A research group with LSU’s Center for Computation & Technology (CCT), has received two awards to provide fundamental technical contributions to the recently announced Defense Advanced Research Projects Agency (DARPA) Ubiquitous High Performance Computing Program (UHPC). This program brings together researchers and scientists from universities, industry, and national laboratories to develop new system architecture and software to prototype next-generation supercomputers. The first models will be completed prior to 2018.
LSU Department of Computer Science Arnaud & Edwards Professor Thomas Sterling and his research group at the CCT, where Sterling has a joint appointment, will lead LSU’s contributions to this project, which include execution models, runtime system software, memory system architecture, and symbolic applications. Under this project, Sterling received $1.2 million from DARPA for four years of work.
LSU is a key academic partner of the X-caliber project, a lead component of the UHPC program led by Sandia National Laboratories to develop advanced computing systems capable of running complex calculations a thousand times faster, using orders of magnitude more processors simultaneously, than today’s most powerful supercomputers before the end of this decade.
National science and engineering organizations have identified developing new supercomputing systems as a top research priority because today’s largest and most powerful supercomputers will be too slow and obsolete to run the kinds of applications needed to use computational science efficiently for military defense and other areas.
“Such systems will enable applications that require months of computation time today to take hours by the end of the decade and make possible real-time applications that cannot be done at all now,” Sterling said.
Sterling has spent the past several years working to address problems in developing Exascale-class supercomputer systems, which would be capable of running a million trillion calculations per second through the simultaneous application of up to a billion processor cores. Currently, the world’s fastest supercomputers are at the Petascale level, running only 200,000 to 300,000 cores cooperatively.
“We are entering the next phase of computing and, essentially, everything has to change,” Sterling said. “We are honored and excited to be selected as part of the national UHPC initiative that will, in the coming years, establish the future direction of American research, development, and application of supercomputer technology.”
Sterling, who in the 1990s collaborated on the Beowulf computing cluster with Penguin Chief Technology Officer Donald Becker, directs the CCT Systems Science and Engineering Focus Area. This group has spent the past four years working on the ParalleX project to devise and test the validity and potential of the ParalleX model of computation.
Parallel computing exploits the concurrency of operation to shorten the amount of time it takes researchers to run applications. Future supercomputers using multicore and accelerator technologies will require dramatic increases in computational parallelism if Exascale performance is to be achieved by the end of this decade. ParalleX replaces conventional practices such as MPI and OpenMP programming models to deliver the needed parallelism that is capable of running on Exascale-class machines combined with dynamic, adaptive resource management for increased efficiency, reliability, and programmability.
The experimental ParalleX execution model is a unique synthesis of several complementing research concepts, some first explored over previous decades, combined with important new ideas of dynamic computation composition and parallel control. ParalleX is embodied by the proof-of-concept HPX runtime system CCT researchers developed, which currently runs on conventional distributed-memory clusters and shared-memory symmetric multi-processors. ParalleX was developed through sponsorship from the National Science Foundation, Department of Defense, DARPA, and Department of Energy, along with major contributions of resources and expertise by CCT.
“Exascale computing is at the forefront of the national science agenda, and it is exciting that research taking place here is establishing LSU as a leader in the next level of computational science research,” Sterling said.