cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Stu Hood (JIRA)" <j...@apache.org>
Subject [jira] Updated: (CASSANDRA-629) antientropyservice patchset nearly doubles compaction time
Date Sun, 13 Dec 2009 08:03:18 GMT

     [ https://issues.apache.org/jira/browse/CASSANDRA-629?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Stu Hood updated CASSANDRA-629:
-------------------------------

    Attachment: 629-2-use-xor-in-merkletree.diff
                629-1-utility-changes.diff

After benchmarking and trying to optimize the current code as much as possible, I gave up
on trying to generate the full range down to hash depth.

Jun suggested using XOR on 193, both for speed and simplicity, but I was worried about hash
collisions. In order to make using XOR a little more secure, I bumped the digests we generate
in AEService to SHA-256, so that we minimize the chance of collision as much as possible.

Time to validate 1 million rows dropped from 84 seconds to 1.9 seconds (including hashing).

> antientropyservice patchset nearly doubles compaction time
> ----------------------------------------------------------
>
>                 Key: CASSANDRA-629
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-629
>             Project: Cassandra
>          Issue Type: Bug
>          Components: Core
>            Reporter: Jonathan Ellis
>            Assignee: Stu Hood
>             Fix For: 0.5
>
>         Attachments: 629-1-utility-changes.diff, 629-2-use-xor-in-merkletree.diff
>
>
> CompactionsTest goes from 2.9s to 5.7s for me, before/after the 193 patchset.  This is
not an acceptable speed hit for when no repair work is actually being done, assuming it is
representative of "real" compactions.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message