giraph-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Raghuraj Mukherjee <Raghuraj.Mukher...@mu-sigma.com>
Subject RE: very important question about partitioning and final combination
Date Thu, 19 Feb 2015 14:16:45 GMT
Hi,

Can we link GIRAPH to some graph DB like Neo4j? The idea is to have the computation done on
GIRAPH while storing the data in graph format itself to save recursive computation. I do know
about Rexster API - http://giraph.apache.org/rexster.html . The problem is that the Rexster’s
architecture diagram shows that even though GIRAPH can read from Neo4j, it won’t be able
to store the computed graph outputs/attributes back on Neo4j. Storing the data on Neo4j is
highly critical since the output might be used for further processing.

Please let me know if we have an more suitable API for converting GIRAPH output from HDFS
to something like a graph DB like Neo4j.


Thanks & Regards
Raghuraj Mukherjee | +91-889-247-3017| raghuraj.mukherjee@mu-sigma.com<mailto:raghuraj.mukherjee@mu-sigma.com>
[cid:image001.png@01CF022C.CA6F92C0]<https://www.facebook.com/musigmaofficial>[cid:image002.png@01CF022C.CA6F92C0]<http://www.linkedin.com/company/mu-sigma>
[cid:image003.png@01CF022C.CA6F92C0] <https://twitter.com/MuSigmaInc>  [cid:image004.png@01CF022C.CA6F92C0]
<https://www.pinterest.com/musigmainc/>



From: Carmen Manzulli [mailto:carmenmanzulli@gmail.com]
Sent: Tuesday, February 10, 2015 9:29 PM
To: user@giraph.apache.org
Subject: Re: very important question about partitioning and final combination

thanks for your help! :) now it's all clearer

2015-02-06 21:03 GMT+01:00 Lukas Nalezenec <lukas.nalezenec@firma.seznam.cz<mailto:lukas.nalezenec@firma.seznam.cz>>:
Hi,
It looks like you need some graph database. Maybe something like neo4j would be better for
you.

Lukas


On 6.2.2015 16:59, Carmen Manzulli wrote:
Hi experts,
i've just two conceptual questions about giraph:
1) because of giraph works on hadoop, if i have more informations about a vertex in different
partitions and also in different machines, for example, for a social graph or a person graph,
about "Alan" (source id) who is (edge value) 24 years old (target id) and, on another machine,
there is a partition with "Alan"(source id) who lives(edge value) in Italy(target id), and
so on: how giraph will  consider these relations if i ask for all people who live in Italy
and are 24 years old ?

2) if giraph doesn't consider relations between same vertex in different partitions and machines
when it runs compute() methos, is it possible to write a  function and make it executable
after all iterations by giraph, so when all vertices are inactive?
thanks and regards,
Carmen.


Disclaimer: http://www.mu-sigma.com/disclaimer.html
Mime
View raw message