In this case, by implementations, I just meant various permutations of the serialization format. Harness was probably to grand of a term. I was thinking a set of benchmarks that could be used to compare different variations on our internal StarGraph serialization as we try them out, really just to make comparison while developing easier.
On Mon, Jun 27, 2016 at 4:49 AM, Stephen Mallette <spmalle...@gmail.com> wrote: > Ted, when you say: > > "benchmarking harness so that it would be easy to swap various > implementations" > > do you mean like gremlin-benchmark would be useful to other graph > providers? like UniPop could run gremlin-benchmark over its implementation > and compare it to Titan? > > On Fri, Jun 17, 2016 at 5:23 PM, Ted Wilmes <twil...@gmail.com> wrote: > > > If folks thought it was a good idea, I'd love to put together a little > > benchmarking harness so that it would be easy to swap various > > implementations & tweaks in and out to measure how they perform. For > > example, benchmarks running against different #'s of vertex properties, > > edge counts, # of properties / edge, measuring things like latency, size > in > > memory, and size on disk. I need to finish up TinkerPop-1254, but then I > > could take a crack at this. I already have done some work towards this > in > > TinkerPop-1287 to confirm improvements from stream removal in key spots > so > > perhaps I could extend upon that. > > > > --Ted > > > > On Fri, Jun 17, 2016 at 3:49 PM, Marko Rodriguez <okramma...@gmail.com> > > wrote: > > > > > Hi, > > > > > > TinkerPop 3.3.0 is not slated for anytime soon, but some buddies are > > > interested in a making the serialization format of StarGraph more > > > efficient. Given it would be a major breaking change, we can’t do it > till > > > TinkerPop 3.3.0, but we can talk about it and design it. Here is a > ticket > > > to get us underway. Your thoughts are more than welcome: > > > > > > https://issues.apache.org/jira/browse/TINKERPOP-1343 < > > > https://issues.apache.org/jira/browse/TINKERPOP-1343> > > > > > > Take care, > > > Marko. > > > > > > http://markorodriguez.com > > > > > > > > > > > > > > >
