Hi, On Wed, Dec 30, 2009 at 4:08 PM, Lorenzo Livi <[email protected]> wrote: > Hi all, > I have two questions about performance. > > 1) I need to calculate the degree of a node (outlinks+inlinks). I can > see that neo4j use lucene to index the node, so each node is a lucene > document, and we can search the nodes by properties. So if I need to > get the degree I can't use the index because the relationships (edges) > are not indexed? It's correct?
Correct. > The getRelationships() method (from Node class or batch inserter mode) > is optimized to do so? The getRelationships methods are optimized to return relationships (for traversal), not to count the number of relationships on a node. > Is there a better way to do this? > I'm thinking abount storing the adiacence of nodes as properties of > the nodes. Then I can lookup the adiacence using lucene. This may be > wrong? Storing the relationship count as a property on the node is a good solution to the problem. I do not know why you would have to index the property using lucene unless you plan to query for for "give me all nodes with relationship count X". > > 2) Do you think that using multiple threads for creating the graph is > a good idea? I mean from performance point of view both using > transaction and batch mode. >From a performance point of view the bottleneck for writes is the disk. If you update different parts of the graph in different threads and have good hardware (raid) you may get a higher insert/update speed. Batch inserter mode does not support concurrent updates. Regards, -Johan _______________________________________________ Neo mailing list [email protected] https://lists.neo4j.org/mailman/listinfo/user
