Towards Modeling Computation Capacity of a Vehicular Cloud While Overcoming Resource Volatility

Future vehicles will become computationally more powerful to be safer, more autonomous, and more convenient for passengers. However, the computing resources onboard the vehicles will often be underutilized and can be pooled to form a computing cluster called vehicular cloud (VC). The VC operators inherently would like to know the capabilities of the VC to predict its performance adequately. Since the vehicles are mobile, their residency time in the VC will be random. As a result, resources in the VC will be volatile. In this work, we analyze computing capacity of a VC while overcoming its volatility characteristics. We assume that computing jobs consist of random number of tasks that can be executed independently. A job is completed when execution of all its tasks are completed. We employ a service strategy that assigns each task to a single vehicle. Further, a task is assigned to a vehicle only if the vehicle can complete its execution during the vehicle's residency time. Using a stochastic modeling approach, we provide a tractable solution to the distribution of the number of completed jobs during the lifetime of a VC, which often can not be obtained through other approaches. Then we employ the Monte Carlo simulation method to verify the numerical results from the analytical model.