A Survey of Computation Offloading for Mobile Systems

Motivation:

Mobile systems have limited resources, such as battery life, network bandwidth,storage capacity, and processor performance. These restrictions may be alleviated by computation offloading. The main reasons of choosing this solution but not traditional solutions are: First, it could migrate programs and computation to servers outside of the users’ immediate computing environment. This could reduce the overhead on mobile system. Second, it could augment mobile systems’ capabilities by using more sources which come from resourceful serves. Third, it typically used to augment the computational capability of a resource constrained device for a single user.

Main Points:

• With the time goes, research of computation offloading focuses on from feasibility, importance to decision, then to infrastructures.

            The paper discuss two significant enablers for offloading:
            (1). Building wireless data networks (in particular WiFi) to facilitate mobility.
                  Mobile agents are autonomous programs that can control their movement from machine to machine in a    heterogeneous network. The author gives Agent TCL and Concordia as examples to explain technologies focus on migrating computation for mobile devices, network connectivity, and leveraging Java for developing platform independent applications.
            (2). Virtualization is a way to logically partition large mainframe computers into smaller, independent computing units. This enabled multitasking: the ability to run multiple applications and processes at the same time.
                  Cloud computing uses virtualization to offer computing as a service; users can “lease” computing resources based on their requirements.

• The paper focuses on the research of making a decision regarding whether and what computation to migrate. It contains two purposes for offloading:

            (1). Through the equation, if offloading wants to improve performance then it should satisfy one or more of these conditions: the program requires heavy computation; the server is fast; a small amount of data is exchanged or the bandwidth is high.
            (2).  Through the equation, if offloading wants to save energy then it should satisfy one or more of these conditions: heavy computation or light communication; also in special case, the bandwidth can be substantially higher.

• The author gives the comparison of static and dynamic decisions. When the decision is static, the program is partitioned during development. Static partition has the advantage of low overload during execution. This approach is valid only when the parameters can be accurately predicted in advance. In contrast, dynamic decisions can adapt to different run-time conditions and also use prediction mechanisms.

• The author explains graph partition approach in details to identify the offload-able parts of a program.

• The paper lists some issues which infrastructures deal with to make offloading practically adopted. Including Inter-operability, mobility and fault tolerance, privacy and security, and context awareness.

• The author combines the trend of this world to predict the development of offloading in the future.

Trade-offs:

• This survey is methodic. First the author tells the motivation. He states the researches in the past. Based on the foundation of the theories, the author explain the useful algorithms to improve performance of offloading. With the time goes, the author predict the development and importance of offloading in the future. From the beginning to last, the readers would be convinced by the author’s powerful logic and facets.

• This is a survey paper. Due to the limited knowledge of authors and the sources of research only just come from IEEE paper pool. So maybe some important papers and research are not in this survey. We should consider this paper discriminately.

• The algorithm of offloading decision is in the condition of wireless network. So, if the conditions change suddenly, such as in some special situation, people only have 3G. The the execute time of changing data would be high. The inequality of decision made in the paper would not be fitted. Then we could not consider the serves. We could consider the usage of local mobiles and the optimal algorithm in own mobile platform. So if author could give some ways of dealing with this condition would be more convincing.

• In the offloading in the future part, the author only consider the problems in the future. I suggest that he could add some superiority. Such as if we could use satellite to send the signal to almost around the world or if people could use WiFi everywhere in anytime. Then what we should do if the condition is so convenient.