The HPC Asia 2009 keynote speech given by Dr. William T.C. Kramer introduces us to the NCSA beginning with its history in field of HPC and continuing on with its latest development, the Blue Waters project. The talk also examines the concepts behind the open science-based projects and applications that are slated to run over Blue Waters. Dr. Kramer then concludes with his insight into Petascale and Exascale era challenges of the future.
Part1 | Part2 | Part3 | Part4 | Part5 | Part6 | Part7 | Part8 [PDF Download] [Video]
Part 3: What Users Want
So, what do users want from these machines? We often hear about performance and you heard a very good discussion this morning about the TOP500 list. The people who use these machines absolutely want high performance. They want to get the most work out of the machine that they can possibly get. But they also want other things as well. High performance is not just a single metric. It is a set of things that you look for in a machine that I call a holistic approach to judging the best types of architectures for science.
The next thing that people want is effectiveness. They want to be able to have their work run at the right time when they need it. They want to be able to have the machine scheduled. They wanna have the machine working for them and working at a reasonably high level of performance. That’s close to the best case scenario.
Reliability is another very important aspect of the machines. The machine has to be able to do the work when the users want to use it. And that’s becoming an increasing challenge as we move into the Petascale and eventually the Exascale era. Consistency is something that doesn’t get discussed very much but it is also very important to the user community in the sense that they have an expectation of how long it will take for a machine to do a certain amount of work. It should take about the same amount of time to do about the same amount of work. And we see in these very complex HPC Systems that very often, a mis-operating component will generate a wide range of inconsistencies and variations. That will not only make the work go slower but it will also perturb the effectiveness of scheduling and, indeed, reliability as well. So consistency is another attribute that people need.
And the last attribute is usability. That is, how easy it is to make the machine operate and perform well for their particular purposes. How much do they have to do to move from a different machine to a new machine, How much work to they have to do to generate an algorithm on a smaller machine or a workstation and move it to a bigger machine? All these are attributes of what our HPC facilities need to address to make the science and engineering community productive in their use of high performance computing.