Performance enhancement framework for cloudlet in mobile cloud computing / Md Whaiduzzaman

Md Whaiduzzaman, Abdullah (2016) Performance enhancement framework for cloudlet in mobile cloud computing / Md Whaiduzzaman. PhD thesis, University of Malaya.

Preview

PDF (Thesis PhD) Download (26Mb) | Preview

Abstract

A tremendous increase in the use of mobile devices, such as smart phones and tablets, has been observed in recent years. Mobile device resources, such as CPU, memory and storage resources, have also experienced dramatic increases in capacity. Simultaneously, a rich variety of mobile applications that require extensive computational resources for mobile application execution have been developed. These new applications are complemented with cloud resources through the emerging Mobile Cloud Computing (MCC) paradigm. MCC augments mobile device capabilities by leveraging the resources of distant clouds, nearby cloudlets or mobile ad-hoc clouds in the vicinity. However, practical utilization of MCC is hindered by limitations associated with network connectivity. In the case of distant clouds, jitter, bandwidth, and propagation delay pose a challenge to realtime application response. On the other hand, cloudlets and mobile ad-hoc clouds are not sufficiently resource-rich to be able to support rich mobile applications. In this research, we consider a mobile user in the vicinity of a cloudlet that is situated at a distance of one hop through aWi-Fi communication medium. We show that a substantial number of users accessing the cloudlet for computation-intensive tasks results in delayed task completion and ultimately diminishes the benefits of using cloudlets. The problem is referred to as the cloudlet resource scarcity problem. To alleviate this problem, various researchers have offered solutions whereby mobile device resources are used for partial task completion. However, the proposed approaches do not consider the mobile device offered-serviceto- load ratio. In this research, we propose the Mobile-device-based Cloudlet Resource Enhancement (MobiCoRE) framework for mobile application augmentation to employ nearby mobile devices while ensuring the following: (i) mobile devices always obtain time benefits for their tasks when submitted to the cloudlet, and (ii) the cloudlet-induced mobile device load is a fraction of its own service requirements. We map MobiCoRE on iii the M/M/c/K queue and model the system using a birth-death Markov chain. We show that for cloudlets, the framework always obtains the maximum advantage for mobile devices in terms of job completion time when the cloudlet service time is set to ¯ c l , where ¯ c is the cloudlet utilization and l is the application arrival rate. Furthermore, the optimal service time is independent of the application’s service requirement. We implemented the MobiCoRE framework using the Openstack cloud. The evaluation shows that MobiCoRE accommodates up to 50% more users when operating at the optimal service time and provides 50% time benefits for mobile users. The empirical analysis and statistical validation demonstrate that our proposed framework, i.e., MobiCoRE, significantly and positively impacts the cloudlet performance by exploiting and orchestrating nearby mobile device resources.

Actions (For repository staff only : Login required)