Didas Turatsinze, Michael Okopa and Tonny Bulega
A fundamental problem faced by any cloud service provider is how to maximize their revenues by allocating re-sources dynamically among the service instances and providing differentiated performance levels. Previous pricing mechanisms have been based on Mean Response Time (MRT) and Instant Response Time (IRT). However, mean response time tends to be representative of the performance of just a few big requests since they count the most in the mean because their response times tend to be highest. In this study, we propose two customer-oriented pricing mechanisms Mean Slowdown (MS) and Instant Slowdown (IS), in which the customers are charged according to achieved service performance in terms of mean slowdown. Analytical models of pricing mechanisms are developed for cloud computing under FCFS and PS scheduling policies. The models are then used to compare the performance of First Come First Served (FCFS) and Processor Sharing (PS) scheduling policies in terms of revenue generated. It is also observed that pricing mechanism based on Slowdown generates more revenue for the service provider than pricing mechanism based on response time. We also observe that revenue generated increases with increase in the number of servers, and arrival rate regardless of the pricing mechanism and scheduling policy used. We further observe that revenue generated in terms of MRT and MS is higher under FCFS policy than under PS policy for lower number of servers, however as the number of servers increase, PS policy outperforms FCFS policy in terms of generating more revenue.
