Temporal Importance Abstraction for Automated Storage Reclamation

This work focuses on scenarios that require the continual storage of large amounts of data. Such systems require the ability to either continually increase the total storage space or reclaim existing contents. Traditional storage systems provide persistence for all stored objects to perpetuity, relegating object reclamation to applications. Our key insight is that not all of the new data are equally important. Exporting this knowledge of the object importance from the application to the storage system could potentially allow the storage system to extend the service range by selectively reducing service levels. In this work, content creators explicitly annotate the object using a temporal importance function. This information allows the storage system to selectively weaken the persistence guarantees offered by the storage and evict less important contents. We also introduce the notion of storage importance density to quantify the importance levels for which the storage is full, allowing the content creators to choose the appropriate importance levels for their objects.