This paper considers the problem of allocating discrete resources according to utility functions reported by potential recipients and relates this abstract problem to resource allocation in a Utility Data Center (UDC). A simple integer program formulation, which generalizes well-known knapsack problems, permits a remarkable breadth of expression while retaining clarity and analytic tractability. In the UDC context, this formulation allows us to incorporate factors such as resource scarcity, user demand, and operating costs in a unified framework. It is equally applicable to long- and short-term allocation. If applied to short-term dynamic re- allocation it allows SLA violation penalties to be enforced or relaxed, thereby permitting principled preemption of resources. Retrospective analysis of past allocator inputs can guide economically-optimal capacity expansion. The proposed problem formulation is suitable both for UDCs that operate exclusively within an enterprise and for those that sell access to computational resources to external customers. The latter case involves multiple divergent interests contending for scarce resources, and this paper surveys the economic issues that arise in such situations and relevant literature, e.g., on mechanism design and auction theory. This paper describes the expressive power of the proposed problem formulation, considers the computational requirements of solution methods, outlines connections with economics literature, and compares the proposed formulation with other UDC allocation schemes. It also presents preliminary computational results showing that a commercial integer-program solver can quickly find near-optimal solutions to random problem instances of reasonable size. Notes: To be published in the First Workshop on Algorithms and Architectures for Self- Managing Systems, 11 June 2003, San Diego, California 6 Pages
PDF
download
878987797
028-85220240
