[
https://issues.apache.org/jira/browse/MAHOUT-319?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13018508#comment-13018508
]
Jake Mannix commented on MAHOUT-319:
------------------------------------
Saikat,
Once this patch is applied, it will make the code for snapshotting both
Lanczos and DistributedLanczos the same: the new class "LanczosState" will
handle persisting itself, and we can subclass or componentize that class in a
variety of ways: a abstract base subclass like PersistentLanczosState which has
a serializer/deserializer object injected into it, and then we can write a
variety of serializer/deserializers: HDFS/local disk/DB, etc.
Apply this patch directly to your checkout if you want to see how it works in
the meantime until this hits trunk.
> SVD solvers should be gracefully stoppable/restartable
> ------------------------------------------------------
>
> Key: MAHOUT-319
> URL: https://issues.apache.org/jira/browse/MAHOUT-319
> Project: Mahout
> Issue Type: Improvement
> Components: Math
> Affects Versions: 0.3
> Reporter: Jake Mannix
> Assignee: Jake Mannix
> Attachments: MAHOUT-319.patch
>
>
> LanczosSolver, DistributedLanczosSolver, and HebbianSolver all keep copious
> amounts of memory-resident data which is lost if the app crashes or is killed
> (OOM, forgetting to run in a screen session, and losing net connectivity to
> the server running it, etc...).
> These algorithms (and many other Mahout processes!) should enable a pluggable
> "persist state" mechanism (to HDFS, RDBMS, local disk, key-value store, etc),
> and similarly, a way to pick up and start from such a state.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
