Kosha: A Peer-to-Peer Enhancement for the Network File System

Abstract

This paper presents Kosha, a peer-to-peer (p2p) enhancement for the widely-used Network File System (NFS). Kosha harvests redundant storage space on cluster nodes and user desktops to provide a reliable, shared file system that acts as a large storage with normal NFS semantics. P2p storage systems provide location transparency, mobility transparency, load balancing, and file replication -features that are not available in NFS. On the other hand, NFS provides hierarchical file organization, directory listings, and file permissions, which are missing from p2p storage systems. By blending the strengths of NFS and p2p storage systems, Kosha provides a low overhead storage solution. Our experiments show that compared to unmodified NFS, Kosha introduces a 4.1% fixed overhead and 1.5% additional overhead as nodes are increased from one to eight. For larger number of nodes, the additional overhead increases slowly. Kosha achieves load balancing in distributed directories, and guarantees 99.99% or better file availability.

Improving Data Availability for Better Access Performance: A Study on Caching Scientific Data on Distributed Desktop Workstations

“…Several software approaches to data availability have also been deployed, especially in distributed and peer-to-peer environments. In particular, replication is widely used and investigated [2,6,7,9,12,21,51]. Similarly, there are many existing studies on erasure coding, and an increasing interest in deploying it in storage and peer-to-peer systems [10,[41][42][43].…”

“…Again, our effort complements the above work and recent research in scientific data caching [36] by investigating the combination of striping and caching on unreliable distributed workstations. This paper targets distributed caches built by amassing donated disk space, and is closely related to storage aggregation systems [2,9] and more generally, resource scavenging systems [32,50]. However, existing storage aggregation systems are typically designed to act as general-purpose file systems and require high data availability.…”

“…However, existing storage aggregation systems are typically designed to act as general-purpose file systems and require high data availability. For instance, both FARSITE [2] and Kosha [9] replicate each dataset at least three times. In contrast, our system is positioned as a best-effort cache for better data access performance.…”

“…We examine two commonly used local recovery mechanisms, replication and erasure coding, in the distributed cache setting and compare them with our proposed strategies. Replication is widely used in parallel or distributed file systems [21,45] and storage aggregation systems [2,9,17]. With replication, a dataset is duplicated a number of times in a storage system, to prepare for disk/network failure.…”

“…Studies have shown that, on average, half of the disk space on desktop workstations is free and that this fraction is only increasing as disks become larger [2,15]. As a result, numerous distributed file systems exist that aggregate desktop storage, both in the local [2,9] and the wide-area [16,17] environments. These distributed storage systems aim at delivering reliable and persistent storage with data availability levels commensurate to centralized file servers.…”