Performability analysis of cloudlet in mobile cloud computing

Abstract

In the Cloudlet architecture of Mobile Cloud Computing (MCC), a mobile device, in order to overcome its lack of resources, offloads resource-intensive tasks on a nearby and self-managed cloud-based data center (i.e., cloudlet). In this architecture, a dedicated Virtual Machine (VM) is provisioned for the mobile device as the VM can be located as a part of the cloudlet or a public cloud. The features of this new technology, including mobility, the instability of wireless connections, and the complexity of virtualization technology, make the prediction of the performance and availability of services challenging. To deal with these issues, the present paper proposes a combined performance and availability (i.e., performability) model using the Stochastic Reward Net (SRN). However, to diminish the complexity of the proposed model for large systems, the model is firstly simplified by disregarding some delays. Then, the simplified model is appropriately separated into two interacting SRN sub-models. The cyclic inter-dependency among these sub-models is also resolved by the fixed-point iteration method. These models illustrate the impact of various sets of parameters, such as the failure/repair of VMs, on two important measures: (1) request rejection probability and (2) availability. To obtain numerical results, all the proposed SRNs are solved by the SPNP software package. To validate and verify the proposed models, discrete-event simulation results are also presented.