Group Leader, Future Technologies Group
Jeffrey S. Vetter, Ph.D., is a Distinguished R&D Staff Member, and the founding group leader of the Future Technologies Group in the Computer Science and Mathematics Division of Oak Ridge National Laboratory. Vetter also holds a joint appointment at the Electrical Engineering and Computer Science Department of the University of Tennessee-Knoxville. From 2005 through 2015, Vetter held a Joint position at Georgia Institute of Technology, where, from 2009 to 2015, he was the Principal Investigator of the NSF Track 2D Experimental Computing XSEDE Facility, named Keeneland, for large scale heterogeneous computing using graphics processors, and the Director of the NVIDIA CUDA Center of Excellence.
Vetter earned his Ph.D. in Computer Science from the Georgia Institute of Technology. He joined ORNL in 2003, after stints as a computer scientist and project leader at Lawrence Livermore National Laboratory, and postdoctoral researcher at the University of Illinois at Urbana-Champaign. The coherent thread through his research is developing rich architectures and software systems that solve important, real-world high performance computing problems. In particular, he has been investigating the effectiveness of next-generation architectures, such as graphics processors, massively multithreaded processors, non-volatile memory systems, heterogeneous multicore processors, and field-programmable gate arrays (FPGAs), for key applications. His recent books, entitled “Contemporary High Performance Computing: From Petascale toward Exascale (Vols. 1 and 2),” survey the international landscape of HPC.
Vetter is a Fellow of the IEEE, and a Distinguished Scientist Member of the ACM. Vetter, as part of an interdisciplinary team from Georgia Tech, NYU, and ORNL, was awarded the Gordon Bell Prize in 2010. Also, his work has won awards at major conferences: Best Paper Awards at the International Parallel and Distributed Processing Symposium (IPDPS) and EuroPar, Best Student Paper Finalist at SC14, and Best Presentation at EASC 2015. In 2015, Vetter served as the Technical Program Chair of SC15.
HPCwire: As someone who’s played a role in spearheading GPU use on very-high end machines, including Titan, what’s your sense of the expanding role of GPUs specifically and accelerated computing more generally? Do you think we will see machines such as NVIDIA’s DGX-1 (essentially a GPU-based supercomputer) moving into mainstream environments?
Jeffrey Vetter: I think that the trend toward specialized architectures including heterogeneous computing and SoCs is natural, given the current and expected limitations of CMOS scaling. The mobile and embedded systems communities have been balancing these system parameters for decades (but for different reasons like battery life). GPUs were created for the specialized market of graphics, but then they made inroads into the HPC and then Deep Learning markets with relatively minor adjustments to their architectures and programming models. I believe that GPUs will continue to grow, but other architectures, such as FPGAs, will also gain more traction in HPC. Two other trends that are equally as important in the short term are tighter integration of these components and emerging memory systems. First, integration is important because it can facilitate heterogeneity, improve performance, and lower power requirements. We already see this integration in the consumer devices with commodity processors that include CPUs, GPUs, and specialized engines for encryption or encoding. Second, memory systems are beginning to fragment and diversify in a similar way to heterogeneous computing. DDR4, GDDR, HMC, HBM, nonvolatile memory, and others occupy specific points in the design space, and now we see them emerging in systems to help satisfy requirements system balance. All of these trends will make it vitally important to develop programming systems that facility performance portability, or our applications teams will suffer.
HPCwire: Fresh from participating in the SC’16 Post Moore’s Law Era workshop, what are the most interesting and perhaps more distant technologies you see emerging and why? Is it quantum or neuromorphic computing, about which quite a bit is already known, or something else whose glimmer outshines others? What is it about them that excites your imagination?
My co-chair, Satoshi Matsuoka, and I were very surprised by the strong interest from SC16 attendees. We had a large number of submissions, and a full room most of the day. Our contributors presented many interesting computing ideas including silicon photonics, neuromorphic, neuro-memristive, quantum, reconfigurable, probabilistic, and approximate. The design space is so broad and the timescales so dissimilar that it often makes it tough to even compare and contrast the approaches with any confidence. There is also levels of disruption that will influence each technologies acceptance. In his talk on the IEEE Rebooting Computing Initiative, Tom Conte presented an informative chart showing the expected levels of disruption to the computing stack for the individual technologies like carbon nanotubes, quantum, and neuromorphic computing. For example, some of the technologies like silicon photonics could potentially be substituted for electrical circuits without regard to the higher levels of software and applications. However, other technologies like quantum or neuromorphic computing may require a comprehensive rethinking of the entire computing stack. We may need to reformulate the algorithms on new languages or libraries to gain full advantage of the new underlying hardware. This ‘switching cost’ will be a major impediment for any new technology, so it is difficult to predict which technology might displace CMOS.
HPCwire: Outside of the professional sphere, what can you tell us about yourself – personal life, family, background, hobbies, etc.? Is there anything about you your colleagues might be surprised to learn?
Outside of work, I enjoy getting outdoors with my family, traveling, and being active. Over the holidays, we spent a few days in Costa Rica. Aside from seeing where most of my Starbucks coffee is grown, we saw some spectacular birds and wildlife. It was nice to be offline for a while!
| Guangwen Yang