An Online Optimization-Based Technique for Dynamic Resource Allocation in GPS Servers

Since web workloads are known to vary dynamically with time, in this paper, the authors argue that dynamic resource allocation techniques are necessary in the presence of such workloads to provide guarantees to web applications running on shared data centers. To address this issue, they present an analytic model of a server resource that services multiple applications using Generalized Processor Sharing (GPS). Using this model, they present online workload prediction and optimization-based techniques to dynamically allocate resources to competing web applications running on shared servers.