The NSF-sponsored workshop on 'Future Directions of Computer System Research in 2010' is to be held in the Waterview Conference Center, Arlington, VA over the two days of March 25-26, 2010.
The outcome of this workshop is a report that outlines new research directions in the general area of Computer Systems. This report will become available on the web within two weeks after the conclusion of the workshop, and will be used later as an inspiration in writing the new solicitation of the CSR program.
" was held in the Waterview Conference Center, Arlington, VA over the two days of March 25-26, 2010. The outcome of this workshop was a report that outlines new research directions in the general area of Computer Systems. Today's computing systems and applications consume more computingpower, distributed over more interconnected machines, running on more efficient and powerful devices, with more varied capabilities, and involving more data than ever before. Computing systems have truly become ubiquitous, and so has systems programming. The result is unparalleled opportunity for innovation. Small programmable personal devices can interact with deep, networked data repositories to improve our health and keep us connected. Enabling technologies like sharedcomputing infrastructure can help make IT sustainable. But, such growth has been largely a patchwork, and we are fast reaching limits to its scope: challenges imposed by limits in energy scaling, increasing heterogeneity in system components, and uncertainty in the environment in which systems operate. Together, these concerns have brought computing systems to an inflection point, where either an incredible proliferation or intractable complexity are possible. The key approaches highlighted in the report for coping with these challenges are: systems that plan for diversity of constituentcomponents as a guiding principle; evolve and improve, not just react; expose information across abstraction boundaries while preserving abstraction's benefits; are reasoned compositions of components with some formally-provable properties; and consider humans in the loop as a first-order concern. These approaches to design and construction have reflections throughout the systems stack, from hardware and architecture, programming languages, operating systems, distributed systems, and in the domains of pervasive, ubiquitous, hybrid, and embedded systems. Applying these approaches in a variety of technical challenges can cope with the inflection points endemic to coming limits on energy reduction, heterogeneity at scale,and uncertainty in components and the environment.
