[
https://issues.apache.org/jira/browse/FLINK-1725?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14727532#comment-14727532
]
ASF GitHub Bot commented on FLINK-1725:
---------------------------------------
Github user tillrohrmann commented on the pull request:
https://github.com/apache/flink/pull/1069#issuecomment-137140076
@gdfm, thanks for the great explanation. A good way to visualize what's
happening.
I'm just wondering why you shouldn't be able to connect more than 2
containers, let's say 3. In cases where you have an extremely high data skew,
this might be helpful. Imagine that you have 10 containers and only 2 of them
are full. Then in the best case you'll get 4 half filled containers after
connecting two containers. But this still leaves 6 unused containers. Wouldn't
it be better to connect for example 5 containers in this case? Then in the best
case you would use all available containers. But of course this strongly
depends on your actual data and therefore I'd vote to make the number of
connected containers configurable.
> New Partitioner for better load balancing for skewed data
> ---------------------------------------------------------
>
> Key: FLINK-1725
> URL: https://issues.apache.org/jira/browse/FLINK-1725
> Project: Flink
> Issue Type: Improvement
> Components: New Components
> Affects Versions: 0.8.1
> Reporter: Anis Nasir
> Assignee: Anis Nasir
> Labels: LoadBalancing, Partitioner
> Original Estimate: 336h
> Remaining Estimate: 336h
>
> Hi,
> We have recently studied the problem of load balancing in Storm [1].
> In particular, we focused on key distribution of the stream for skewed data.
> We developed a new stream partitioning scheme (which we call Partial Key
> Grouping). It achieves better load balancing than key grouping while being
> more scalable than shuffle grouping in terms of memory.
> In the paper we show a number of mining algorithms that are easy to implement
> with partial key grouping, and whose performance can benefit from it. We
> think that it might also be useful for a larger class of algorithms.
> Partial key grouping is very easy to implement: it requires just a few lines
> of code in Java when implemented as a custom grouping in Storm [2].
> For all these reasons, we believe it will be a nice addition to the standard
> Partitioners available in Flink. If the community thinks it's a good idea, we
> will be happy to offer support in the porting.
> References:
> [1].
> https://melmeric.files.wordpress.com/2014/11/the-power-of-both-choices-practical-load-balancing-for-distributed-stream-processing-engines.pdf
> [2]. https://github.com/gdfm/partial-key-grouping
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
