hadoop-mapreduce-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From jeremy p <athomewithagroove...@gmail.com>
Subject Re: Will hadoop always spread the work evenly between nodes?
Date Tue, 19 Mar 2013 20:49:36 GMT
Thank you, Jeff.  Actually, it turns out the scenario you outlined is much
closer to the situation I'm dealing with.  I'm going to ask a separate
question about that.

--Jeremy

On Wed, Mar 13, 2013 at 1:47 PM, Jeffrey Buell <jbuell@vmware.com> wrote:

> I think in your case it will have to be even, because all the slots will
> get filled.  A more interesting case is if you have 40 nodes, will you get
> exactly 5 slots used for each of the nodes?  Or will some nodes get more
> than 5 mappers, and others less?  I don't remember the details, but I've
> had problems with unevenness in such scenarios.  At least in MR1, you can
> usually force evenness by adjusting the number of map and reduce slots per
> node.  In MR2 the slots are combined so achieving evenness will be more
> difficult.
>
> Jeff
>
> ------------------------------
> *From: *"jeremy p" <athomewithagroovebox@gmail.com>
> *To: *user@hadoop.apache.org
> *Sent: *Wednesday, March 13, 2013 1:01:46 PM
> *Subject: *Will hadoop always spread the work evenly between nodes?
>
>
> Say I have 200 input files and 20 nodes, and each node has 10 mapper
> slots.  Will Hadoop always allocate the work evenly, such that each node
> will get 10 input files and simultaneously start 10 mappers?  Is there a
> way to force this behavior?
>
> --Jeremy
>
>

Mime
View raw message