Time-varying management of data storage

The efficiency of large-scale applications is strongly de-pendent on good data management techniques. In this pa-per, we claim that the ability to specify data requirements in a time-varying manner facilitates data management and im-proves application efficiency. This is because requirements such as availability, bandwidth and latency can vary sig-nificantly with time. Consequently, the storage system can dynamically change the allocation of resources to data ob-jects. We describe how an application may specify these dynamic requirements using utility functions, and outline a strategy towards achieving an optimal allocation of re-sources to data objects.