The emergence of network-attached disks provides the possibility of transferring data between the storage system and the client directly. This offers new possibilities in building a distributed storage system. In this paper, we examine different storage organizations based on network-attached disks and compare the performance of these systems to a traditional system. Trace-driven simulations are used to measure the average response times of the client requests in two different workloads. We show that the advantages of distributing the server's network processing load to disks can be offset by losses in cache hits in a system based on network-attached disks. The two workloads we considered highlight this impact. We show that it is possible to offload significant load from the server by utilizing network-attached disks and point out that specific applications may be able to better exploit the features of network-attached disks.
展开▼
