The goals of the workshop are to discuss research directions and challenges in power, energy, and temperature management for high-performance computing (HPC) systems; identify opportunities for collaborations between some of the key research groups in this area; and develop a research agenda for the future.
The workshop will involve a keynote address, two invited talks, two panels, and a final open discussion. The presenters will discuss architectures, systems, applications, and algorithms designed for power, energy, and temperature management, as well as current and upcoming national initiatives and funding opportunities.
The growing scale of HPC systems and data centers has made power, energy, and temperature become critical issues for these systems. High-end HPC systems today consume several megawatts of power, enough to power small towns. The cost of powering these systems is in the millions of dollars per year, and is increasing fast as they target Petascale and eventually Exascale performance.
At the same time, new hardware and software technologies are creating exciting opportunities for power, energy, and temperature management in these systems. As a few examples of these technologies, designers have been introducing a variety of lower power modes into hardware components; advances in virtualization are enabling the efficient consolidation of data- and compute-intensive applications to increase efficiency; and researchers have been introducing new energy-aware algorithms for many existing problems. It is incumbent upon academia and industry to explore and exploit these technologies and opportunities to the fullest extent. Moreover, it is time for us to devise the next wave of technologies that can facilitate power, energy, and temperature management further.