A context-aware application offloading scheme for a mobile peer-to-peer environment

Abstract

Improved capabilities and the subsequent ability to run applications with increasing complexity transform smart phones in a valid alternative for users` next personal computer. However, smart phones are still far from desktops or laptops` capabilities, and fail to execute complex tasks.

The integration of smart phones into computing clusters, such as cloud or grids, is seen as one alternative to overcome this issue. However, this alternative is limited by several aspects, among them the lack of access to contextual information, the need of infrastructure and network connectivity and privacy considerations. Given the potential pervasiveness of smart phones in the near future, we propose a new paradigm in which smart phones play a crucial role as both providers and users of their resources. In this concept, smart phones in the vicinity can create a mobile peer-to-peer network and share their computational tasks to execute complex applications.

An approach used for sharing computational tasks among devices is based on application offloading. This approach works by selecting part of an application and delegating its execution to another device in order to overcome resource constraints - such as low CPU power, short battery life and limited memory.

A major challenge when applying such approach in a mobile peer-to-peer environment is how to reflect the intrinsic characteristics of mobile devices - such as limited resources, mobility and frequent disconnection - in the offloading scheme.

In this dissertation we present an application offloading scheme for mobile peer-to-peer environments, in which the intrinsic characteristics of mobile devices - such as limited resources, mobility and frequent disconnection - are reflected.

Evaluation results show that by considering contextual aspects, the proposed scheme presents a higher success ratio than existing approaches when applied to mobile peer-to-peer environments.