giraph-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jyoti Yadav <rao.jyoti26ya...@gmail.com>
Subject Fwd: About Giraph Design
Date Wed, 06 Nov 2013 15:50:37 GMT
---------- Forwarded message ----------
From: Jyoti Yadav <rao.jyoti26yadav@gmail.com>
Date: Wed, Nov 6, 2013 at 9:08 PM
Subject: Re: About Giraph Design
To: ssc@apache.org


Hi ..
in your earlier posts,you mentioned that running
SimpleShortestPathsComputation example given in giraph examples for v
number of times where v=total no of vertices in graph,(ie. realizing
AllPairShortestPath),is not a scalable approach ...But  for measuring
performance purpose, how can make it possible??

in configuration option,we can set source id =1 as shown below in the code
snippet,

but let me know how we can change this configuration option ie. source
id,every time we want to execute for all vertices??

public static final LongConfOption SOURCE_ID =
      new LongConfOption("SimpleShortestPathsVertex.sourceId", 1,
          "The shortest paths id");


Any idea??

Thanks
Jyoti


On Mon, Nov 4, 2013 at 8:27 PM, Jyoti Yadav <rao.jyoti26yadav@gmail.com>wrote:

> Thanks Sebastian...I will try this ..
>
>
> On Mon, Nov 4, 2013 at 4:50 PM, Sebastian Schelter <ssc@apache.org> wrote:
>
>> It is possible: You have to represent each matrix as a graph and "tag"
>> the elements so that you know to which matrix they belong.
>>
>> On 04.11.2013 10:48, Jyoti Yadav wrote:
>> > Hi Sebastian ...
>> > You suggested me to implement LineRank algorithm..In this algorithm one
>> > large matrix is divided into two sub  matrices.Is there any way to give
>> > these two matrices as input to Giraph.?.Does Giraph design support
>> this..?
>> >
>> > Thanks
>> > Jyoti
>> >
>> >
>> > On Mon, Oct 28, 2013 at 9:59 PM, Sebastian Schelter <ssc@apache.org>
>> wrote:
>> >
>> >> You could implement the LineRank algorithm, which was proposed as a
>> >> scalable substitute for betweeness centrality:
>> >>
>> >> http://www.cs.cmu.edu/~ukang/papers/CentralitySDM2011.pdf
>> >>
>> >> --sebastian
>> >>
>> >> On 28.10.2013 17:09, Jyoti Yadav wrote:
>> >>> Thanks Sebastian for your reply.
>> >>> If we want to implement Betweenness Centrality algo then what should
>> we
>> >> do?
>> >>>
>> >>> Regards
>> >>> Jyoti
>> >>>
>> >>>
>> >>> On Mon, Oct 28, 2013 at 6:10 PM, Sebastian Schelter <ssc@apache.org>
>> >> wrote:
>> >>>
>> >>>> Hi Jyoti,
>> >>>>
>> >>>> All-Pairs-Shortest-Path is a problem with a solution quadratic in
the
>> >>>> number of vertices of the input graph. For a reasonably large graph,
>> you
>> >>>> cannot even store the result.
>> >>>>
>> >>>> For most algorithms that map well to systems like Giraph the amount
>> of
>> >>>> messages per iteration is linear in the number of edges of the graph
>> and
>> >>>> the number of iterations necessary is small by default or bound
by
>> the
>> >>>> diameter of the graph.
>> >>>>
>> >>>> --sebastian
>> >>>>
>> >>>>
>> >>>> On 28.10.2013 05:09, Jyoti Yadav wrote:
>> >>>>> Hi..
>> >>>>> one doubt is disturbing me..Would anyone suggest an idea?
>> >>>>>
>> >>>>> Is GIRAPH is designed for solving problems like AllPairShortestPath
>> ?
>> >>>>>
>> >>>>> Thanks in advance.
>> >>>>>
>> >>>>> Jyoti
>> >>>>>
>> >>>>
>> >>>>
>> >>>
>> >>
>> >>
>> >
>>
>>
>

Mime
View raw message