giraph-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Claudio Martella" <claudio.marte...@gmail.com>
Subject Re: Review Request 13484: GIRAPH-692: Bracha Toueg deadlock detection algorithm example
Date Mon, 12 Aug 2013 12:36:29 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/13484/#review24993
-----------------------------------------------------------


Nice code, very clean and lean. Three things:

1) it *could* be nice to implement this via the new Computation API, with three classes. It
is not mandatory, but AFAIK no code in the examples exploits it, so this one could have a
double role (showing how to use that API)
2) I think the plural of vertex is "vertices"
3) I'd keep a space between variable type and variable name, for consistency with the rest
of the codebase


- Claudio Martella


On Aug. 11, 2013, 9:59 p.m., Armando wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/13484/
> -----------------------------------------------------------
> 
> (Updated Aug. 11, 2013, 9:59 p.m.)
> 
> 
> Review request for giraph.
> 
> 
> Bugs: GIRAPH-692
>     https://issues.apache.org/jira/browse/GIRAPH-692
> 
> 
> Repository: giraph-git
> 
> 
> Description
> -------
> 
> The Bracha Toueg deadlock detection algorithm is a distributed asynchronous deadlock
detection algorithm which is applied on a wait-for graph which is generated by taking a snapshot
of the graph to be analysed. This is the last patch on which I have worked on. Is significantly
better organized and cleaner than the one before. I have also finally added the N-out-of-M
semantics which can be provided by tagging the edges with the same tag.
> I have added the termination in the algorithm by tweaking the algorithm itself. Another
possibility (maybe cleaner but less efficient) would be to add a termination algorithm (for
example a wave-termination algorithm) initiated by the same initiator of the deadlock detection
algorithm so that all the nodes vote to halt.
> 
> The algorithm passed "mvn verify"
> 
> 
> Diffs
> -----
> 
>   giraph-examples/src/main/java/org/apache/giraph/examples/BrachaTouegDeadlockComputation.java
PRE-CREATION 
>   giraph-examples/src/main/java/org/apache/giraph/examples/io/formats/BrachaTouegDeadlockInputFormat.java
PRE-CREATION 
>   giraph-examples/src/main/java/org/apache/giraph/examples/io/formats/package-info.java
PRE-CREATION 
>   giraph-examples/src/main/java/org/apache/giraph/examples/utils/BrachaTouegDeadlockMessage.java
PRE-CREATION 
>   giraph-examples/src/main/java/org/apache/giraph/examples/utils/BrachaTouegDeadlockVertexValue.java
PRE-CREATION 
>   giraph-examples/src/main/java/org/apache/giraph/examples/utils/package-info.java PRE-CREATION

>   giraph-examples/src/test/java/org/apache/giraph/examples/BrachaTouegDeadlockComputationTest.java
PRE-CREATION 
> 
> Diff: https://reviews.apache.org/r/13484/diff/
> 
> 
> Testing
> -------
> 
> The patch includes a set of tests to verify the correctness. I am checking different
graphs that stimulate several combinations algorithm states.
> 
> 
> Thanks,
> 
> Armando
> 
>


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