Hi,
I was in Santa Clara last week for Cassandra Summit and was talking with Jon
Haddad about Spark after he gave a presentation on it. He was asking what I
thought of GraphX. I said GraphX is sort of a "big hammer" that is only useful
in limited situations where you have a multi-relational to single-relational
projection of your GraphRDD. In short, not really good for adhoc querying (just
standard graph algorithms over a transformed graph). I gave him a demo of
TinkerPop's SparkGraphComputer and he was like: "Whoa. Thats amazing." I was
like: "Yes, of course it is."
http://tinkerpop.incubator.apache.org/docs/3.0.1-incubating/#sparkgraphcomputer
I was wondering if we should get in touch with the Apache Spark guys and see if
they are interested in linking/collaborating on SparkGraphComputer as this
provides Spark a graph query language (Gremlin). If anything, be good for them
to know about it…Jon was mentioning he would like to collaborate on getting
Spark's DataFrame API integrated with SparkGraphComputer so you can (as I
understand it) make "any data blob a graph" and query it with Gremlin.
Thoughts?,
Marko.
http://markorodriguez.com
