[
https://issues.apache.org/jira/browse/MAPREDUCE-2841?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Todd Lipcon updated MAPREDUCE-2841:
-----------------------------------
Attachment: fb-shuffle.patch
I forward-ported Facebook's shuffle from their github 'production' branch into
MR2. There are a few changes I had to make to adjust to the latest interfaces,
and some counter-related stuff I commented out. I also fixed a perf issue where
it spent a lot of time creating new Configuration objects. But the core of the
code is the same.
The patch also has a simple benchmark similar to what Binglin described above
-- a standalone piece of code to exercise the mapside sort/spill with nothing
else in the way. This makes it very easy to compare different implementations.
In my testing, the FB implementation is about 10% faster on a wall clock basis,
and perhaps a bit better on CPU. I didn't spend a lot of time looking at
results as of yet - but figured I'd post this here in case anyone else was
interested.
This is in no way meant for commit - just sharing code so other people can
tweak. If someone has a version of the native mapside sort already available
against trunk MR2, it would be great to have that available on the JIRA so we
can compare all of the options.
> Task level native optimization
> ------------------------------
>
> Key: MAPREDUCE-2841
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-2841
> Project: Hadoop Map/Reduce
> Issue Type: Improvement
> Components: task
> Environment: x86-64 Linux/Unix
> Reporter: Binglin Chang
> Assignee: Binglin Chang
> Attachments: DESIGN.html, MAPREDUCE-2841.v1.patch,
> MAPREDUCE-2841.v2.patch, dualpivot-0.patch, dualpivotv20-0.patch,
> fb-shuffle.patch
>
>
> I'm recently working on native optimization for MapTask based on JNI.
> The basic idea is that, add a NativeMapOutputCollector to handle k/v pairs
> emitted by mapper, therefore sort, spill, IFile serialization can all be done
> in native code, preliminary test(on Xeon E5410, jdk6u24) showed promising
> results:
> 1. Sort is about 3x-10x as fast as java(only binary string compare is
> supported)
> 2. IFile serialization speed is about 3x of java, about 500MB/s, if hardware
> CRC32C is used, things can get much faster(1G/
> 3. Merge code is not completed yet, so the test use enough io.sort.mb to
> prevent mid-spill
> This leads to a total speed up of 2x~3x for the whole MapTask, if
> IdentityMapper(mapper does nothing) is used
> There are limitations of course, currently only Text and BytesWritable is
> supported, and I have not think through many things right now, such as how to
> support map side combine. I had some discussion with somebody familiar with
> hive, it seems that these limitations won't be much problem for Hive to
> benefit from those optimizations, at least. Advices or discussions about
> improving compatibility are most welcome:)
> Currently NativeMapOutputCollector has a static method called canEnable(),
> which checks if key/value type, comparator type, combiner are all compatible,
> then MapTask can choose to enable NativeMapOutputCollector.
> This is only a preliminary test, more work need to be done. I expect better
> final results, and I believe similar optimization can be adopt to reduce task
> and shuffle too.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
