Sign up for our newsletter and get the latest HPC news and analysis.

Interview: IBM’s David Ungar on End-to-End Non-Determinism

In this interview, cloud consultant Miha Ahronovitz discusses parallel programming with Dr. David Ungar from IBM Research. Ungar is author of the popular article Many Core processors: Everything You Know (about Parallel Programming) Is Wrong!

Miha Ahronovitz: In parallel distributed computing, using a product like Grid Engine we got different results if the round robin servers were processing some sequential jobs mixed with parallel, or if we dedicated the servers to parallel MPI processing exclusively. We didn’t know why at the time. How can you explain this?

David Ungar: I think the problem you described, is that you have different results if you configure the servers differently. One of the principles in our project is what we call “end-to-end non-determinism”. The idea is if you go into these parallel systems to get performance – this holds for multi-core, many-core, distributed, then you need to take an approximate route.

Read the Full Story.

Resource Links: