This shows you the differences between two versions of the page.

Link to this comparison view

seminars:seminar_02_28_14 [2017/09/20 22:02] (current)
Line 1: Line 1:
 +====== Big Data and Dark Silicon: Taming Two IT Inflection Points on a Collision Course ======
 +==== Babak Falsafi (EPFL) ====
 +== Tuesday, Feb. 28th, 11:​30am-1:​00pm ==
 +== Hamerschlag Hall D-210 ==
 +Information technology is now an indispensable pillar of a modern-day
 +society, thanks to the proliferation of digital platforms in the past
 +several decades. We are now witnessing two inflection points, however, that
 +are about to change IT as we know it. First, we are entering the Big Data
 +era where demand on robust and economical data processing, communication and
 +storage is growing faster than technology can sustain. Second, while
 +forecasts indicate that chip density scaling will continue for another
 +decade, the diminishing returns in supply voltage scaling and the impending
 +"​energy wall", is leading server designers towards energy-centric solutions
 +and eventually Dark Silicon. In this talk, I will motivate these two IT
 +trends and present promising research avenues for server chip design
 +including specialized scale-out processors, on-chip networks and die-stacked
 +DRAM cache hierarchies.
 +Babak is a Professor in the School of Computer and Communication Sciences and
 +the founding director of the EcoCloud research center pioneering future
 +energy-efficient and environmentally-friendly cloud technologies at EPFL. He
 +has made numerous contributions to computer system design and evaluation
 +including a scalable multiprocessor architecture which was prototyped by Sun
 +Microsystems (now Oracle), snoop filters and temporal stream prefetchers
 +that are incorporated into IBM BlueGene/P and BlueGene/Q, and computer
 +system simulation sampling methodologies that have been in use by AMD and HP
 +for research and product development. His most notable contribution has been
 +to be first to show that contrary to conventional wisdom, multiprocessor
 +memory programming models -- known as memory consistency models -- prevalent
 +in all modern systems are neither necessary nor sufficient to achieve high
 +performance. He is a recipient of an NSF CAREER award, IBM Faculty
 +Partnership Awards, and an Alfred P. Sloan Research Fellowship. He is a
 +fellow of IEEE.