How to improve performance in titan for the given graph

[Vijayalakshmi S B]

Hi,

I have a graph which has around 14000 a type nodes, each has relation with 4 x type nodes (total nodes around 1500+). 

Similarly, I have a node b  type which has relation with 1500+ x type nodes

In the computation,

1.  I need to get 1500+ x type nodes and compute values (fetching property from x type node)

2. Fetch all a type node with filter such has not "imageurl != "" , type = 'a', ...

3. For each node of a type, fetch x type nodes

4. and then compute final score 

i have added indexes of type composite on id and type, vertex centric index on the edge between b type and its property

when i run my algorithm with above mentioned graph design

 for (1000 + x type nodes, 3500 a type nodes)its taking around 6 to 7 minutes and the expectation is in milli sec

Can someone help , how can i improve performance. 

what kind of indexes should be added

It would be great help

Thanks in advance

Vijayalakshmi

 

[Bryn Cooke]

Hi,

Please can detail the gremlin queries that you are using? For instance the filter step will not use the indexes, you need to use 'has'.

Bryn