[
https://issues.apache.org/jira/browse/CASSANDRA-2045?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13012813#comment-13012813
]
Jonathan Ellis commented on CASSANDRA-2045:
-------------------------------------------
You want to use the pointer approach when your ratio of overwrites : row size
is sufficiently high -- the biggest win there is when you can turn dozens or
hundreds of mutations, into replay of just the latest version.
Not sure what the best way to estimate that is -- Brandon suggested checking
SSTable bloom filters on writes. Which is probably low-overhead enough,
especially if we just do it only every 10% of writes for instance. I kind of
like that idea, I think it will be useful in multiple places down the road.
("Sufficiently high" depends on SSD vs magnetic -- time to introduce a
postgresql-like random vs sequential penalty setting?)
> Simplify HH to decrease read load when nodes come back
> ------------------------------------------------------
>
> Key: CASSANDRA-2045
> URL: https://issues.apache.org/jira/browse/CASSANDRA-2045
> Project: Cassandra
> Issue Type: Improvement
> Reporter: Chris Goffinet
> Fix For: 0.8
>
>
> Currently when HH is enabled, hints are stored, and when a node comes back,
> we begin sending that node data. We do a lookup on the local node for the row
> to send. To help reduce read load (if a node is offline for long period of
> time) we should store the data we want forward the node locally instead. We
> wouldn't have to do any lookups, just take byte[] and send to the destination.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
