I've read that "The clustering features of J2EE are intended to distribute requests from many simultaneous clients, not to distribute a large single computation (to be run in sandboxed local user environments) to many individual machines." Does that mean for scientific
computations which contain single but heavy computing tasks,EJB infrastructure is of little significance because it can't make its nodes solve one computation together?
-
EJB clustering features (6 messages)
- Posted by: spears tian
- Posted on: August 29 2003 12:51 EDT
Threaded Messages (6)
- EJB clustering features by Paul Strack on August 29 2003 13:51 EDT
- EJB clustering features by Ian Mitchell on August 29 2003 14:27 EDT
- Distributed Computation by Rob Misek on August 29 2003 16:05 EDT
- Distributed Computation by Ian Mitchell on August 29 2003 16:20 EDT
- Distributed Computation by Cameron Purdy on August 29 2003 05:55 EDT
- about distributed computations in Cohenrence by spears tian on August 29 2003 23:09 EDT
- Distributed Computation by Ian Mitchell on August 29 2003 16:20 EDT
-
EJB clustering features[ Go to top ]
- Posted by: Paul Strack
- Posted on: August 29 2003 13:51 EDT
- in response to spears tian
That is correct. EJBs are optimized to support many distributed clients, not to support distributed computations. -
EJB clustering features[ Go to top ]
- Posted by: Ian Mitchell
- Posted on: August 29 2003 14:27 EDT
- in response to Paul Strack
Distributed computation usually relates to parallel processing and its associated architectures. Generally, this discipline tackles the problem of defining a contiguous distributed address space for algorithmic computation. J2EE solves a very different set of problems: the encapsulation of remote computational objects via facades (servlets, clusters) and proxying (EJB's).
I suppose you could say that J2EE distributes the black box, while a parallel architecture distributes the white box... -
Distributed Computation[ Go to top ]
- Posted by: Rob Misek
- Posted on: August 29 2003 16:05 EDT
- in response to spears tian
Hi Spears,
Coherence provides the ability to perform complex distributed computations within the Coherence cluster. Through the Invocation Service (our grid-computing support) you can 'invoke' tasks on each node (or a subset of nodes) within the cluster in a synchronous or asynchronous manner. You can also configure the number of threads being utilized for the task on each node.
Some of our customers are using this feature to bring overnight computations into the 'almost-real-time' realm of performance.
Later,
Rob Misek
Tangosol Coherence: Cluster your Work. Work your Cluster.
Coherence Forums. -
Distributed Computation[ Go to top ]
- Posted by: Ian Mitchell
- Posted on: August 29 2003 16:20 EDT
- in response to Rob Misek
"Work your Cluster"?
Oh well, its more memorable than your mate Cameron's sig! ;-P -
Distributed Computation[ Go to top ]
- Posted by: Cameron Purdy
- Posted on: August 29 2003 17:55 EDT
- in response to Ian Mitchell
-
about distributed computations in Cohenrence[ Go to top ]
- Posted by: spears tian
- Posted on: August 29 2003 23:09 EDT
- in response to Rob Misek
Hi,Rob,
If a EJB cluster cannot disassemble a large single computation among its nodes,it seems that there is little meaning registering EJB services as Grid services.Of course we can get a Grid Service Handle and manage the life cycle in a soft-state way,but I think what really accounts is :Grid computing make us use comprehensive computing resources to solve complicated computations,then how can we perform that if our business logic is based on EJB infrastructure?