Analysis and Optimization of Overflow Loss Systems for Cost-Effective Operations of Video Server Farms, Call Centers, and Intensive Care Units

Overflow loss systems are an important teletraffic model originally arising in telecommunication systems. Recently, they are applied in service sectors to model video server farms, call centers, intensive care units, and lead to challenging management science problems for cost-effective operations of such systems subject to stringent quality of service requirements. They are characterized by customers requiring immediate service in a system composed of multiple stations. An arriving customer either is admitted to a station with at least one idle server or “overflows” to another station. If all stations accessible to the customer are busy, the customer is blocked. It is well known in teletraffic theory that analysis of overflow loss systems is difficult. Such models involving significant dependencies are not amenable to an exact analysis, and finding accurate and scalable approximations is challenging. This project will contribute to the development of a robust approximation framework for analysis of overflow loss systems. The pilot study will enable us to develop optimal policies for resource management in video server farms, staff configuration in call centers, and capacity planning for intensive care units. Our research in these important areas will have long-term social and economic impact on the largely service-oriented Hong Kong economy.