We have a highly parallel application that will be scaled to work in a clustered environment. We had planned on using JINI and JavaSpaces as our technology of choice for this development but having just discovered Terracotta I am investigating its possible application for this problem.
However, before I even get started I wanted to first ask you guys if Terracotta is suitable for parallelizing an application (my understanding is that it is)?
The parallel area of the application looks something like this:
for dataItems = 0 to 1,000,000
for itemFeature = 0 to 1000
sharedData3[itemFeature] += sharedData1[itemFeature] * sharedData2[dataItems][itemFeature]
end
end
So calculations and updates of shared data within nested loops, nothing fancy but inherantly parallel.
If Terracotta can help me scale this in a clustered environment without code changes that would make it worthwhile evaluating.
Thanks!
Ian
|