[
https://issues.apache.org/jira/browse/TEZ-3673?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15983850#comment-15983850
]
Siddharth Seth commented on TEZ-3673:
-------------------------------------
bq. TEZ_RUNTIME_UNORDERED_PARTITIONED_KVWRITER_BUFFER_PARTS
Think this is better as a fraction/percent (instead of buffers will be divided
by this). Easier to understand.
bq. new SpillCallable(filledBuffers, codec, spilledRecordsCounter,
spillPathDetails))
Nit: Clone filledBuffers right here to ensure we don't accidentally change it
in a future patch (e.g. calling a filledBuffers.clear instead of allocating a
new instance)
Tests should verify numSpills.
Other than that, looks good.
> Allocate smaller buffers in UnorderedPartitionedKVWriter
> --------------------------------------------------------
>
> Key: TEZ-3673
> URL: https://issues.apache.org/jira/browse/TEZ-3673
> Project: Apache Tez
> Issue Type: Improvement
> Reporter: Harish Jaiprakash
> Assignee: Harish Jaiprakash
> Attachments: TEZ-3673.01.patch, TEZ-3673.02.patch
>
>
> UnorderedPartitionedKVWriter allocates in bigger chunks. It may or may not
> get filled up. In PipelinedSorter, we start off with 32MB chunks. But
> UnorderedPartitionedKVWriter can be worse as it allocates bigger blocks. Need
> to revisit this allocation.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
