Fulgora Graph Computer - Where is she hiding...

[Dylan Bethune-Waddell]

I wanted to inquire about FulgoraGraphComputer, which by my understanding is an in-memory graph computer that runs across active Titan instances on your nodes, and you can use just by calling g = graph.traversal(computer()) - I haven't come across much material on its use, design, or limitations. I saw it mentioned at the end of a slideshow or two, referenced in some TP3 JIRA tickets, and also in the Titan JavaDocs - can it do some stuff better than or "orthogonal" to SparkGraphComputer and the rest of the APIs, like bulk graph mutations involving inferring or dropping new edges, vertices, and properties globally, or possibly doing a mass rename/repurpose of schema elements when the data model is adjusted?

Best,

Dylan

[Marko Rodriguez]

Hi,

FulgoraGraphComputer is a single-machine, in-memory graph computer built for Titan. Matthias did some really cool work on compressing immutable OLAP graphs into a tiny memory footprint. The benefit, if the graph you are analyzing can fit within the confines of a single-machine, FulgoraGraphComputer can be used and thus, no socket based message passing as everything is in local RAM. 

Choosing over Fulgora vs. Spark will become more apparent in TinkerPop 3.2.0 when we will allow users to specify which subgraph to pull out of the underlying data source (e.g. "only the friendship network since 2005").

https://issues.apache.org/jira/browse/TINKERPOP3-962

Both Fulgora and Spark support bulk mutations/etc. via BulkLoaderVertexProgram and (soon) "BulkMutationVertexProgram" (https://issues.apache.org/jira/browse/TINKERPOP3-942). Thus, they have the same capabilities, one is just single machine in-memory and the other is multi-machine in-memory.

HTH,

Marko.

http://markorodriguez.com

--
You received this message because you are subscribed to the Google Groups "Aurelius" group.
To unsubscribe from this group and stop receiving emails from it, send an email to aureliusgraph...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/aureliusgraphs/48b01442-0e49-4b84-8e7d-ed68ed032f73%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Dylan Bethune-Waddell]

Gotcha - thanks Marko.