Evaluating the Usefulness of Content Addressable Storage

Content Addressable Storage (CAS) is increasingly being used as a technique to provide for space savings when storing datasets. In this paper we analyze the performance of CAS on real-world applications and discuss the effects on space savings, savings in network bandwidth and on resultant error resilience of data. We find that a chunksize of 1 KB can provide up to 84% space savings and even higher savings in network bandwidth whilst trading off error resilience. We also find that a chunksize of 1 KB or larger provides good file-system performance while incurring about 14% overhead from use of a CAS based naming scheme.