I would like to pose a question to see if HAMA could be a practical solution to a problem I am facing. Any help would be appreciated.
I am interested in the performance of several thousand (2,000 or 3,000) independent Cholesky decompositions on a large symmetric positive definite matrix (10,000 x 10,000 up to 20,000 x 20,000). Processing time is the most important performance goal (under 5-10 minutes) for all 2,000 or 3,000 decompositions. Is HAMA a viable solution for achieving this task. If so, do you have a ballpark estimate of the size and power of nodes needed to achieve something like this? Is this testable on amazon ec2? I appreciate any help in advance. Thanks, Mark Brown
