[
https://issues.apache.org/jira/browse/HADOOP-2560?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12644062#action_12644062
]
Matei Zaharia commented on HADOOP-2560:
---------------------------------------
Instead of using a limit N on the number of splits, I'd have a limit on the
number of MB of input data, call it L. Then a default value can be set for the
cluster as a whole and it will do something reasonable in each job. If some
jobs have really lightweight mappers they can also choose to increase it.
I would first look to see if there are enough data-local input chunks to make L
MB of total data. If not, put all of the data-local ones go on to rack-local
ones. This is strictly better than picking just random rack-local splits.
> Processing multiple input splits per mapper task
> ------------------------------------------------
>
> Key: HADOOP-2560
> URL: https://issues.apache.org/jira/browse/HADOOP-2560
> Project: Hadoop Core
> Issue Type: Bug
> Reporter: Runping Qi
> Assignee: dhruba borthakur
> Attachments: multipleSplitsPerMapper.patch
>
>
> Currently, an input split contains a consecutive chunk of input file, which
> by default, corresponding to a DFS block.
> This may lead to a large number of mapper tasks if the input data is large.
> This leads to the following problems:
> 1. Shuffling cost: since the framework has to move M * R map output segments
> to the nodes running reducers,
> larger M means larger shuffling cost.
> 2. High JVM initialization overhead
> 3. Disk fragmentation: larger number of map output files means lower read
> throughput for accessing them.
> Ideally, you want to keep the number of mappers to no more than 16 times the
> number of nodes in the cluster.
> To achive that, we can increase the input split size. However, if a split
> span over more than one dfs block,
> you lose the data locality scheduling benefits.
> One way to address this problem is to combine multiple input blocks with the
> same rack into one split.
> If in average we combine B blocks into one split, then we will reduce the
> number of mappers by a factor of B.
> Since all the blocks for one mapper share a rack, thus we can benefit from
> rack-aware scheduling.
> Thoughts?
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
