Jane C. Blake,
Managing Editor

Scientists have long been motivators for the development of powerful computing environments. Two sections in this issue of the Journal address the requirements of scientific and technical computing. The first, from Digital's High Performance Technical Computing Group, looks at compiler and development tools that accelerate performance in parallel environments. The second section looks to the future of computing; University of California and Digital researchers present their work on a large, distributed computing environment suited to the needs of earth scientists studying global changes such as ocean dynamics, global warming, and ozone depletion. Digital was an early industry sponsor and participant in this joint research project, called Sequoia 2000.

To support the writing of parallel programs for computationally intense environments, Digital has extended DEC Fortran 90 by implementing most of High Performance Fortran (HPF) version 1.1. After reviewing the syntactic features of Fortran 90 and HPF, Jonathan Harris et al. focus on the HPF compiler design and explain the optimizations it performs to improve interprocessor communication in a distributed-memory environment, specifically, in workstation clusters (farms) based on Digital's 64-bit Alpha microprocessors.

The run-time support for this distributed environment is the Parallel Software Environment (PSE). Ed Benson, David LaFrance-Linden, Rich Warren, and Santa Wiryaman describe the PSE product, which is layered on the UNIX operating system and includes tools for developing parallel applications on clusters of up to 256 machines. They also examine design decisions relative to message-passing support in distributed systems and shared-memory systems; PSE supports network message passing, using TCP/IP or UDP/IP protocols, and shared memory.

Michael Stonebraker's paper opens the section featuring Sequoia 2000 research and is an overview of the project's objectives and status. The objectives encompassed support for high-performance I/O on terabyte data sets, placing all data in a DBMS, and providing new visualization tools and high-speed networking. After a discussion of the architectural layers, he reviews some lessons learned by participants--chief of which was to view the system as an end-to-end solution--and concludes with a look at future work.

An efficient means for locating and retrieving data from the vast stores in the Sequoia DBMS was the task addressed by the Sequoia 2000 Electronic Repository project team. Ray Larson, Chris Plaunt, Allison Woodruff, and Marti Hearst describe the Lassen text indexing and retrieval methods developed for the POSTGRES database system, the GIPSY system for automatic indexing of texts using geographic coordinates discussed in the text, and the TextTiling method for automatic partitioning of text documents to enhance retrieval.

The need for tools to browse through and to visualize Sequoia 2000 data was the impetus behind Tecate, a software platform on which browsing and visualization applications can be built. Peter Kochevar and Len Wanger present the features and functions of this research prototype, and offer details of the object model and the role of the interpretive Abstract Visualization Language (AVL) for programming. They conclude with example applications that browse data spaces.

The challenge of high-speed networking for Sequoia 2000 is the subject of the paper by Joseph Pasquale, Eric Anderson, Kevin Fall, and Jon Kay. In designing a distributed system that efficiently retrieves, stores, and transfers very large objects (in excess of tens or hundreds of megabytes), they focused on operating system I/O and network software. They describe two I/O system software solutions--container shipping and peer-to-peer I/O--that avoid data copying. Their TCP/IP network software solutions center on avoiding or reducing checksum computation.

The editors thank Jean Bonney, Digital's Director of External Research, for her help in obtaining the papers on Sequoia 2000 research and for writing the Foreword to this issue.

Our next issue will feature papers on multimedia and UNIX clusters topics.