Mass / Group Delete OrientDB / Java

问题

I am trying to prototype performance results of orientdb mass delete of vertexes. I need to prototype trying to delete more than 10000 upto million.

Firstly, I am using the light weight edges property to be false while creating my vertices and edges following this Issue with creating edge in OrientDb with Blueprints / Tinkerpop

When I try deleting (Please see below the code)

private static OrientGraph graph = new OrientGraph(
        "remote:localhost/WorkDBMassDelete2", "admin", "admin");
private static void removeCompleatedWork() {

    try {
        long startTime = System.currentTimeMillis();
        List params = new ArrayList();
        String deleteQuery = "delete vertex Work where out('status') contains (work-state = 'Not Started')";
        int no = graph.getRawGraph().command(new OCommandSQL(deleteQuery))
                .execute(params);
        // graph.commit();
        long endTime = System.currentTimeMillis();
        System.out.println("No of activities removed : " + no
                + " and time taken is : " + (endTime - startTime));
    } catch (Exception e) {
        e.printStackTrace();
    } finally {
        graph.shutdown();
    }
}

The Results are good if I tring to delete in 100's 500 activities take ~500 ms. But when i trying to delete 2500/5000 activities the no's are high for 2500 deletions it takes ~6000.

A) I also tried creating index. What is the best practise to create a index on the attribute work-state or to create index on the edge status? I tried both while creating the vertex and edge. But both are not improving the performance a lot.

((OrientBaseGraph) graph).createKeyIndex("Status", Edge.class);
//or on the vertex
((OrientBaseGraph) graph).createKeyIndex("work-state", Vertex.class);

What is the best practice to delete mass/group data using the query like mentioned above? Any help is greatly appreciated.

UPDATE:

I downloaded orientdb-community-1.7-20140416.230539-144-distribution.tar.gz from https://oss.sonatype.org/content/repositories/snapshots/com/orientechnologies/orientdb-community/1.7-SNAPSHOT/.

When I try deleting using the subquery from the studio / program I get the following error:com.orientechnologies.orient.core.sql.OCommandSQLParsingException: Error on parsing command at position #0: Class 'FROM was not found . I had modified my query like this:

delete vertex from (select in('status') from State where work-state = 'Complete')

Also while I ran it through program I updated my maven dependencies to 1.7-SNAPSHOT libraries. My old query was still producing the same numbers and the subquery deletion was giving errors even in studio. Please let me know if I am missing anything. Thanks !!

回答1:

First, please try the same exact code with 1.7-SNAPSHOT. It should be faster.

Then in 1.7-SNAPSHOT we just added the ability to delete vertices from a sub-query. This is because why browsing all the Work when you could delete all the incoming vertex from the Status vertex "Not Started"?

So if you've 1.7-SNAPSHOT change this query from:

delete vertex Work where out('status') contains (work-state = 'Not Started')

to (assuming the status vertex is called "State"):

delete vertex from (select in('status') from State where work-state = 'Not Started')

来源：https://stackoverflow.com/questions/23156580/mass-group-delete-orientdb-java

标签

java

performance

graph-databases

OrientDB