A model for power efficiency of mobile devices through lightweight method level computational offloading
Mobile devices have become an integral part of our daily lives. However, the restricted battery timing curtails longer operational hours. To tackle the limited battery timing issue, a technique, computational offloading is used. In computational offloading, the intensive tasks are offloaded from mob...
| Summary: | Mobile devices have become an integral part of our daily lives. However, the restricted battery timing curtails longer operational hours. To tackle the limited battery timing issue, a technique, computational offloading is used. In computational offloading, the intensive tasks are offloaded from mobile devices to remote server in order to execute the task remotely and save battery life. Computational offloading frameworks/models based on VM migration, whole application migration, or traditional method level offloading are resources intensive and time consuming. The dynamic partitioning of application, execution of task at cloud server, service call by Simple Object Access Protocol (SOAP) and no defined mechanism for predefined parameters, make the previous method level computational frameworks/models inefficient for energy saving. In order to address the inefficiencies of previous method level computational offloading frameworks/models, a lightweight method level computational offloading model is proposed. Four distinct components are deployed in the proposed model which eliminates the shortcomings of previously developed frameworks/models. A Representational State Transfer (REST) based technique developed for calling the remote services which is based on JSON instead of XML, and hence is lightweight. REST also reduces the size of communication data at approximately 100% as compared to SAOP service call. Surrogate server is configured at a single hop distance which reduces the RTT and ultimately reduces the power consumption. The application is partitioned at method level by a novel dynamic technique in source code, which counters the inefficiencies of existing partitioning techniques. A mechanism for selection of predefined parameters is defined. These parameters are important to consider before each offload. The predefined parameters consist of battery level, network type, and execution time which affirms the energy saving during offloading. The proposed framework is implemented in the real mobile cloud computing environment. Execution time and energy consumption of both local execution and traditional offloading are benchmarked in order to investigate and validate the performance of the proposed lightweight method level model. The prototype is developed with three components which are REST-Offload, Local Execution and Traditional- Offload and then tested in real mobile cloud environment for Execution Time and Energy Consumption. The result of this research indicates that the proposed solution diminishes resources utilization. The REST-Offload is significantly useful compared to both Local Execution and Traditional Offloading methods. It reduces about 50% Execution Time and approximately 38% Energy Consumption. |
|---|