Director of Collaborative eScience Programs at the National Center for Supercomputing Applications (NCSA) at the University of Illinois
Behind a lot of the research success that you read on the pages of HPCwire are the people dedicating their lives to making these discoveries – and behind these people are the individuals who dedicate their lives to ensuring that there is enough computing power to make these discoveries possible. John Towns, a leader with the National Science Foundation (NSF), National Center for Supercomputing Applications (NCSA), and XSEDE – the Extreme Science and Engineering Discovery Environment, is one of these people. This year, John will collaborate on the merging of two of the largest cyberinfrastructure projects in existence – XSEDE and Blue Waters. HPCwire caught up with John to find out more, and to get his perspective on the trends in HPC.
HPCwire: John, you wear a lot of hats. You’re involved with so much, from the eScience Programs at the NCSA, to the XSEDE project with the NSF and more. The list goes on. Is there a theme in your professional life that you recognize that you believe tells the story of your work in HPC?
John Towns: Having been at NCSA for more than 25 years now, I certainly have developed the need for a rather large hat rack in my office. HPC has always been a centerpiece of the efforts I have been involved in, though the scope of those activities has certainly expanded significantly over time. Underlying it all has always been the fact that I derive a great deal of personal satisfaction from being able to play a role in the research successes of others. This has ranged from code optimization and scaling in the early 1990s, to deployment of productive and effective HPC resources in the late 1990s and 2000s, on to development and coordination of large scale research computing resources, infrastructure and support services at national and international levels.
HPCwire: One of the more interesting things going on with you is the collaboration that you are working on between XSEDE and Blue Waters to bring together two of the National Science Foundation’s largest cyberinfrastructure projects. Can you tell us the significance of this new alignment, and what it means to the world at large?
John Towns: It would seem natural and simple for these two projects to be highly coordinated given that NCSA leads both of them. As it turns out, this is somewhat harder than expected primarily due to the differences in goals for each project—Blue Waters is strongly focused on a small number of users executing very large-scale runs on HPC resources while XSEDE is more broadly focused on the larger community of researchers needing computational resources to support their work.
What we recognize, however, is that there is still significant overlap with the set of users of Blue Waters with the community that XSEDE supports and a need for coordinating their use of Blue Waters with resources they are using in the XSEDE portfolio. Alignment certainly facilitates this for the set of users spanning these environments, but it also facilitates XSEDE and Blue Waters being able to leverage one another’s activities. In addition, from an XSEDE perspective, it allows insight into a broader set of community needs as input into our ongoing development of national/international cyberinfrastructure.
HPCwire: You’ve told us that one of your hopes is to see the rise of the integration of multiple resources, saying that too many people in the HPC community put emphasis on monolithic HPC systems. Can you talk a little about the importance of this?
John Towns: At the risk of being misunderstood, I am happy to comment on this. As a community we are too driven by a small number of metrics that we use to indicate success. In particular, we tend to place more importance on certain benchmarks than we should when what is really important to most of those making use of the resources and services we provide is their productivity and accomplishing research, education, development or other goals.
Accomplishing these requires many things, not just single HPC systems. Lacking a simple and straightforward mechanism to measure productivity that applies across many endeavors, complicated by the fact that very few vendors provide anything more than components to the overall environment in which folks work, we focus on performance of those individual components and particularly on floating point performance. This de-emphasizes the importance of the productivity of the overall environment and the ability of the components of that environment to operate in an integrated fashion. To be clear, performance of individual systems is very important, but it is only one facet of the complex ecosystems in which we work.
HPCwire: On a personal note, can you talk about your personal life? Your family, background, any hobbies?
John Towns: As a single parent of four sons, my personal life has been, to say the least, a bit challenging and hectic. Outside of the demands of work, there has been little time for hobbies. I have been lucky to find a partner willing to put up with my personal life which she has described as a reality TV show. (Note: I would prefer to watch a reality TV show than to live one.)
Much of my available time for a good number of years has gone into coaching soccer—a sport I am passionate about—and this has allowed me to spend a lot of time with my sons in the process. As a player since early childhood and then as a coach, this sport has taught me many life lessons and allowed me to share what I have learned with so many players.
As my sons have gotten older (the youngest is now 18), I have found time to put into a couple of my other hobbies, cooking and candle making. With four athletic sons, cooking is a handy hobby; the candle making is purely for fun. I do want to try my hand at one more activity I am very interested in: brewing beer. I have the equipment at hand and just need to carve out a bit more time.
HPCwire: One last question – is there anything about yourself that you can share that you think your colleagues would be surprised to learn?
John Towns: Well, I suspect I may have done that with my response to the last question; there are not many candle makers out there.