Re: [eclipselink-users] Performance questions EclipseLink vs. Hibernate

[James Sutherland <jamesssss@xxxxxxxxx>]

This is very very odd, in our extensive performance testing we have done
comparing with Hibernate we have greatly exceed their performance is almost
every possible usecase we have tested.  Could you please include more
details on exactly what your test is doing and exactly what your environment
is (db, os, jvm, server, jdbc).

My guess is you issue is related one of the following,
1 - You do not have weaving enabled in EclipseLink.  EclipseLink requires
this to do lazy loading and change tracking.  You must enable weaving either
through the agent, or static weaver.
2 - If using Oracle Hibernate enabled batch writing by default, where as you
must turn in on in EclipseLink through a persistence property.
3 - Do you have a security manager set in your jvm?  There is a setting you
need to set currently for this in EclipseLink.




Leon Derks-2 wrote:
> 
> Hello,
> 
> Today we noticed differences in performance between EclipseLink and 
> Hibernate.
> 
> We had made up a test case with a big list of objects(Entities). Each 
> Enitity has references to other Entities / list of Entities.
> Some of these Entities are shared and others are new.
> 
> To save one Entity from the main list took about 780 milliseconds to 
> persist the whole Object graph into the db.
> There was not much difference between EclipseLink and Hibernate in this 
> case.
> 
> But when we iterated through the big list of entities and persisted them 
> one by one, there was a big difference between EclipseLink and Hibernate.
> We tried one transaction for the whole list and also a single 
> transaction per Entity in the list.
> 
> Hibernate succeeded in every case and was much faster then EclipseLink 
> (sometimes 50% faster).
> EclipseLink only succeeded when saving one Entity per transaction, 
> during the other cases (saving the whole list of objects in one 
> transaction) we did get "java.lang.OutOfMemoryError: Java heap space".
> EclipseLink used almost 15 minutes to complete and Hibernate did the 
> same in 9 minutes.
> 
> Based on our results we would go for Hibernate at the moment. But are 
> there ways to speed up performance in EclipseLink?
> Why is there so much difference in performance between Hibernate and 
> EclipseLink.
> 
> Leon
> 


-----
---
http://wiki.eclipse.org/User:James.sutherland.oracle.com James Sutherland 
http://www.eclipse.org/eclipselink/
 EclipseLink ,  http://www.oracle.com/technology/products/ias/toplink/
TopLink 
Wiki:  http://wiki.eclipse.org/EclipseLink EclipseLink , 
http://wiki.oracle.com/page/TopLink TopLink 
Forums:  http://forums.oracle.com/forums/forum.jspa?forumID=48 TopLink , 
http://www.nabble.com/EclipseLink-f26430.html EclipseLink 
Book:  http://en.wikibooks.org/wiki/Java_Persistence Java Persistence 
-- 
View this message in context: http://www.nabble.com/Performance-questions-EclipseLink-vs.-Hibernate-tp16768260p16807905.html
Sent from the EclipseLink - Users mailing list archive at Nabble.com.