Harmonic placement: File system support for scalable streaming of layer encoded object

Abstract

From file system's point of view, scalable streaming introduces another dimension of complexity in disk scheduling. This is particularly because when a subset of information is retrieved, the playback does not necessarily coincides with the sequential access. We propose a new file organization technique called Harmonic placement. The basic idea is to cluster the frequently accessed layers together. We develop elaborate analytical models for different file organization techniques: Progressive placement, Interleaved placement and Harmonic placement. We investigate the performance of file organization techniques under varying workload conditions. The models developed in this work enables us to predict the performance and efficiency of the storage system in scalable streaming environment. We find that the Harmonic placement outperforms other schemes in scalable streaming environment.