Hi Stack, I'm not really an expert in Titan or S2Graph, but I believe user-supplied IDs would be a big change for Titan. As for the abstraction layer, I've read that DataStax specifically moved away from it for DSE Graph as it prevented them from taking advantage of store-specific features. Also Titan uses a wide table schema, whereas S2Graph and Zen both use a tall table schema. I think I prefer a tall table schema for graph databases, especially in the context of social networks where the incident edge count can be quite high.
I believe S2Graph is looking to implement TinkerPop, but I'm not sure when it will be completed. https://issues.apache.org/jira/browse/S2GRAPH-72 Regards, Robert On Thu, Nov 10, 2016 at 10:16 AM, Stack <[email protected]> wrote: > On Thu, Nov 10, 2016 at 6:28 AM, Robert Yokota <[email protected]> wrote: > > > Hi, > > > > If anyone is interested in using HBase to work with graphs, I wrote a > > TinkerPop layer for HBase: > > > > https://rayokota.wordpress.com/2016/11/10/hgraphdb-hbase- > > as-a-tinkerpop-graph-database/ > > > > https://github.com/rayokota/hgraphdb > > > > > Thanks for posting here Robert so others learn of your work. > > Out of interest, no hurry: > > + Would it have taken a while to add userids to Titan (not withstanding > current flux around project Titan)? > + You didn't like the abstraction layer because adds complexity, dents > perf? > + Did you try reaching out to the good folks of s2graph to see if tinkerpop > API of interest? > > Thanks again for the notice, > S >
