Hi Alex, I just have one general comment regarding the read workloads and that is > that they appear to be very "global", that is, the operations access most > or all of the graph dataset: Find neighbors OF ALL NODES, Find adjacent > edges OF ALL EDGES, etc. Graph databases are definitely capable of this > type of workload, but it is not where they truly shine, nor is it what real > world usage looks like. Graph databases are designed with interactive, low > latency queries in mind; similar in nature to how you would use a > relational database but far more capable at querying connected, complex > data.
Besides the global queries you mention I utilized also a FindShortestPath query, which run just for 100 nodes. Also, the proposed clustering algorithm uses a lot of not-global queries, suchs as "get neighbors of the nodes with a specific community id property". I am aware that global graph traversing is not very frequent use case scenario, but I though it would be useful to benchmark the databases with global queries too. There are some analytics that need global traversing, such as the neighborhood distribution, the average clustering coefficient of the graph. If you're interested, Neo Technology is one of the major contributors in an > EU-funded research project tasked with developing graph database benchmarks > (participating in this is my daily job). For more information about the > project: > > - Site: http://ldbcouncil.org/ > > > - Github: https://github.com/ldbc > > > - Specification for one of the benchmarks we are developing: > https://github.com/ldbc/ldbc_snb_docs > > This seems really interesting. I'll take a look and if anything comes up I'll drop you an email. On a related topic, I saw that LDBC uses synthetic data from with social network characteristics. In the last days I'm trying to find either a dataset or a generator that models re-tweet chains in Twitter. Have you worked on this topic? Cheers, Sotiris On Wed, Sep 24, 2014 at 9:51 AM, Alex Averbuch < [email protected]> wrote: > Hi Sotiris, > Sorry about the belated reply, I read your paper with interest about a > month ago but didn't see this post until now. > Interesting work and results, thanks for sharing (both the links and code)! > I just have one general comment regarding the read workloads and that is > that they appear to be very "global", that is, the operations access most > or all of the graph dataset: Find neighbors OF ALL NODES, Find adjacent > edges OF ALL EDGES, etc. > Graph databases are definitely capable of this type of workload, but it is > not where they truly shine, nor is it what real world usage looks like. > Graph databases are designed with interactive, low latency queries in mind; > similar in nature to how you would use a relational database but far more > capable at querying connected, complex data. > > If you're interested, Neo Technology is one of the major contributors in > an EU-funded research project tasked with developing graph database > benchmarks (participating in this is my daily job). For more information > about the project: > > - Site: http://ldbcouncil.org/ > - Github: https://github.com/ldbc > - Specification for one of the benchmarks we are developing: > https://github.com/ldbc/ldbc_snb_docs > > Feel free to ask any questions. > Best, > Alex > > On Wed, Sep 3, 2014 at 12:36 PM, Sotiris Beis <[email protected]> wrote: > >> Hi all, >> >> we have released a grapah database benchmark that includes Neo4j. If you >> like take a look here https://github.com/socialsensor/graphdb-benchmarks >> . You can find the presentation here >> http://www.slideshare.net/sympapadopoulos/adbis2014-presentation. Feel >> free to suggest any fix in order to be as objective as possible. >> >> Thanks in advance, >> Sotiris >> >> -- >> You received this message because you are subscribed to the Google Groups >> "Neo4j" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to [email protected]. >> For more options, visit https://groups.google.com/d/optout. >> > > -- > You received this message because you are subscribed to a topic in the > Google Groups "Neo4j" group. > To unsubscribe from this topic, visit > https://groups.google.com/d/topic/neo4j/s9KdpkzCahM/unsubscribe. > To unsubscribe from this group and all its topics, send an email to > [email protected]. > For more options, visit https://groups.google.com/d/optout. > -- You received this message because you are subscribed to the Google Groups "Neo4j" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
