Heterogeneous Processors and Hybrid Systems for Big Data and Analytics

▣ Title : Heterogeneous Processors and Hybrid Systems for Big Data and Analytics

▣ Speaker

: Dr.

Peter Hofstee (IBM)

▣ Date

& Time : Tuesday, May 21 (3:30 ~ 4:45pm)

▣ Place

: LG Research Building, Room #105

▣ Host

: Prof. Sungjoo Yoo (Tel. 2379)

▣ Abstract : Technology constraints increasingly allow heterogeneous processors, that combine multiple types of compute cores sharing memory, and hybrid systems, where different types of processors are combined at a system level, to gain a market advantage in spite of the challenges associated with programming such systems. This talk will briefly review the experience gained with the Cell Broadband Engine processor, including some of its applications in high-performance computing and embedded systems. Next the role of heterogeneity in commercial systems is discussed, and a hardware and software approach is discussed that allows for the integration of reconfigurable accelerators in high-performance and commercial systems. We discuss how these technologies can help us store and analyze the vast amounts of digital data of various types that are collected and the new levels of insight, foresight, and optimization that can be achieved.