BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:ThinkAir: Dynamic resource allocation and parallel execution in cl
 oud for mobile code offloading - Andrius Aucinas (University of Cambridge)
DTSTART:20120322T160000Z
DTEND:20120322T163000Z
UID:TALK36700@talks.cam.ac.uk
CONTACT:Eiko Yoneki
DESCRIPTION:Smartphones have exploded in popularity in recent years\,\nbec
 oming ever more sophisticated and capable. As a result\,\ndevelopers world
 wide are building increasingly complex applications\nthat require ever inc
 reasing amounts of computational\npower and energy. In this paper we propo
 se ThinkAir\, a\nframework that makes it simple for developers to migrate\
 ntheir smartphone applications to the cloud. ThinkAir exploits\nthe concep
 t of smartphone virtualization in the cloud and\nprovides method-level com
 putation offloading. Advancing on\nprevious work\, it focuses on the elast
 icity and scalability of the\ncloud and enhances the power of mobile cloud
  computing by\nparallelizing method execution using multiple virtual machi
 ne\n(VM) images. We implement ThinkAir and evaluate it with a\nrange of be
 nchmarks starting from simple micro-benchmarks\nto more complex applicatio
 ns. First\, we show that the execution\ntime and energy consumption decrea
 se two orders of\nmagnitude for a N-queens puzzle application and one orde
 r of\nmagnitude for a face detection and a virus scan application. We\nthe
 n show that a parallelizable application can invoke multiple\nVMs to execu
 te in the cloud in a seamless and on-demand\nmanner such as to achieve gre
 ater reduction on execution\ntime and energy consumption. We finally use a
  memoryhungry\nimage combiner tool to demonstrate that applications\ncan d
 ynamically request VMs with more computational power\nin order to meet the
 ir computational requirements.
LOCATION:FW26\, Computer Laboratory\, William Gates Builiding
END:VEVENT
END:VCALENDAR