Parallel And Distributed Programming Samples - (Multithreading, MPI, Hybrid)
Laplace's equation, a particular 2D partial differential equation (pde), solved using Gauss-Seidel Red/Black Successive Over-Relaxation (SOR) employing following schemes:
- Sequential computation
- Parallel computation using multithreading
- Distributed computation using MPI
- Distributed and Parallel Computation using MPI and multithreading We measure & compare the time taken to solve the pde using each of the above strategies.
Create a task graph to find out the critical path in a distributed computing system to identify the node(s) that might be slowing down the entire system.
Create redundant execution nodes at each rank so that failure of one or more nodes at each rank doesn't stop the execution. Execution fails only when all nodes at a particular rank fail.