BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20101118T181500Z
DTEND:20101118T191500Z
LOCATION:384-385
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: The main technological challenge for Exa-scale computing is power consumption. Accelerators such as modern GPUs now offer attractive power/watt performances and can thus already be seen as part of the current top-of-the-line supercomputers (www.top500.org). At the same time, modern embedded processors such as the ARM architecture found in many mobile devices, offer even more compute power per watt and have recently also started to include floating point operations. Interconnecting these devices may hence be the technology that can be pushed beyond Exa-scale. However, there are many challenges to overcome for this to happen.=0A=0ACombing these technologies with Cloud computing, offers an interesting avenue. Cloud computing is a scalable services delivery platform that allows users to access resources in a flexible way, making the infrastructure adapt to the application and not vice versa. This is particularly attractive from an energy point of view.=0A=0AThe first objective of this BoF Session is to set up a meeting point for researchers and developers in these Computing Paradigms in order to share knowledge and best practices. The second (but not secondary) objective is to find a beneficial convergence that could lead to a roadmap for a flexible Exa-scale HPC infrastructure.=0A=0ATopics for this Session include (but are not limited to):=0A=0A- Power efficient HPC using future GreenIT technologies=0A=0A- Impact of GPU Computing in the Cloud Computing business model=0A=0A- Virtualization technologies in GP-GPU based clouds=0A=0A- Evolution of mobile and embedded devices for their integration in Computing Clouds=0A=0A- Common heuristics and execution models of heterogeneous applications
SUMMARY:Towards Exa-Scale: Heterogeneous Clouds (CPUs, GPUs and Embedded Devices)
PRIORITY:3
END:VEVENT
END:VCALENDAR