Performance evaluation of wide-spread assignment schemes in a vehicular cloud

A parking lot based vehicular cloud is a cloud system where the resources are parked vehicles with onboard computing units. The dynamic and random process of vehicles joining and leaving a vehicular cloud results in a randomly changing number of available resources. This dynamic nature results in tasks assigned to departing resources to fail which reduces system efficiency. In this paper, we introduce a vehicular cloud model that considers multiple task request types and multiple resource capability types. We studied the performance of each type of user task request in tour model when using classical assignment policies, i.e. First Fit, Best Fit, and Last Fit. We also propose and investigate two techniques, task handoff and queueing, to increase the efficiency of the system and reduce the failure rate in the cloud. System simulation results show that the Last-Fit technique having the lowest task completion failure, while Best-Fit having the lowest task assignment failure. Simulations also show that hand off and queueing are very effective in increasing efficiency and reducing failure rates.