A number of ongoing research projects follow a partition-based approach to provide highly scalable distributed storage services. These systems maintain namespaces that reference objects distributed across multiple locations in the system. Typically, atomic commitment protocols, such as 2-phase commit, are used for updating the namespace, in order to guarantee its consistency even in the presence of failures. Atomic commitment protocols are known to impose a high overhead to failure-free execution. Furthermore, they use conservative recovery procedures and may considerably restrict the concurrency of overlapping operations in the system. This paper proposes a set of new protocols implementing the fundamental operations in a distributed namespace. The protocols impose a minimal overhead to failure-free execution. They are robust against both communication and host failures, and use aggressive recovery procedures to re-execute incomplete operations. The proposed protocols are compared with their 2-phase commit counterparts and are shown to outperform them in all critical performance factors: communication roundtrips, synchronous I/O, operation concurrency. 1.

In this paper, we describe an architecture, NFS^2, for uniting several NFS servers under a single namespace. This architecture has some interesting properties. First, the physical file systems that make up an NFS^2 instance, i.e., the file systems on the individual NFS servers, may be heterogeneous. This, combined with the way the NFS^2 namespace is constructed, allows files of different types (text, video, etc.) to be served from file servers (potentially) optimized for each type. Second, NFS^2 storage is strictly partitionedach NFS server is solely responsible for allocating the resources under its control. This eliminates resource contention and distributed lock management, commonly found in cluster file systems. Third, because the system may be constructed with standard NFS servers, it can benefit from existing high-availability solutions for individual nodes, and performance improves as NFS servers improve. Last, but not least, the system is extremely easy to manage--new resources may be added to a configuration by simply switching on a new server, which is then seamlessly integrated into the cluster. An extended version of this architecture is the basis for a completed prototype in Linux [5].

P2P overlays offer a convenient way to host an infrastructure that can scale to the size of the Internet and yet manageable. Current proposals, however, do not offer support for structuring data, other than assuming a distributed hash table.In reality, both applications and users typically organize data in a structured form. One such popular structure is tree as employed in a file system, and a database. A naïve approach such as hashing the pathname not only ignores locality in important operations such as file/directory lookup, but also results in uncontrollable, massive object relocations when rename on a path component occur. In this paper, we investigate policies and strategies that place a tree onto the flat storage space of P2P systems. We found that, in general, there exists a tradeoff between lookup performance and balanced storage utilization, and attempts to balance these two requirements calls for intelligent placement decision. 1

Storage is increasingly becoming a commodity shared in global scale, either within the infrastructure of large organizations or by outsourcing to Storage Service Providers. Storage resources are managed and shared in the form of logical volumes; that is, virtual disks that aggregate resources from multiple, distributed physical devices and storage area networks. Logical volumes are dynamically assigned to servers according to a global resource utility model.

Current peer-to-peer storage utilities offer a convenient flat storage space, delegating the organization and presentation of data to upper layers. In reality, both applications and users typically organize data in a structured form. One such popular structure is hierarchical namespace as employed in a file system. A nave approach such as hashing the pathname of file system not only ignores locality in important operations such as file/directory lookup, but also results in uncontrollable, massive object relocations when rename on path component occurs.