Posted: Tue May 30, 2006 11:31 am Post subject: Java compute node (or) ESQL compute node?
Voyager
Joined: 26 Mar 2004 Posts: 76
For development in Message Broker, we have to choose between the tried and tested ESQL compute node and Java Compute node.
I have gone thru the MB infocenter and found that both nodes are pretty similar in functionality except that in Java node we cant access some of the Broker environment Properties.
Also there is XPATH support in Java node which helps in coding and in accessing elements.
The main thing that is against the Java compute node is that as we are coding in java the problem of Garbage collection eventually comes up. Is this problem applicable to this Java compute node too? Has anyone tried performance testing with Java compute node?. How did it fared against the ESQL node? How has IBM addressed the garbage collection issue in Java node?
Also is there any additional functionality available in ESQL compute node not available in Java node? I'm sure many of you wud have gone thru the process/trouble of choosing between ESQL node and Java node in development. Upon what factors you made a particular decision?
Any pointers on the above would be really helpful.
My understanding is that the Java compute node should have roughly the same performance characteristics as the ESQL compute node. This is supported by the performance report, as I recall.
On the other hand, it's my belief that it's easier to code ESQL to perform well than to code the same stuff in Java to perform well.
For garbage collection - the rule with the Java plugin node was to delete everything that you allocated after you called propagate - at least in terms of messages and assemblies. I'm sure this holds true for Java compute nodes. _________________ I am *not* the model of the modern major general.
But is there any apparent limitations in Java compute node apart from the inability to access some of the Broker properties?
Bcoz right now we are not able to visiualize clearly now on what capabilities we will need in development. In the case of general transformation, parsing etc both nodes can be used interchangeably. But i'm not sure abt the other aspects.
We did our previous project in MB5.0 using ESQL coding. But now as we have java compute node with same performance level as ESQL we are thinking of giving it a try.
True. In java there is always a possibility (especially with inexperienced java developers) to code not keeping performance in mind, like not deleting unused objects etc.
But i just wanted to know abt the limitations in java compute node and also abt how the garbage collection is handled.
I guess this forum also has some folks from IBM. So any sort of official info wud also be welcome.
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum