Execution of load balancing algorithm on cloud using matlab and distributed computing server.

I have written a load balancing algorithm on MatLab. i want to execute it on cloud using Matlab and 6 machines. I know that MATLAB provides the Parallel computing tool for this purpose. Please any one is there to guide me how should i proceed.

[Akula Nishanth]

hai can you provide me the code for the algorithm which you have written

[Abdulrauf Gara]

Please help with code of load balance on matlab Thank you

Answer 1

[Walter Roberson]

parfor() manages loads automatically, so you possibly want to work with spmd() and use labSend() to send new work to a worker.

Note: you will need the Distributed Computing license in order to work with multiple machines.

I do not know if the Distributed Computing has facilities to build your own scheduler; it might.

[Hemant Petwal]

i have distributed computing server with license and six machine on which matlab client is installed. Now please tell me how to proceed for next set up. what should i need to do.

[Walter Roberson]

If you are wanting to do all the system-level work of monitoring processes, determining workloads, and so on, your own job scheduler, then see http://www.mathworks.com/help/mdce/configure-parallel-computing-products-for-a-generic-scheduler.html

The alternative is that you could define your own "work queue" routines and start spmd labs and use labSend() and labReceive() to distribute tasks from the workers and get back results.

Are you interested in doing things like monitoring system workload including non-MATLAB work, and watching out for servers that have hung, and worry about disks filling up, and all the other overhead of running reliable computer systems? If so then you should create your own generic scheduler.

If not, if you are interested in "well-behaved" ideal systems and you are doing something like modelling industrial process systems that build parts at different rates, then model each industrial component as a lab and communicate with it using the lab* routines and let the DCS handle the overhead.