An economic approach to adaptive resource management

Resource management is a fundamental concept in operating system design. In recent years it has become fashionable to consider the problem as an aspect of heterogeneous support for quality of service ('QoS'). The desire for QoS support leads to the dual management goals of global (system) and local (application) optimisation. We propose an architecture based on an economic model of resource management using frequently renegotiated timed resource contracts. We use dynamic pricing as a congestion feedback mechanism to enable applications to make system policy controlled adaptation decisions. We argue that this scheme has many advantages over a traditional central resource management entity, including scalability and application specific adaptation.