Split an action and run on the servers

[Venkat]

Hi,



Ours is a windows based application and I have to perform an
action which takes much time (as it requires much CPU time) on a single pc.
I want to split the action to more than two and I want to run the action on
different servers and mix the result at end and do the rest on the result.



Can any one tell me how to split an action and run on servers?



Cheers,

Venkat

 

[Guest]

Venkat,
The concept you describe of "split an action" is far too generic to get an
appropriate answer. There are literally dozens of different ways to handle
distributed processing, and each has a particular approach that is usually
designed to fit the work that needs to be done.

Perhaps if you could be more specific about exactly what your business logic
of "an action" actually is, you could get some better responses.
Peter

 

[Venkat]

Hi Peter,

Thanks for your reply. I have a method which fetches information from
database (mostly thousands of records). I do some calculations based on the
data and display it to the user. Here when I run this action on my PC it is
taking much time. So I am looking to use Distributed processing to divide
the data into multiple blocks and run separately on different servers. Once
the calculation is done merge the results and display back it to the user.

Hope it gives much information about my problem.

Thanks in advance.

 

[John Timney \( MVP \)]

Have you identified where the delay is caused, is it heavy processing that
may be improved by more efficient code, or is it a slow machine, could it be
improved by multi-threading the application or making more use of stored
procs for example, could data caching help?.

If it were me I would look at something like PVM (Parallel virtual
machines), they are designed to do this type of work if you want a very
scalable solution. That said, you will be coding in C probably so the
easiest approach may be to have your application populate some form form of
job queue, allowing each distributed application to pick up their share of
the work and write the results to a central DB table. When all work was
processed the app could then inspect the table and present the results.

--
Regards

John Timney
Microsoft MVP