views:

153

answers:

1

Hi I am developing an algorithm to find out a fraud-detecting system using a math formula. I need to work the program in different computer. How can I do that in Java? How can I do parallel programing in java? using RMI? or any other method? can u guys give me a code sample?

+2  A: 

First figure out how to split your algorithm so that it can be divided and conquered in parallel. That's the hard part.

There are probably frameworks for doing the rest. For example, Hazelcast claims to have an implementation of java.util.concurrent.ExecutorService that works over a cluster. Implement your program using a simple executor service first and you can probably drop in that framework, configure it, and away you go.

Mark Peters
After Dividing the Algorithm can i make parallel processing in RMI?
Why the focus on RMI?
David Zaslavsky
Not Focusing, i know RMI Better, so...any scope in that?
@Binu you can do it using RMI, and some frameworks might use RMI, but even RMI might be more low-level than you need to go. An example isn't really well-suited to SO though IMO, it's too complex to put in an answer here.
Mark Peters
@Mark Thks for reply,what about JPPF framework?, can i use that?
@Binu: Are you seriously asking if you can use a parallel computing framework for parallel computing? Yes, I'd imagine you could. You know better than me, it's your project.
Mark Peters
@Mark: What i Know is My project Need So many computers to Work out,about 100 PC, i dont know how to implement is
@Binu: I would start with the documentation for one of these frameworks then. I can't do your project for you, and I haven't worked with these frameworks, and as I already said a full-out example of using the frameworks is probably out of scope for StackOverflow.
Mark Peters
@Mark: K Mark, Thks for the info, even thought, i am waiting for some one who will gave me another info ;)
@Binu, Analysis Paralysis is a real problem. I think @Mark Peters' initial intuition of implementing the algorithm with an ExecutorService first and then later dropping in Hazelcast (or some other framework) is a good approach.
Tim Bender