Towards Energy-Centric Computing and Computer Architecture


Formale Metadaten

Titel Towards Energy-Centric Computing and Computer Architecture
Herausgeber European Organization for Nuclear Research (CERN)
Erscheinungsjahr 2011
Sprache Englisch

Inhaltliche Metadaten

Fachgebiet Physik
Abstract Technology forecasts indicate that device scaling will continue well into the next decade.  Unfortunately, it is becoming extremely difficult to harness this increase in the number of transistors into performance due to a number of technological, circuit, architectural, methodological and  programming challenges.In this talk, I will argue that the key emerging showstopper is power.  Voltage scaling as a means to maintain a constant power envelope with an increase in transistor  numbers is hitting diminishing returns. As such, to continue riding the Moore's law we need to look  for drastic measures to cut power. This is definitely the case for server chips in future datacenters, where abundant server parallelism, redundancy and 3D chip integration are likely to remove  programming, reliability and bandwidth hurdles, leaving power as the only true limiter.I will present  results backing this argument based on validated models for future server chips and parameters  extracted from real commercial workloads. Then I use these results to project future research  directions for datacenter hardware and software.About the speakerBabak Falsafi is a Professor in the School of Computer and Communication Sciences at EPFL, and an  Adjunct Professor of Electrical and Computer Engineering and Computer Science at Carnegie Mellon.  He is the founder and the director of  the Parallel Systems Architecture Laboratory (PARSA) at EPFL where he conducts research on architectural support for parallel programming, resilient systems, architectures to break the memory  wall, and analytic and simulation tools for computer system performance evaluation.In 1999, in  collaboration with T. N. Vijaykumar he showed for the first time that, contrary to conventional wisdom, multiprocessors do not need relaxed memory consistency models (and the resulting convoluted programming interfaces found and used in modern systems) to achieve high performance. He is a recipient of an NSF CAREER  award in 2000, IBM Faculty Partnership Awards between 2001 and 2004, and an Alfred P. Sloan  Research Fellowship in 2004. He is a senior member of IEEE and ACM. 

Ähnliche Filme