Can you try it to with an embedded setup? Sent from mobile device
Am 05.07.2014 um 12:19 schrieb ashish jindal <[email protected]>: > it is hosted as a webapp in tomcat7 with maxThreads = 100 and minSpareThreads > = 10 > and the load test shoots N threads using java.util.concurrent.ExecutorService > with fixed size of N threads. > > On Saturday, July 5, 2014 2:34:31 PM UTC+5:45, Michael Hunger wrote: >> >> How do you schedule the threads? >> >> Sent from mobile device >> >> Am 05.07.2014 um 10:28 schrieb ashish jindal <[email protected]>: >> >>> Hi Michael, >>> I tried what you suggested, it has increased the performance by a factor of >>> ~5 but the time spent still increase linearly with the number of threads. >>> >>> On Saturday, July 5, 2014 2:22:47 AM UTC+5:45, ashish jindal wrote: >>>> >>>> let me try that. >>>> >>>> On Saturday, July 5, 2014 2:15:38 AM UTC+5:45, Michael Hunger wrote: >>>>> >>>>> I just saw. ... If you have already n >>>>> >>>>> Then it should be much faster to check outgoing rels by property than a >>>>> index lookup. >>>>> >>>>> Sent from mobile device >>>>> >>>>> Am 04.07.2014 um 22:25 schrieb ashish jindal <[email protected]>: >>>>> >>>>>> yes, each thread has its own transaction. >>>>>> >>>>>> On Saturday, July 5, 2014 2:08:45 AM UTC+5:45, Michael Hunger wrote: >>>>>>> >>>>>>> But tge tx is created per thread? >>>>>>> >>>>>>> >>>>>>> >>>>>>> Sent from mobile device >>>>>>> >>>>>>> Am 04.07.2014 um 22:20 schrieb ashish jindal <[email protected]>: >>>>>>> >>>>>>>> It is in a single transaction, which is on the outermost layer. >>>>>>>> >>>>>>>> On Saturday, July 5, 2014 2:01:44 AM UTC+5:45, Michael Hunger wrote: >>>>>>>>> >>>>>>>>> What is your transactional scope? >>>>>>>>> >>>>>>>>> Michael >>>>>>>>> >>>>>>>>> Sent from mobile device >>>>>>>>> >>>>>>>>> Am 04.07.2014 um 22:06 schrieb ashish jindal <[email protected]>: >>>>>>>>> >>>>>>>>>> Hi Michael, >>>>>>>>>> number of rels returned is 10-50 . Yes, i close the hits iterator as >>>>>>>>>> soon as its purpose is served. >>>>>>>>>> >>>>>>>>>> On Saturday, July 5, 2014 1:08:09 AM UTC+5:45, Michael Hunger wrote: >>>>>>>>>>> >>>>>>>>>>> How many rels are returned from the call? >>>>>>>>>>> >>>>>>>>>>> Do you close the hits afterwards to release resources? >>>>>>>>>>> >>>>>>>>>>> Sent from mobile device >>>>>>>>>>> >>>>>>>>>>> Am 04.07.2014 um 10:51 schrieb ashish jindal <[email protected]>: >>>>>>>>>>> >>>>>>>>>>>> Hi, >>>>>>>>>>>> I am relatively new to neo4j. I have implemented a traversal >>>>>>>>>>>> algorithm using neo4j as underlying graph storage. >>>>>>>>>>>> Graph has about 40K nodes and about 2M edges. >>>>>>>>>>>> Using 2.1.2 community version of neo4j embedded. It is hosted as >>>>>>>>>>>> service in tomcat7 with allocated 6Gb of memory on 14.04 Ubuntu. >>>>>>>>>>>> Following is a piece of code which does numeric index queries. >>>>>>>>>>>> RelationshipIndex numericIndex = >>>>>>>>>>>> graphDb.index().forRelationships(Constants.RELATIONSHIP_INDEX); >>>>>>>>>>>> public Relationship getRelationship(Node n, int lowerlimit, int >>>>>>>>>>>> upperlimit) throws Exception { >>>>>>>>>>>> >>>>>>>>>>>> long t1 = System.currentTimeMillis(); >>>>>>>>>>>> IndexHits<Relationship> hits = >>>>>>>>>>>> numericIndex.query(QueryContext.numericRange("start_time", >>>>>>>>>>>> lowerlimit, upperlimit, true, true), n, null); >>>>>>>>>>>> long t2 = System.currentTimeMillis(); >>>>>>>>>>>> time_spent_in_numeric_index_queries += t2-t1; >>>>>>>>>>>> // global variable >>>>>>>>>>>> // do something on hits and return a >>>>>>>>>>>> relationship from it >>>>>>>>>>>> } >>>>>>>>>>>> >>>>>>>>>>>> for a single threaded traversal >>>>>>>>>>>> time_spent_in_numeric_index_queries ~= 1 sec. >>>>>>>>>>>> but if i do multiple traversals ( same traversal ) in parallel , >>>>>>>>>>>> time_spent_in_numeric_index_queries increases linearly . for 10 >>>>>>>>>>>> calls in parallel it takes ~= 5 sec. >>>>>>>>>>>> Also, i am only doing read operations throughout . >>>>>>>>>>>> I want to know, how does querying work? Does it each query lock >>>>>>>>>>>> the entire index? >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> Thanks, >>>>>>>>>>>> Ashish >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> -- >>>>>>>>>>>> You received this message because you are subscribed to the Google >>>>>>>>>>>> Groups "Neo4j" group. >>>>>>>>>>>> To unsubscribe from this group and stop receiving emails from it, >>>>>>>>>>>> send an email to [email protected]. >>>>>>>>>>>> For more options, visit https://groups.google.com/d/optout. >>>>>>>>>> >>>>>>>>>> -- >>>>>>>>>> You received this message because you are subscribed to the Google >>>>>>>>>> Groups "Neo4j" group. >>>>>>>>>> To unsubscribe from this group and stop receiving emails from it, >>>>>>>>>> send an email to [email protected]. >>>>>>>>>> For more options, visit https://groups.google.com/d/optout. >>>>>>>> >>>>>>>> -- >>>>>>>> You received this message because you are subscribed to the Google >>>>>>>> Groups "Neo4j" group. >>>>>>>> To unsubscribe from this group and stop receiving emails from it, send >>>>>>>> an email to [email protected]. >>>>>>>> For more options, visit https://groups.google.com/d/optout. >>>>>> >>>>>> -- >>>>>> You received this message because you are subscribed to the Google >>>>>> Groups "Neo4j" group. >>>>>> To unsubscribe from this group and stop receiving emails from it, send >>>>>> an email to [email protected]. >>>>>> For more options, visit https://groups.google.com/d/optout. >>> >>> -- >>> You received this message because you are subscribed to the Google Groups >>> "Neo4j" group. >>> To unsubscribe from this group and stop receiving emails from it, send an >>> email to [email protected]. >>> For more options, visit https://groups.google.com/d/optout. > > -- > You received this message because you are subscribed to the Google Groups > "Neo4j" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > For more options, visit https://groups.google.com/d/optout. -- You received this message because you are subscribed to the Google Groups "Neo4j" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
