09:00 - 10:00 Keynote

Mobile Cloud Computing: The Beginning of the End? Abstract

Dr. Baochun Li (Bell Canada Endowed Chair in Computer Engineering, Professor, Department of Electrical and Computer Engineering, University of Toronto) Bio


Mobile cloud computing has received a substantial amount of academic research attention and interests in the past three years, represented by not only the number of papers published, but also by the advent of dedicated workshops (such as the MCC workshops at SIGCOMM since last year).  In essence, mobile cloud computing is about remedying the limited resource availability on mobile devices with the resource abundance in the cloud.  Since the most important resources on mobile devices are computing cycles and battery energy, existing works in the literature are primarily concerned with offloading computing cycles to the cloud, in order to improve performance of computational-intensive applications, or to reduce energy consumption.  Such computational offloading is performed at the granularity of a method or a thread in mobile applications, as represented by recent papers such as MAUI (MobiSys 2010), CloneCloud (EuroSys 2011), and COMET (OSDI 2012).

However, it is unfortunate that every time computation is offloaded from a mobile device to the cloud, we have to transmit the application states at runtime over the network, with the potential risk of consuming even more energy than performing the same computation locally on the mobile device.  From this perspective, such computational offloading only makes practical sense if the performance gain is worth the energy cost of transmitting data over the network.  Since the questions of how and when computation should be offloaded have been thoroughly answered by existing works in the literature, one would naturally wonder (with a touch of pessimism) whether we have entered the beginning of the end of research on mobile cloud computing.

In this talk, we advocate that mobile cloud computing should not be limited to offloading computing cycles to the cloud, but should also be concerned with the use of the cloud to assist interactive and delay-sensitive applications.  As two examples, we briefly introduce our recent work on streaming gestures among users in gesture-intensive interactive applications, and on the use of inter-datacenter networks in the cloud to improve the performance of multi-party video conferencing.  In fact, a recent public message from the chief architect of Skype has explained how mobile devices have accelerated the conversion of Skype from a peer-to-peer architecture to a cloud-assisted design.  Regardless of how practical computational offloading is, we believe that mobile applications will forever be tightly integrated with the cloud, which is what mobile cloud computing is all about.