Search

All too often, decisions about whom to trust in computer systems are driven by the needs of system management rather than data security. In particular data storage is often entrusted to people who have no role in creating or using the data-through outsourcing of data management, hiring of outside consultants to administer servers, or even collocation servers in physically insecure machine rooms to gain better network, connectivity. This paper outlines the design of SUNDR, a network file system designed to run on untrusted servers. SUNDR servers can safely be managed by people who have no permission to read or write data stored in the file system. Thus, people can base their trust decisions on who needs to use data and their administrative decisions on how best to manage the data. Moreover, with SUNDR, attackers will no longer be able to wreak havoc by compromising servers and tampering with data. They will need to compromise clients while legitimate users are logged on. Since clients do not need to accept incoming network connections, they can more easily be firewalled and protected from compromise than servers.

The first generation of peer-to-peer file sharing systems followed the traditional client-server paradigm. However, legality and scalability issues have driven the development of decentralized file sharing protocols; the most popular of these being Gnutella. To date, such systems have been unable to match the quality of service (QoS) offered by centralized architectures. AGnuS improves QoS on Gnutella by increasing file availability, improving network friendliness and increasing file quality. This is achieved by layering caching, load balancing, content-based routing and filtering services on top of the core Gnutella protocol.

In this paper we present a new web based Distributed File System server management tool to perform load balancing across multiple servers. The Distributed File System from Distributed Computing Environment (DCE DFS) is a collection of many file systems mounted onto a single virtual file system space with a single namespace. The tool is based on rule-based data mining techniques and graph analysis algorithms. The data mining procedures identify DFS file access patterns and the graph analysis and statistical information relocates the filesets between different file servers. We demonstrate our tool on data collected for five months on DFS servers in a production environment. Experiments with this data show that our load balancing tool is useful to file system administrators to monitor, evaluate DFS state and to make intelligent decisions about file system transfers in order to balance the access request load on "read-write" filesets across DFS servers.

Many peer-to-peer (P2P) file sharing systems have been proposed to take advantage of high scalability and abundant resources at end-user machines. Previous approaches adopted either simple flooding or routing with complex structures, such as distributed hashing tables (DHT). However, these approaches did not consider the heterogeneous nature of the machines and the hierarchy of networks on the Internet. This paper presents the peer-to-peer asymmetric file sharing system (PASS), a novel approach to P2P file sharing, which accounts for the different capabilities and network locations of the participating machines. Our system selects only a portion of high-capacity machines (supernodes) for routing support, and organizes the network by using location information. We show that our key-coverage based directory replication improves the file search performance to a small constant number of routing hops, regardless of the network size.

In this paper we propose and evaluate a new NAS based system which is a consolidated file system in a network environment. The consolidated file system utilizes storage boxes cooperating as peers to provide all file system services, and presents itself to its NFS clients as a unified storage server. We have prototyped several key operations in this system. To provide more performance insights, we build a comprehensive system simulator and conduct extensive performance studies for the system via the simulator This simulator has been run against a workload of a real Web server and the system behavior of this workload is studied. In light of the experimental results, a consolidated file system is being implemented in full scale.

With the advancement of wireless networks and mobile computing, there is an increasing need to build a mobile file system that can perform efficiently and correctly for accessing online information. Previous system research on mobile file systems is based on some experimental platforms. We describe the design and implementation of a mobile file system on an open platform, the Linux kernel, and at the same time, our mobile file system is compatible with the popular NFS 2.0 protocol. We formally define the file semantics of our mobile file system, which we called NFS/M. We also specify the conditions of object conflict as well as our conflict resolution algorithms. NFS/M supports client side caching, data prefetching, file system service during the disconnected mode, data reintegration and conflict resolution on various file system objects. Since the NFS/M is based on an open platform, it serves as a basic building block for developing future mobile computing applications.

This paper presents the design and implementation of the Mobile File and Adaptation System (MFAS), a system to provide efficient data access mechanisms over variable quality of service networks. MFAS supports fine-grained file caching and consistency, and provides a mechanism for mobility-aware applications to adapt the caching and consistency behavior as network conditions change. Over low-bandwidth links, MFAS shows substantial performance gains over typical network file systems by sending only the modified portions of files. To our knowledge, MFAS is the first system that efficiently supports application-directed consistency policies at the subfile level.

The Internet File System (IFS) extends the scope of file systems from LANs to the Internet, encouraging collaboration and information dissemination on a much broader scale. In addition, Internet resources (e.g. Web pages, Gopher Information, and Network News) become files in IFS, allowing file system APIs and existing tools and commands to be used to manipulate Internet resources just like conventional files. IFS is unique in that it introduces a logical layer between applications and operating systems, and integrates widely-used protocols like FTP, HTTP, GOPHER, NNTP, RSH in this layer. The system is transparent to applications and operating systems, and requires no modification of software nor change in management of Internet file servers. A prototype of IFS is currently running on Sun OS 4.1, Solaris, HP-UX, SVR4, SGI MIPS, and Linux.

Recently, LINUX cluster file systems using the storage area network (SAN) have been developed. The Global File System (GFS) implemented at the University of Minnesota is such an example. These systems do not use a central file server, and instead, multiple clients sharing the whole disk storage through Fibre Channel can freely access disk storage and act as file servers. Accordingly, they offer advantages such as availability, load balancing, and scalability.

Global File Store (GFS) is a pervasive computing oriented file storage system. It puts forward an idea - put all files on the Internet. There are 3 features associated with GFS: massive storage, high performance and mobile computing support. The approach to construct GFS is to implement a layer on top of existing heterogeneous file systems. The above features are realized by means of multiple namespace, extensive use of caching and replication, and highly scalable architecture. The system is in the stage of prototype development.

In this paper we have carried out an experimental study of the I/O latency of a distributed file system over a parallel multimedia platform. Our experimental platform uses NFS (Network File System) to support an application characterized by its high processing and storage requirements. We evaluate the overall system by varying several NFS parameters as well as the underlying network topology and data access patterns. We have also examined the relevance of two other important file management access mechanisms: caching and prefetching. Our experimental results have allowed us to identify the major system parameters affecting the performance of NFS. We then propose the use of a predictive data access mechanisms. Under this mechanism, the file server should detect the similarities among the most recent I/O requests incoming from a set of cooperating clients. Once having identified the access pattern, the file server will be able to multicast the required data even before receiving explicit I/O data requests from the clients. Under this predictive mechanism, the system performance should improve as the number of explicit I/O requests directed to the file server decreases.

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 round-trips, synchronous I/O, operation concurrency.

In a distributed file system built on network attached storage, client computers access data directly from shared storage, rather than submitting I/O requests through a server. Without a server marshaling access to data, if a computer fails or becomes isolated in a network partition while holding locks on cached data objects, those objects become inaccessible to other computers until a locking authority can guarantee that the lock holder will not again directly access these data. We describe a server that acts as the locking authority and implements a lease-based protocol for revoking access to data objects locked by an isolated or failed computer. When a lease expires, the server can be assured that the client no longer acts on locked data, and can safely redistribute locks to other clients. During normal operation, this protocol invokes no message overhead, and uses no memory and performs no computation at the locking authority.