Mechanisms for Optimally Scheduling and Pricing Pleasingly Parallel Jobs in Service Clouds

Recently, more and more users begin to outsource their job executions to service clouds, which brings benefits for both users and providers. Unfortunately, the fixed-price models which are commonly used by most clouds have several shortcomings, such as requesting users to specify and monitor their resources as well as being difficult to set prices. These shortcomings limit the usage of fixed-price models for service clouds and, thus, service providers with finite resource urgently need effective approaches to schedule and price user’s jobs, with the goal of social welfare maximization. In response to the need of service providers, this paper designs new auction mechanisms for service clouds, with unique features of job-oriented users, pleasingly parallel jobs, and soft deadline constraints. However, several challenges should be addressed when designing mechanisms, such as the NP-hardness of finding the optimal job scheduling and possible misreports of selfish users for private information. To deal with these challenges, we first propose a new randomized scheduling mechanism for optimally scheduling and pricing pleasingly parallel jobs in service clouds. This mechanism is truthful in expectation, while achieving $\alpha $ -approximation to the social welfare. However, potential collusion among cloud users which may result in significant effects has been ignored by this mechanism. To handle the collusion problem, we further propose a collusion-resistant mechanism which achieves $(t,P)$ -truthful while scheduling and pricing jobs. Both of these two mechanisms are computationally efficient and individually rational, and they can schedule jobs in a way without preemption. Finally, the theoretical analysis and extensive simulations based on synthetic data and real-world job traces validate the effectiveness of our mechanism.