An application-aware data storage model

We describe a new application-controlled file persistence model in which applications select the desired stability from a range of persistence guarantees. This new abstraction extends conventional abstractions by allowing applications to specify a file's volatility and methods for automatic reconstruction in case of loss. The model allows applications, particularly ones with weak persistence requirements, to leverage the memory space of other machines to improve their performance. An automated (filename-matching) interface permits legacy applications to take advantage of the variable persistence guarantees without being modified. Our prototype implementation shows significant speed-ups, in some cases more than an order of magnitude over conventional network file systems such as NFS version 3.