To overprovision or to share via QoS-aware resource management?

With the rapid progress in processor speeds and large network bandwidth availability, it is legitimate to ask the following question: Do we still need quality of service (QoS) mechanisms and policies for shared high-performance distributed computing when timely guarantees are required for their applications? Many researchers believe that through overprovisioning each application will get its QoS without any specific mechanisms and policies. This paper argues that even with overprovisioning we do not achieve always QoS guarantees, especially in the timing domain. By discussing and analyzing currently evolving QoS-aware resource management in a shared computing and communication environment, we show the necessity of QoS provision and its role in high-performance distributed computing environments.