Dynamic Scheduling of a Multi-Class Fluid Model with Transient Overload under QoS Contracts
We study the scheduling of a multi class single server system in a non-stationary setting, where the server is overloaded for a period of time due to the increase in customer arrival rates. As a result of the quality of service contracts, this overload situation incurs a penalty cost. The QoS level depends on the customer type and the customer arrival rates vary in a cyclic pattern. In each cycle, there is a high period, when the server is overloaded,
i.e. the total traffic intensity is bigger than 1, and a low period when the total traffic intensity is less than 1. We study the fluid model of this system under a special type of QoS contract and provide thes cheduling policy that minimizes the total cost in a cycle.