaults.patch,
> 0.6-0003-Stop-aggregating-individual-key-caches-into-one.patch,
> 0.6-0004-Migrate-rowCache-into-SSTableTracker.patch,
> 0.6-0005-SSTableTracker-sets-the-cache-for-tracked-Readers.patch,
> 0.6-0006-Remove-JMXAggregatingCache.patch,
> trunk-0001-Convert-KeysCachedFraction-to-KeysCached.patch,
> trunk-0001-Convert-KeysCachedFraction-to-KeysCached.patch,
> trunk-0002-Update-config-values-and-defaults.patch,
> trunk-0002-Update-config-values-and-defaults.patch,
> trunk-0003-Stop-aggregating-individual-key-caches-into-one.patch,
> trunk-0004-Migrate-rowCache-into-SSTableTracker.patch,
> trunk-0005-SSTableTracker-sets-the-cache-for-tracked-Readers.patch,
> trunk-0006-Remove-JMXAggregatingCache.patch
>
>
> Just like the row cache, the key cache should be configurable by absolute
> numbers (not just a fraction). It would be nice to configure it by either
> memory size or entries (like the row cache).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
