Network File System Research Articles

DAFS: A New High-Performance Networked File System

The Direct Access File System (DAFS) is a remote file-access protocol designed to take advantage of new high-throughput, low-latency network technology.

NFS‐based secure file sharing over multiple administrative domains with minimal administration

AbstractAlthough several schemes of file sharing for Unixes have been proposed so far, file sharing by these schemes between different administrative domains is difficult and imposes enormous burden on the administrators even if file sharing is possible since the user information differs by such computers in general. In this paper, the file system PNFS (Proxy/Private Network File System), which makes file sharing based on NFS easy between different administrative domains, is proposed. With PNFS, a user can share files of a server that has his account without intervention of its administrator. Although, in general, a decrease in security occurs when intervention of an administrator is omitted, security is enhanced by encryption and the like by the PNFS. In addition, the efficacy of the proposed system has been verified by performance evaluation of a prototype system, which shows a practically sufficient throughput even with simultaneous accesses from a multiple number of clients. © 2002 Wiley Periodicals, Inc. Syst Comp Jpn, 33(14): 50–58, 2002; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/scj.10027

Load balancing across a highly heterogeneous processor cluster using file status probes

Ray tracing for Kirchhoff migration is an “embarrassingly parallel” operation, and so can be distributed across a cluster of workstations or personal computers. However, since the computation time associated with individual rays varies with the length of a ray and with the complexity of the velocity model and processors can have widely different computation speeds, load balancing (and hence computational efficiency) across a heterogeneous cluster of processors is a significant problem. Load balancing operations are traditionally implemented by the operating system outside of the program to be executed, and generally require a priori knowledge of the number and the types of processors to be used. We develop an alternative load balancing scheme that is implemented by UNIX Network File System functions for interrogating the existence of files in a working directory from within a running program and so requires no specialized algorithms for monitoring job flow, nor a priori knowledge of the number or types of processors that are participating in the job execution. Processors can be added or removed at any time during the job execution, and the system is crash-proof in the sense that a job will successfully complete as long as at least one processor remains running; there is no master node, so no individual node is indispensable. These ideas are successfully applied to ray tracing for 3-D prestack Kirchhoff depth migration of a field ground-penetrating radar data set from the Ferron Sandstone in east-central Utah.

Interposed request routing for scalable network storage

This paper explores interposed request routing in Slice, a new storage system architecture for high-speed networks incorporating network-attached block storage. Slice interposes a request switching filter---called a μ proxy ---along each client's network path to the storage service (e.g., in a network adapter or switch). The μproxy intercepts request traffic and distributes it across a server ensemble. We propose request routing schemes for I/O and file service traffic, and explore their effect on service structure. The Slice prototype uses a packet filter μproxy to virtualize the standard Network File System (NFS) protocol, presenting to NFS clients a unified shared file volume with scalable bandwidth and capacity. Experimental results from the industry-standard SPECsfs97 workload demonstrate that the architecture enables construction of powerful network-attached storage services by aggregating cost-effective components on a switched Gigabit Ethernet LAN.

On the Performance and Scalability of NFS over a Myrinet-based Network of Workstations

The development of high performance networks and multimedia applications has spurred the need of developing more efficient distributed file systems. In the past, distributed file systems, such as NFS (Network File System), have been developed aiming to provide transparent remote access to shared file systems across computer platforms. Even though, NFS has been widely studied and new releases are being developed, there are still some open issues regarding its performance and scalability. In this work we explore the operation and performance of NFS when used to support parallel data intensive-processing applications over a cluster of workstations interconnected by a 1-Gbps network. We first consider a parallel implementation of a video compression software encoder. The application is characterized by the large amount of data to be distributed to the processing workstations. Through an exhaustive campaign of experiments, we evaluate the impact of several system key parameters over the performance and scalability of NFS. We then explore the operation of NFS using a synthetic load pattern. Our main results provide the guidelines to set-up some of the key parameters for a more efficient operation of NFS over a cluster of workstations interconnected by a high-speed interconnection network.

Failure-Atomic File Access in the Slice Interposed Network Storage System

This paper presents a recovery protocol for block I/O operations in Slice, a storage system architecture for high-speed LANs incorporating network-attached block storage. The goal of the Slice architecture is to provide a network file service with scalable bandwidth and capacity while preserving compatibility with off-the-shelf clients and file server appliances. The Slice prototype virtualizes the Network File System (NFS) protocol by interposing a request switching filter at the client's interface to the network storage system. The distributed Slice architecture separates functions typically combined in central file servers, introducing new challenges for failure atomicity. This paper presents a protocol for atomic file operations and recovery in the Slice architecture, and related support for reliable file storage using mirrored striping. Experimental results from the Slice prototype show that the protocol has low cost in the common case, allowing the system to deliver client file access bandwidths approaching gigabit-per-second network speeds.

A low-bandwidth network file system

Users rarely consider running network file systems over slow or wide-area networks, as the performance would be unacceptable and the bandwidth consumption too high. Nonetheless, efficient remote file access would often be desirable over such networks---particularly when high latency makes remote login sessions unresponsive. Rather than run interactive programs such as editors remotely, users could run the programs locally and manipulate remote files through the file system. To do so, however, would require a network file system that consumes less bandwidth than most current file systems.This paper presents LBFS, a network file system designed for low-bandwidth networks. LBFS exploits similarities between files or versions of the same file to save bandwidth. It avoids sending data over the network when the same data can already be found in the server's file system or the client's cache. Using this technique in conjunction with conventional compression and caching, LBFS consumes over an order of magnitude less bandwidth than traditional network file systems on common workloads.

The design and implementation of the Gecko NFS Web proxy

AbstractThe World Wide Web uses a naming scheme (URLs), transfer protocol (HTTP), and cache algorithms that are different from traditional network file systems. This makes it impossible for unmodified Unix applications to access the Web; as a result, the Web is only accessible to special Web‐enabled applications. Gecko bridges the gap between Web‐enabled applications and conventional applications, allowing any unmodified Unix application to access the Web. We developed two prototypes of the Gecko system at the University of Arizona and incorporated the many lessons learned from the first into the second. Minimizing the amount of state was a key goal of the Gecko redesign and the second prototype uses a unique compression technique to reduce server state. Experiments performed on the prototype show that Gecko not only makes the Web accessible to unmodified applications, but does so at a performance that meets or exceeds that of HTTP. Copyright © 2001 John Wiley & Sons, Ltd.

COMPUTATIONAL ASPECTS OF THE LOCAL ITERATIVE MONTE CARLO TECHNIQUE

Lately, the Local Iterative Monte Carlo technique was introduced for an efficient simulation of effects connected to sparsely populated regions in semiconductor devices like hot electron effects in silicon MOSFETs. This paper focuses on computational aspects of this new Monte Carlo technique, namely the reduction of the computation time by parallel computation and the reuse of drift information. The Local Iterative Monte Carlo technique combines short Monte Carlo particle flight simulations with an iteration process to a complete device simulation. The separation between short Monte Carlo simulations and the iteration process makes a simple parallelization strategy possible. The necessary data transfer is small and can be performed via the Network File System. An almost linear speed up could be achieved. Besides by parallelization, the computational expenses can be significantly reduced, when the results of the short Monte Carlo simulations are memorized in a drift table and used several times. A comparison between a bulk, a one-dimensional and the two-dimensional Local Iterative Monte Carlo simulation reveals that by using the drift information more than once, becomes increasingly efficient with increasing dimension of the simulation.

Measurement and in-service monitoring for QoS violations and spare capacity estimations in ATM network

In this paper, we investigate the traffic behaviour of a local ATM network. The major loading application in this network is network file system (NFS). The cell data are collected between switches. The results show that the traces present two different kinds of properties: long-range dependence (LRD) or short-range dependence (SRD). Then we propose a new method for in-service QoS violations and spare capacity estimations. For a typical ATM network, a large number of cells must be observed before we can get a statistically meaningful cell loss probability (CLP) result but these results may be too obsolete for effective network management. On the contrary, it is difficult to choose a suitable representative analytical model for many applications in ATM networks that exhibit varied statistical properties. Besides, finding analytical models may be complex and slow to fit. For both SRD and LRD processes, there exists a relationship between the logarithm of the CLP and the buffer size. The relationship for SRD processes is often linear while it is polynomial for LRD processes. If we can know the traffic presents LRD or SRD, the above relationships help us estimate the CLP of the queuing behaviour. The proposed method in this paper uses a simple regression test to select the suitable model: LRD or SRD. Then it employs the relationships between log(CLP) and buffer size of the selected model. With observations of cell loss for several small-buffer pseudo-queues and the selected regression model, the scheme can estimate the QoS and the spare capacity of the actual system. It requires only a short observation period and it does not need analytical models describing the statistics of traffic. The collected data are used for the basis of the simulation sources and the results of the simulations show the effectiveness of the method for both LRD and SRD trace data.

Separating key management from file system security

No secure network file system has ever grown to span the Internet. Existing systems all lack adequate key management for security at a global scale. Given the diversity of the Internet, any particular mechanism a file system employs to manage keys will fail to support many types of use.We propose separating key management from file system security, letting the world share a single global file system no matter how individuals manage keys. We present SFS, a secure file system that avoids internal key management. While other file systems need key management to map file names to encryption keys, SFS file names effectively contain public keys, making them self-certifying pathnames. Key management in SFS occurs outside of the file system, in whatever procedure users choose to generate file names.Self-certifying pathnames free SFS clients from any notion of administrative realm, making inter-realm file sharing trivial. They let users authenticate servers through a number of different techniques. The file namespace doubles as a key certification namespace, so that people can realize many key management schemes using only standard file utilities. Finally, with self-certifying pathnames, people can bootstrap one key management mechanism using another. These properties make SFS more versatile than any file system with built-in key management.

Special Issue on Software Support for Distributed Computing: Guest Editors' Introduction

Rapid advances in communications technology and the proliferation of inexpensive PCs and workstations have created a wide avenue for distributed computing to move into mainstream computing. Distributed computing is no longer confined to research laboratories but is available to the mass community of computer users. The shift of status would not have been possible if it were not for the immense effort that researchers and developers have put into creating the necessary supporting software networking protocols, distributed operating systems, network file systems, remote procedure calls, message passing interfaces, and most recently, object-oriented middleware. The advent of the Internet and the Web has also added fuel to the shift. The Internet as one gigantic distributed system is no longer a farfetched idea of the visionaries; it is now being used by some to try to solve really tough problems to search for extraterrestrial intelligence [1], to find the largest prime number [2], and to crack the RC5 block cipher [3], just to name a few. Many wide-area distributed systems that are in the making, the Globe system [6] for example, have worldwide scalability as one of their main goals. But distributed computing is not about just IP and HTTP, nor is it about setting up an Ethernet network. Distributed computing, as it always has been, is about achieving high performance at low cost; it is a kind of ``lowly parallel computing'' that dwindling costs and increasing power of the hardware have brought about [5]. Easy as it might sound, distributed computing is still struggling to reach the goal of high performance that is commensurate with the amount of hardware invested. To the researchers, distribution is a long-standing problem. We used to focus a lot on absolute performance, but when machines became faster, the focus seemed to have shifted; it did not, however, completely go away. Today, communication latency is a prime concern affecting performance in the design of high-performance distributed systems, especially when transparency is much demanded at application or user levels and when the system extends beyond a local area. Article ID jpdc.1999.1579, available online at http: www.idealibrary.com on

Design and implementation of a video network server

In this article, we propose a new video server method, which is an extension of an existing network file system to include video functions. Conventional network file systems cannot transfer video data in real time, because they do not have a mechanism to guarantee the video data transfer rate. Therefore, we use a new time slot method that combines the divided time slot method and the elevator seek method. Furthermore, we use a new method for sending video data from the server to the network. It performs rate control on existing network protocols that do not have rate control functionality. The result of implementation of these methods shows that the new time slot method decreases client response time to one-half to one-tenth compared with conventional methods, and we confirmed that the worst response delay by the new rate control method is within the permissible time that guarantees data transfer rate. We decreased response time for clients in our new video server method, by employing the time slot method to guarantee video data read-bit-rate from disks, we employ the time slot method to guarantee the bit rate in reading video data from disks, and we decreased the response time for clients (by using the time slot method in reading from disks).

Home network file system for home network based on IEEE-1394 technology

This paper proposes a network file system for home network based on IEEE-1394 technology. The home network file system enables real-time playback, recording of audio/video file and file sharing for HAVi (home audio video interoperability) compliant consumer devices.

Native-language-based distributed computing across network and filesystem boundaries

This paper discusses how the aspects unique to the Java programming language can be combined with complementary and unique aspects of other languages such as C and Fortran. This combining of the strong features of Java, such as portability and platform independence, with packages and legacy codes written in traditional languages such as C and Fortran results in a program blend which exhibits portability and speed not realizable by any of these languages individually. One area where this confluence of previously disparate language features has strong potential is in the area of distributed, concurrent computing over heterogeneous platforms and across local network and filesystem boundaries – the setting addressed within this paper. Also addressed in this paper are the pivotal aspects of the Java bytecode representation of a class object which makes the porting of shared libraries across network boundaries, filesystems, and architectures possible. © 1998 John Wiley & Sons, Ltd.

Data base and management system for clinical positron emission tomography (PET) studies.

A data base and management system connected to an image analysis system has been developed and utilized for clinical positron emission tomography (PET). This data base system, 1) is based on "GBASE", a general purpose data base, which runs on a UNIX work station, 2) works on a network file system and is connected to PET cameras and other data acquisition devices as well as to an image analysis system "Dr.View", 3) centrally manages the data stored in a data storage unit, 4) is easily modifiable and expandable, and 5) has a human friendly interface which requires minimum operation for registration, retrieval and management. We have been using this system to handle clinical PET data for seven years and have optimized the data base schema. As a result, this system has become a truly practical tool for the daily operation and is well-received by technologists, nuclear physicians and attending physicians.

Development of a collaborative CAD/CAM system

This paper presents the development of a collaborative CAD/CAM system ( COCADCAM). COCADCAM extends an existing single-location CAD/CAM system to a multi-location CAD/CAM application so that two geographically dispersed CAD/CAM users can work together on a three-dimensional CAD-geometry coediting and CAD-related tasks collaboratively and dynamically. COCADCAM dynamically supports CAD data communication that are not available in traditional single-location CAD/CAM. The dynamic data communication is achieved through the development of networking algorithms and CAD/CAM functions in this paper. The networking algorithms based on UNIX Interprocess Communication (IPC), the Network File System (NFS), and a connection-oriented client and server model under the Transmission Control Protocol/Internet Protocol (TCP/IP) suite. The CAD/CAM functions included surface modelling, simulation of a milling toolpath, and post-processing of an NC program following collaborative CAD-geometry coediting, which are directly or indirectly supported by the Application Programming Interface (API) of the CAD/CAM software. The networking algorithms and CAD/CAM functions together can facilitate an environment for CAD-geometry coediting and related tasks such as design, analysis and manufacture. COCADCAM has been successfully implemented through local area network (LAN) and the Internet; a remote machining cell is also linked so that the generated NC program based on a coedited free-form surface can be used for the physical machining operation. The algorithm proposed by COCADCAM can be referenced for the extension of other single-location CAD/CAM systems to multi-location applications.

Optimizing throughout in a workstation-based network file system over a high bandwidth local area network

This paper describes methods of optimizing a client/server network file system to advantage of high bandwidth local area networks in a conventional distributed computing environment. The environment contains hardware that removes network and disk bandwidth bottlenecks. The remaining bottlenecks at clients include excessive context switching, inefficient data translation, and cumbersome data encapsulation methods. When these are removed, the null-write performance of a current implementation of Sun's Network File System improves by 30%. A prototype system including a high speed RAM disk demonstrates an 18% improvement in overall write throughput. The prototype system fully utilizes the available peripheral bandwidth of the server.

The transfer protocols

There are many Internet-scale messaging systems: Telnet for duplex byte streams, File Transfer Protocol (FTP) for pulling files, Simple Mail Transfer Protocol (SMTP) for reliable e-mail messaging, Network News Transfer Protocol (NNTP) for pushing e-mail, HyperText Transfer Protocol (HTTP) for pulling Web pages, and a host of other special-purpose schemes, from Finger to Network File System (NFS). How can we make sense of it all? Understanding the history of these transfer protocols (TPs) is the key to charting their future, especially as new contenders like HTTP-NG (next-generation) emerge. In this article, I set forth a general ontology-a vocabulary for describing and classifying TPs, with the aim of developing a framework to help navigate among the alternatives.

Studying vertical dependence to improve NFS performance in wireless networks

Wireless networks experience a high level of errors and losses. These physical layer characteristics have an impact on the performance of the higher layers. In addition, the performance of each protocol layer is contingent on the behavior of the other layers. Vertical dependency is a term which describes this inter-dependence between layers. In the wireless and mobile environment, the effects of vertical dependence are particularly pronounced due to the dynamic nature of the environment and due to the fact that traditional assumptions about protocol layer interactions do not always hold. In this paper, we consider the vertical dependencies between various layers in the protocol stack, studying the performance of the Network File System under various error models and improvement techniques. Our experimental results demonstrate the dependency of the application performance on the details of the error characteristics and other protocol layers. After studying the vertical dependencies, we improve NFS performance by implementing changes to the application level reliability mechanisms. Understanding of the vertical dependencies enables development of effective methods for performance enhancement and efficient reaction to errors and changes on the wireless media.