Fluid Limits and Optimal Task Assignment Policies for Locally Pooled Service Systems

Task assignment policies play a central role in many online applications, where service requests or tasks arrive over time and are distributed across parallel servers in a data center or cloud computing platform. The way in which the tasks are distributed across the servers has a tremendous impact on the performance perceived by users and the efficiency of server usage, which has attracted strong interest from our research community. Most of this interest has been directed to the so-called supermarket model, but real-life systems have features that fall outside of this standard framework, raising significant challenges which have been pursued in my doctoral thesis [3]. In particular, my thesis provides novel modeling frameworks and performance analysis techniques that enable the study of interactive online applications with fluctuating demand patterns and networked systems.