How fast should OrientGraph.query() be?

[Nick DeYoung]

I have an orient db 2.0.10 community server up and running

I have created a simple test where I have placed 300,000 vertices in an orient database each with a property "testProperty" with a value - UUID.randomUUID().toString

I have created an index on the property

  graph.createKeyIndex("testProperty", classOf[Vertex])  // this is scala :)

In a  test I create an OrientGraph connected via remote:127.0.0.1:2424/test-database 

and do 

g.query().has("testProperty", "cbc6b219-9df3-44c5-90b2-ab7a6bd9bf7e") 

This takes 20 seconds to return the vertex with that UUID. 

In another test I do 

g.getRawGraph.query(new OSQLSynchQuery[ODocument]("select * from INDEX:V.testProperty where key = 'cbc6b219-9df3-44c5-90b2-ab7a6bd9bf7e'"))

which takes 1 millisecond to return the vertex. 

Is the java graph query api really this slow or am I doing something wrong? 

[W. Craig Trader]

Nick ...

Much like you told the SQL engine to use an index (select * from INDEX:V.testProperty ...), you need to explicitly use indexes from the graph API, as follows:

Java:

OrientVertex result = g.getVertices( "V.testProperty", "cbc6b219-9df3-44c5-90b2-ab7a6bd9bf7e" ).iterator().next();

Gremlin:

GremlinPipeline p = g.getVertices( "V.testProperty", "cbc6b219-9df3-44c5-90b2-ab7a6bd9bf7e" )._()

I think you'll find that using indexes with the Java API is even faster than the SQL query API.

- Craig -

--

---
You received this message because you are subscribed to the Google Groups "OrientDB" group.
To unsubscribe from this group and stop receiving emails from it, send an email to orient-databa...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Nick DeYoung]

Craig, 

I have tried that, it takes 11 ms. but that is not a Pipe - (g.getVertices())

our code base is leveraging another TinkerPop enabled graph databse and are researching the idea of switching it out. 

my research is trying to figure out how does orient perform on the generic Tinkerpop API. 

I want to do things like g.V().has("someProperty","someValue").outE("someLabel")....somemoreStepsInMyPipe..

and know that that is going to hit indexes. 

Does it ?

this is a simple test of the different ways I am accessing the graph and their timing on 300,000 vertices (not connected at the momen)

 time("orient graph api 'orient specific'") {

            val v = g.getVertices("V.testProperty", "cbc6b219-9df3-44c5-90b2-ab7a6bd9bf7e")

        } 11 ms

        time("tinkerpop pipeline 'generic'") {

          val v = new GremlinPipeline(g, true).V("testProperty", "cbc6b219-9df3-44c5-90b2-ab7a6bd9bf7e").next()

        } 21 seconds

        time ("another blueprints? " ) {

          val v = g.query().has("testProperty", "cbc6b219-9df3-44c5-90b2-ab7a6bd9bf7e").vertices()

        } 19 seconds

        time("sql count all 'orient specific'") {

          val v:OResultSet[ODocument] = g.getRawGraph.query(

            new OSQLSynchQuery[ODocument]("select * from INDEX:V.testProperty where key = 'cbc6b219-9df3-44c5-90b2-ab7a6bd9bf7e'")

          ) 1 ms

        }

        // throws an error in distrubuted mode 

//        time ("gremlin command 'generic' ") {

//          val r = g.getRawGraph.command(new OCommandGremlin("g.V().has('testProperty')")).execute()

//        }

If I understand Orient's documentation right, I am going to have to either change our pipes (traversals in tinkerpop) into SQL+ queries in order to leverage the claims of orient's speed. 

If i'm wrong in this, let me know

--

---
You received this message because you are subscribed to a topic in the Google Groups "OrientDB" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/orient-database/Ff5Fbxi6p20/unsubscribe.
To unsubscribe from this group and all its topics, send an email to orient-databa...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--

// ndy 

[W. Craig Trader]

Nick ...

(1) I don't work for OrientDB, (2) I am not familiar with the Tinkerpop APIs outside of OrientDB.

OrientGraph.getVertices() returns an iterator, not a pipe. Gremlin adds a DSL method _() that will turn an iterator into a pipeline, from which you can do the normal Gremlin queries.

AFAIK, the OrientDB Java/Groovy/Gremlin APIs require you to explicitly select an index to use. The SQL queries can either explicitly specify an index (SELECT * from INDEX:...) or they may take advantage of an existing, properly specified, index, maybe. This is an area where OrientDB still needs a lot of work, IMO.  Or at least a lot more documentation.

FWIW, I've been working with databases all my career, and I've never moved from one database to a different but comparable database without having to make changes to some aspect of my application. Blueprints is nice, but it is not a cross-graph-database panacea.

- Craig -

[Nick DeYoung]

" Blueprints is nice, but it is not a cross-graph-database panacea."

there is much wisdom in this, that I think the graph community could use to hear from.