Performance metrics
Nixon Michaelangelo
Has anyone implemented HyperGraphDB as a backend datastore in an
interactive, low-latency web app instead of an RDBMS?
How does HyperGraphDB perform when compared with Neo4J, AllegroGraph,
BigData, etc?
Where can i find some performance metrics?
Please share your thoughts.
regards,
Nixon
Borislav Iordanov
On Sun, Jun 6, 2010 at 12:37 PM, Nixon Michaelangelo
<nmichae...@gmail.com> wrote:
> Greetings Group,
>
> Has anyone implemented HyperGraphDB as a backend datastore in an
> interactive, low-latency web app instead of an RDBMS?
We have, and here is a publicly accessible example: http://ask.miamidade.gov
> How does HyperGraphDB perform when compared with Neo4J, AllegroGraph,
> BigData, etc?
>
> Where can i find some performance metrics?
>
> Please share your thoughts.
No serious performance metrics have been actually done yet. A few
speculative thoughts: HyperGraphDB is somewhat a middle ground between
a pure graph database and a relational style database. So, for example
graph traversals will be faster than dbs relying on the relational
model, but most likely slower than a pure graph database that
essentially replicates an in-memory pointer structure on disk. On the
other hand, conventional set-oriented querying (like "get me all edges
of such and such type with such and such properties) will be faster
than in a pure graph database.
Using it as an embedded OO/graph/relational/key-value db for a website
is a very good fit. The main thing that you will be missing, and we
hope to remedy that soon, are more powerful pattern-matching (or SQL
style) queries and things like paging and sorting of result sets which
you will need to do at the application level.
Best,
Boris
mtm
It is 20 times faster than JDO with MYSQL. (some graph traversal +
object queries).
// Magnus
On Jun 7, 4:22 pm, Borislav Iordanov <borislav.iorda...@gmail.com>
wrote:
>
> On Sun, Jun 6, 2010 at 12:37 PM, Nixon Michaelangelo
>