Global Graph Analytics - andrew-nguyen/titan GitHub Wiki

Titan is designed to compute numerous, concurrent, short-lived, local graph traversals. Titan is not designed for global graph analysis. In order to understand the difference, a itemization of each term above is defined below.

• Numerous concurrent interactions: the graph is being queried, in parallel, by different users solving different problems.
• Short-lived transactions: each interaction with Titan is intended to be real-time on the order of hundreds of milliseconds.
• Local graph traversals: each transaction is touching a small fraction of the entire graph and as such is solving a problem within a localized region of the graph.

Titan is a multi-client database in the classic sense. For performing global graph analytics, Titan relies on Faunus. Faunus is a batch processing framework that is optimized for computing a small number of concurrent, long running, global graph traversals.

Please review Faunus documentation for instructions on how to connect Titan to Faunus. However, a short example is provided below. With Faunus it is possible to use Titan as the source of a Hadoop MapReduce job chain. Like Titan, Faunus uses Gremlin as its traversal language. However, Faunus compiles its Gremlin statements down to MapReduce jobs. Assume the Titan/Cassandra maintains The Graph of the Gods. The Faunus/Gremlin shell can be used to create a FaunusGraph. When the traversal is evaluated, Titan serves as the data source and Hadoop serves as the data processing system.

faunus$ bin/gremlin.sh 

         \,,,/
         (o o)
-----oOOo-(_)-oOOo-----
gremlin> g = FaunusFactory.open('bin/titan-cassandra-input.properties')
==>faunusgraph[titancassandrainputformat]
gremlin> g.V.count()
12/12/15 23:58:38 INFO mapreduce.FaunusCompiler: Compiled to 1 MapReduce job(s)
12/12/15 23:58:38 INFO mapreduce.FaunusCompiler: Executing job 1 out of 1: MapSequence[com.thinkaurelius.faunus.mapreduce.transform.VerticesMap.Map, com.thinkaurelius.faunus.mapreduce.util.CountMapReduce.Map, com.thinkaurelius.faunus.mapreduce.util.CountMapReduce.Reduce]
...
12/12/15 23:59:16 INFO mapred.JobClient:     Total committed heap usage (bytes)=269619200
12/12/15 23:59:16 INFO mapred.JobClient:     Combine input records=12
12/12/15 23:59:16 INFO mapred.JobClient:     SPLIT_RAW_BYTES=120
12/12/15 23:59:16 INFO mapred.JobClient:     Reduce input records=1
12/12/15 23:59:16 INFO mapred.JobClient:     Reduce input groups=1
12/12/15 23:59:16 INFO mapred.JobClient:     Combine output records=1
12/12/15 23:59:16 INFO mapred.JobClient:     Reduce output records=0
12/12/15 23:59:16 INFO mapred.JobClient:     Map output records=12
==>12
gremlin>