Smart offloading Algorithms and Profiling Techniques for Cloud Computing, PI: Hyesoon Kim (May 2012-April 2013)

Hyesoon Kim, COC

 

Georgia Tech PI: Hyesoon Kim, COC
Co-PI: Mayur Naik, Santosh Pande
Students: Dilan Manatunga, Pranith Kumar

Project Goals

As the computing capability in mobile platforms becomes powerful, more and more users are using mobile devices as one of the main computing platforms. To provide a more powerful and comfortable environment, using both the server computing power and client computing power is a demanding need. However, what to off-load and when are still challenging questions. Furthermore, because the latency of networks varies significantly, depending on network traffic, it is not easy to make these decisions at static time.


Hence, in this project, we study the characteristics of applications and provide an understanding of the trade-offs between local computation and cloud computation.

The foci of this project are

  1. to understand what kind of applications or what kind of scenario would enjoy the benefits of off loading;
  2. to profile techniques to estimate the trade-offs between execution time savings and network latency;

Figure 1

Scope of the work

  1. Developing cloud computing version of vide editor
  2. Developing off-line profiler techniques to identify off-load candidates