Alfred Z. Spector

The ITC distributed file system: principles and design

This paper presents the design and rationale of a distributed file system for a network of more than 5000 personal computer workstations. While scale has been the dominant design influence, careful attention has also been paid to the goals of location transparency, user mobility and compatibility with existing operating system interfaces. Security is an important design consideration, and the mechanisms for it do not assume that the workstations or the network are secure. Caching of entire files at workstations is a key element in this design. A prototype of this system has been built and is in use by a user community of about 400 individuals. A refined implementation that will scale more gracefully and provide better performance is close to completion.

Synchronizing shared abstract types

Experimental data on transient errors from several digital computer systems is presented and analyzed. This is the first large scale public study on the statistical distribution of transient errors. The systems for which data has been collected are the DEC PDP-10 scries computers, the Cm* multiprocessor, and the C.vmp fault tolerant microprocessor. Statistical tests indicate that transient errors follow a decreasing hazard rate distribution. This is at variance with the standard assumption of constant hazard rates (exponential distribution) used in reliability modeling, and requires models of greater complexity for accurate results. Models of common fault tolerant redundant structures are developed using the Weibul! distribution, which has a time-varying hazard rate. Both analytical and simulation models arc used to analyze die differences between the reliabilities predicted by Weibull based transient error models and those predicted by exponential based models. The analysis indicates a significant difference between the models based on the exponential distribution and those based on the decreasing hazard rate Weibull distribution. Reliability differences ranging from -0.10 to +0.20 and factors greater than 2.0 in Mission Time Improvement for Weibull parameters equivalent to measured system behavior arc seen in the model results. System designers should be aware of these differences. Transient Error Reliability Models Based on Data Analysis

Support for Distributed Transactions in the TABS Prototype

The TABS prototype is an experimental facility that provides operating system-level support for distributed transactions that operate on shared abstract types. The facility is designed to simplify the construction of highly available and reliable distributed applications. This paper describes the TABS system model, the TABS prototypes structure, and certain aspects of its operation. The paper concludes with a discussion of the status of the project and a preliminary evaluation.

Distributed logging for transaction processing

Increased interest in using workstations and small processors for distributed transaction processing raises the question of how to implement the logs needed for transaction recovery. Although logs can be implemented with data written to duplexed disks on each processing node, this paper argues there are advantages if log data is written to multiple log server nodes. A simple analysis of expected logging loads leads to the conclusion that a high performance, microprocessor based processing node can support a log server if it uses efficient communication protocols and low latency, non volatile storage to buffer log data. The buffer is needed to reduce the processing time per log record and to increase throughput to the logging disk. An interface to the log servers using simple, robust, and efficient protocols is presented. Also described are the disk data structures that the log servers use. This paper concludes with a brief discussion of remaining design issues, the status of a prototype implementation, and plans for its completion.

Transactions: a construct for reliable distributed computing

Transactions have proven to be a useful tool for constructing reliable database systems and are likely to be useful in many types of distributed systems. To exploit transactions in a general purpose distributed system, each node can execute a transaction kernel that provides services necessary to support transactions at higher system levels. The transaction model that the kernel supports must permit arbitrary operations on the wide collection of data types used by programmers. New techniques must be developed for specifying the synchronization and recovery properties of abstract types that are used in transactions. Existing mechanisms for synchronization, recovery, deadlock management and communication are often inadequate to implement these types efficiently, and they must be adapted or replaced.

The TWA reservation system

Where can you find a solid, forthright overview of the computer systems and management behind airline reservations? NASAs space shuttle? Or any of the multitude of other large computer systems that support important projects or national activities? Its hard, sometimes impossible: partly because the people who worked on such systems often do not have the time to write about their experiences: and partly because many professional journalists who interview these people do not have the technical background to ferret out answers to the fundamental design questions addressed in these systems.

Camelot: a flexible, distributed transaction processing system

The Camelot Project has constructed a distributed transaction facility intended to support widespread use of transaction processing techniques. Camelot executes on a variety of uni- and multiprocessors on top of the Unix-compatible, Mach operating system. The authors describe the design decisions that make Camelot a flexible, easy-to-use system and briefly describe Camelots programming interfaces and algorithms. They discuss two applications of Camelot: an implementation of distributed ET-1 and a graphical room reservation system that uses the X Window Manager.<<ETX>>

An algorithm for replicated directories

This paper describes a replication algorithm for directory objects based upon Giffords weighted voting for files. The algorithm associates a version number with each possible key on every replica and thereby resolves an ambiguity that arises when directory entries are not stored in every replica. The range of keys associated with a version number changes dynamically; but in all instances, a separate version number is associated with each entry stored on every replica. The algorithm exhibits favorable availability and concurrency properties. There is no performance penalty for associating a version number with every possible key except on Delete operations, and simulation results show this overhead is small.

A weighted voting algorithm for replicated directories

Weighted voting is used as the basis for a replication technique for directories. This technique affords arbitrarily high data availability as well as high concurrency. Efficient algorithms are presented for all of the standard directory operations. A structural property of the replicated directory that permits the construction of an efficient algorithm for deletion is proven. Simulation results are presented and the system is modeled and analyzed. The analysis agrees well with the simulation, and the space and time performance are shown to be good for all configurations of the system.

The CIRRUS banking network

The CIRRUS banking network makes coast-to-coast automatic banking transactions possible. The system will soon be able to handle international currency transactions and point-of-sale transactions in stores.