[
https://issues.apache.org/jira/browse/CASSANDRA-5515?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13713789#comment-13713789
]
Tyler Hobbs commented on CASSANDRA-5515:
----------------------------------------
bq. I think we do need to persist this one across restarts, e.g., in a system
table. (Maybe just a Map would be fine, since pulling 1000s of entries into
memory once on startup to initialize state should be okay.)
I take it we will want to periodically sync the system table with the latest
counts? (Just relying on some kind of shutdown hook seems like a bad idea.)
Given that there can be 10's of thousands of sstables, I think we'll want to
throttle that sync.
By "a Map", I assume you mean a map column collection. I am a little concerned
about pulling in a map with 10's of thousands of entries all at once, even if
it's just on startup.
As far clearing entries goes, directly deleting when the sstable is removed
combined with a TTL of 10 days to handle odd cases seems reasonable to me.
Thoughts?
> Track sstable coldness
> ----------------------
>
> Key: CASSANDRA-5515
> URL: https://issues.apache.org/jira/browse/CASSANDRA-5515
> Project: Cassandra
> Issue Type: New Feature
> Components: Core
> Reporter: Jonathan Ellis
> Assignee: Tyler Hobbs
> Fix For: 2.0.1
>
> Attachments: 0001-Track-row-read-counts-in-SSTR.patch
>
>
> Keeping a count of reads per-sstable would allow STCS to automatically ignore
> cold data rather than recompacting it constantly with hot data, dramatically
> reducing compaction load for typical time series applications and others with
> time-correlated access patterns. We would not need a separate age-tiered
> compaction strategy.
> (This will really be useful in conjunction with CASSANDRA-5514.)
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
