giraph-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Maja Kabiljo" <majakabi...@fb.com>
Subject Review Request: Properly size netty buffers when encoding requests
Date Wed, 05 Dec 2012 20:05:56 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/8361/
-----------------------------------------------------------

Review request for giraph.


Description
-------

For some type of requests, it's easy to calculate the exact size of requests, and create the
buffers as big as we need.


This addresses bug GIRAPH-443.
    https://issues.apache.org/jira/browse/GIRAPH-443


Diffs
-----

  http://svn.apache.org/repos/asf/giraph/trunk/giraph/src/main/java/org/apache/giraph/comm/netty/handler/RequestEncoder.java
1417254 
  http://svn.apache.org/repos/asf/giraph/trunk/giraph/src/main/java/org/apache/giraph/comm/requests/ByteArrayRequest.java
1417254 
  http://svn.apache.org/repos/asf/giraph/trunk/giraph/src/main/java/org/apache/giraph/comm/requests/SaslTokenMessageRequest.java
1417254 
  http://svn.apache.org/repos/asf/giraph/trunk/giraph/src/main/java/org/apache/giraph/comm/requests/SendPartitionCurrentMessagesRequest.java
1417254 
  http://svn.apache.org/repos/asf/giraph/trunk/giraph/src/main/java/org/apache/giraph/comm/requests/SendPartitionMutationsRequest.java
1417254 
  http://svn.apache.org/repos/asf/giraph/trunk/giraph/src/main/java/org/apache/giraph/comm/requests/SendVertexRequest.java
1417254 
  http://svn.apache.org/repos/asf/giraph/trunk/giraph/src/main/java/org/apache/giraph/comm/requests/SendWorkerMessagesRequest.java
1417254 
  http://svn.apache.org/repos/asf/giraph/trunk/giraph/src/main/java/org/apache/giraph/comm/requests/WritableRequest.java
1417254 
  http://svn.apache.org/repos/asf/giraph/trunk/giraph/src/main/java/org/apache/giraph/utils/ByteArrayVertexIdMessages.java
1417254 

Diff: https://reviews.apache.org/r/8361/diff/


Testing
-------

Using this patch, here is the performance gain on one of our large applications:
- about 10% of overall superstep time
- about 25% of compute time
- about 10% less GC

Passes mvn verify.


Thanks,

Maja Kabiljo


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message