[
https://issues.apache.org/jira/browse/MAHOUT-1573?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14036694#comment-14036694
]
ASF GitHub Bot commented on MAHOUT-1573:
----------------------------------------
Github user tdunning commented on the pull request:
https://github.com/apache/mahout/pull/13#issuecomment-46509902
On Tue, Jun 17, 2014 at 10:43 PM, Dmitriy Lyubimov <[email protected]
> wrote:
> Ted, are you ready to help with a concrete alternative? This is a very
> small issue compared to even the patch, lets build a list of alternatives
> and vote. But lets get it done
>
> My additional variants
>
> minSplits,...
> minPar, exactPar, autoPar (consitent with scala's collection.par())
>
> To give something to vote down for Ted
> >=|| :=||
> :||=
>
> Not ok with me
>
> minParts
> minParallelism
> minPartitions
> repartition
> reshuffle
> and other do-something kind
>
minSplits is fine by me. I strongly discourage abbreviations because they
are hard for non-native English speakers to generate well and hard for
non-native English speakers to understand.
Actually, they are often very hard for me to understand and I claim to be a
native English speaker some days.
If more than nine letters is too hard to type (even with an IDE to help
you) then minSplits seems to be reasonable common ground.
> More explicit parallelism adjustments in math-scala DRM apis; elements of
> automatic parallelism management
> ----------------------------------------------------------------------------------------------------------
>
> Key: MAHOUT-1573
> URL: https://issues.apache.org/jira/browse/MAHOUT-1573
> Project: Mahout
> Issue Type: Task
> Affects Versions: 0.9
> Reporter: Dmitriy Lyubimov
> Assignee: Dmitriy Lyubimov
> Fix For: 1.0
>
>
> (1) add minSplit parameter pass-thru to drmFromHDFS to be able to explicitly
> increase parallelism.
> (2) add parrallelism readjustment parameter to a checkpoint() call. This
> implies shuffle-less coalesce() translation to the data set before it is
> requested to be cached (if specified).
> Going forward, we probably should try and figure how we can automate it, at
> least a little bit. For example, the simplest automatic adjustment might
> include re-adjust parallelims on load to simply fit cluster size (95% or 180%
> of cluster size, for example), with some rule-of-thumb safeguards here, e.g.
> we cannot exceed a factor of say 8 (or whatever we configure) in splitting
> each original hdfs split. We should be able to get a reasonable parallelism
> performance out of the box on simple heuristics like that.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
