Title: High Performance User - Level Thread Migration for SMP Clusters


Abstract:

The need to compute processor intensive algorithms is not a modern day issue. It has been around since the early days of programming.

Supercomputers and mainframe computers have been proposed from the engineering side. However these solutions are incredibly expensive. In
this literature review, we are proposing a software based solution.

Rather than using a large and expensive machines, our proposition is to use a cluster of workstations. An application, in order to execute on the said
cluster, should be written using the thread-based model and submitted as a job on a single workstation. The threads spawned off by the original
process will be distributed to execute on the workstations present in the cluster. This problem will be tackled using thread migration where a thread living
on one workstation will be migrated to another workstation and resume its process seamlessly. Strong migration will be applied where along with the thread, its state will be migrated and no code will be executed more than
once. Using this network of workstations, we will be able to provide enough processing power for processor intensive algorithms at a much cheaper
price then supercomputers and mainframes.