Per Brinch Hansen
|Born||13 November 1938|
|Died||31 July 2007 (aged 68)|
Syracuse, New York
|Alma mater||Technical University of Denmark|
Per Brinch Hansen was born in Frederiksberg, an enclave surrounded by Copenhagen, Denmark. His father, Jørgen Brinch Hansen, worked as a civil engineer, becoming a leading expert in soil mechanics, and later accepting a professorship at Technical University of Denmark. His mother, Elsebeth Brinch Hansen (née Ring), was the daughter of Danish composer Oluf Ring and worked as a hairdresser before marrying.
Brinch Hansen attended St. Jørgens Gymnasium and then studied electrical engineering at Technical University of Denmark where he sought an area to pursue that "was still in its pioneering phase" on the belief that "If a subject was being taught, it was probably already too late to make fundamental contributions." After a seven-week student internship at IBM's Hursley Laboratory in England, he decided to dedicate his career to computers. Initially focused on computer construction, reading a book on IBM's Project Stretch that described computer organization from a programmer's point of view refocused his interest toward becoming a computer architect.
After completing a master's degree in electronic engineering in 1963, Brinch Hansen landed a job at Regnecentralen, then a research institution under The Danish Academy of Technical Sciences There, his first significant project was writing a parser for a COBOL compiler for the Siemens 3003 computer., working in the compiler group, led by Peter Naur and Jørn Jensen.
I now understand that it was really a small operating system, I had programmed. However, in the mid 1960s, the dividing line between language implementation and operating systems was still not clearly understood.
In 1966, Brinch Hansen moved to Henning Isaksson's hardware group at Regnecentralen, by then a company with shareholders. Together with Peter Kraft, he defined the architecture and instruction set for Regnecentralen's third computer, the RC 4000, using Algol 60 as a hardware definition language to produce a formal specification.
Inexperienced with multiprogramming, he used a copy of Cooperating Sequential ProcessesEdsger Dijkstra had sent him to understand process synchronization using semaphores, and then implemented a specialized RC 4000 real-time monitor for use in managing a fertilizer plant. Peter Kraft and Charles Simonyi, who was still a teenager, wrote a p-code interpreter and data logging task programs that were compiled to p-code.
In the summer of 1967, Brinch Hansen left Regnecentralen's hardware group to become head of RC 4000 software development, where he led a team including Jørn Jensen, Peter Kraft and Søren Lauesen in defining a general-purpose RC 4000 multiprogramming system, with a goal to avoid developing a custom real-time control operating system for every RC 4000 installation, and to support batch processing and time-sharing as well. The resulting system was not a complete operating system, but a small kernel providing the mechanisms upon which operating systems for different purposes could be built. By the spring of 1969, a well-documented, reliable version of the RC 4000 multiprogramming system was running.
In late 1970, Brinch Hansen moved to Pittsburgh, accepting an invitation from Alan Perlis to visit the Department of Computer Science at Carnegie Mellon University as a research associate, while he wrote the first systematic textbook on operating system principles. During this time, at the 1971 Summer School in Marktoberdorf and a symposium in Belfast, Brinch Hansen, Tony Hoare and Dijkstra began to discuss ideas that evolved into the monitor concept. In the spring of 1972, after reading about the class concept invented by Ole-Johan Dahl and Kristen Nygaard for Simula 67, Brinch Hansen completed his text with a chapter on resource protection that proposed the first monitor notation, using shared classes. In 1973, Operating System Principles was published, becoming the first comprehensive textbook on operating systems.
In July 1972, Brinch Hansen joined the faculty of Caltech as an Associate Professor of computer science, where he began work on defining a programming language with concurrent processes and monitors. In April 1974, he distributed a technical report on Concurrent Pascal. A Concurrent Pascal compiler for the PDP 11/45, written by Brinch Hansen's doctoral student, Al Hartmann, was released in January 1975. Subsequently, Brinch Hansen began writing model operating systems in Concurrent Pascal, to evaluate the language. In May 1975, he completed Solo, a single-user operating system for development of Concurrent Pascal programs. Next, he rewrote the original RC 4000 real-time scheduler in Concurrent Pascal, taking three days to write it, and three hours of machine time to systematically test it.
When the Caltech computer science department shifted focus toward computer engineering and away from programming Brinch Hansen decided to leave, rather than to seek tenure there.
In 1976, Brinch Hansen chose USC for his next post, so that his family could remain in their Altadena home. Joining the faculty as a tenured full professor, and first chair of a newly created computer science department, he led efforts to identify and attract top-notch faculty to build a first rate department. By 1980, USC's computer science department was ranked by the National Research Council as one of the top ten in the US.
While at USC, Brinch Hansen wrote his second book, The Architecture of Concurrent Programs, based on his work developing operating systems with Concurrent Pascal. Published in 1977, it was the first book on concurrent programming. In 1978, Brinch Hansen became the first computer scientist awarded the Doctor Technices degree, the highest academic distinction within engineering and technological science in Denmark, for the work documented in The Architecture of Concurrent Programs. Later in 1978, Brinch Hansen published the Distributed Processes language concept, proposing the use of remote procedure calls to synchronize processes running across a microcomputer network.
Also in 1978, L. J. Sevins and Steve Goings from Mostek visited Brinch Hansen at USC, where he outlined a low-cost multiprocessor architecture. Mostek began a project to implement such a multiprocessor, with Brinch Hansen working as a consultant. Brinch Hansen developed a new concurrent programming language, Edison, for the project. As with the RC 4000 project, Edison was also used as a formal specification language for the hardware. Mostek got an initial 4-node multiprocessor working and Brinch Hansen wrote a portable Edison compiler on a PDP 11/55, but shortly after, United Technologies acquired Mostek and cancelled the project. In 1982, Brinch Hansen moved the Edison system to an IBM PC, and then published his third book, Programming a Personal Computer.
In 1982, Brinch Hansen was named the first Henry Salvatori Professor of Computer Science at USC.
In 1984, feeling homesick for Denmark after fourteen years abroad, Brinch Hansen left USC and joined the faculty of the University of Copenhagen as a Professor of datalogy. In 1985, he was elected a Fellow of the IEEE. Later in 1985, his fourth book Brinch Hansen on Pascal Compilers, which he used for his own compiler course, was published.
While designing a multicomputer operating system for Danish company GN Elmi, Brinch Hansen concluded he needed a new language, this time leveraging the message passing paradigm of Hoare's CSP. The resulting language, Joyce, removed a major limitation of CSP by introducing parallel recursion. Brinch Hansen developed a portable implementation on an IBM PC.
After finding that neither he nor his family felt at home in Denmark, Brinch Hansen decided to return to the US, but discovered that their immigration status required them to do so very quickly. Brinch Hansen contacted John Reynolds at Syracuse University and in 1987 he joined the faculty as a Distinguished Professor.
Working with his student Rangachari Anand, Joyce was moved to an Encore Multimax 320 multiprocessor at SU's Northeast Parallel Architectures Center. Recognizing the scaling limitations of multiprocessors, however, Brinch Hansen sought a suitable multicomputer for further work. Acquiring a Meiko Computing Surface in 1989, he began experimenting with scientific applications by developing parallel programs for Householder reduction and then n-body simulation as learning exercises, and was surprised to find that both programs had nearly identical control structures. Concluding that both fit an "all-pairs paradigm," he then focused on exploring reusable parallel algorithm structures he termed "programming paradigms" or "generic programs" (later, popularly known as "design patterns"). In 1995, Brinch Hansen's fifth book, Studies in Computational Science: Parallel Programming Paradigms was published, with programs rewritten in SuperPascal, a fully implemented publication language he created for parallel algorithms.
In his later years, Brinch Hansen published a retrospective of his most important papers, The Search for Simplicity (1996), a text for a course in programming for non-majors, Programming for Everyone in Java (1999), a retrospective on the evolution of operating systems, Classic Operating Systems: From Batch Processing to Distributed Systems (2001), and a retrospective on the evolution of concurrent programming, The Origin of Concurrent Programming: From Semaphores to Remote Procedure Calls (2002). He self-published a 2004 memoir, A Programmer's Story: The Life of a Computer Pioneer, on his website.
In 2002, Brinch Hansen was awarded the IEEE Computer Pioneer Award "For pioneering development in operating systems and concurrent programming exemplified by work on the RC 4000 multiprogramming system, monitors, and Concurrent Pascal."
On July 31, 2007, Brinch Hansen died, shortly after being diagnosed with terminal cancer.
Brinch Hansen met his wife Milena (née Hrastar) on a class trip to Yugoslavia in 1962. They married in 1965 and had two children, daughter Mette and son Thomas.
In a career spanning five decades, Brinch Hansen made seminal contributions in the areas of operating systems, concurrent computing and parallel and distributed computing that were influential in shaping the development of those fields and advancing them from ad hoc techniques to systematic engineering disciplines. The impact of his work was amplified by his "relentless search for simplicity," his insistence on clarity, writing and rewriting his papers many times before publishing them, becoming "one of the best explainers in the business," and his commitment to testing concepts in working systems – Brinch Hansen implemented and distributed working systems for new concepts he developed, noting:
It is not uncommon for a computer scientist to make a proposal without testing whether it is any good in practice. After spending 3 days writing up the monitor proposal and 3 years implementing it, I can very well understand this temptation. It is perhaps also sometimes a human response to the tremendous pressure on university professors to get funding and recognition fast. Nevertheless, we must remember that only one thing counts in engineering: Does it work (not "might it work" or "wouldn't it be nice if it did")?
The RC 4000 multiprogramming system introduced the now-standard concept of an operating system kernel and the separation of policy and mechanism in operating system design. Modern microkernel architectures trace their roots to the extensible nucleus architecture of the RC 4000. Improving microkernel performance was a major theme in operating system research for three decades after the RC 4000.
Brinch Hansen's text, Operating System Principles, was the first comprehensive textbook on operating systems. Eventually published in six languages (English, Japanese, German, Czech, Polish and Serbo-Croatian), it remained in print for decades, and years after the RC 4000 system it described had become outdated. In 1990, nearly two decades after its initial publication, P.J. Plauger reviewed it, saying:
This book is terribly dated. It describes many of the RC4000's operating-system algorithms. The RC4000 is slow, small, and starved for peripherals by today's standards. The algorithms are presented in a subset of Pascal. Why bother? The answer is that Brinch Hansen is one of the best explainers in the business. He explains things clearly and to the point. He has an eye for the general principle behind the example, but manages to avoid unnecessary abstraction. After all these years, he is still a pleasure to read.
In 2005, ACM members voted Operating System Principles a top 25 classic book, in a survey to choose out-of-print classic books to add to the ACM Digital Library. Operating System Principles ranked 15th in the survey, appearing on 8.8% of ballots. It is now available from the ACM Digital Library.
Using Concurrent Pascal, Brinch Hansen demonstrated that it was feasible to fully implement operating systems in high level languages, and that doing so reduced the development effort by one to two orders of magnitude. Furthermore, entire operating systems could be published, and understood in their entirety. Brinch Hansen did precisely that in The Architecture of Concurrent Programs, leading Roy Maddux and Harlan Mills to declare:
Part two of the book is indeed remarkable. Here, an entire operating system is visible, with every line of program open to scrutiny. There is no hidden mystery, and after studying such extensive examples, the reader feels that he could tackle similar jobs and that he could change the system at will. Never before have we seen an operating system shown in such detail and in a manner so amenable to modification.
Brinch Hansen and Hoare developed the monitor concept. Brinch Hansen published the first monitor notation, adopting the class concept of Simula 67, and invented a queueing mechanism. Hoare refined the rules of process resumption. Brinch Hansen created the first implementation of monitors, in Concurrent Pascal. More than a dozen additional monitor-based languages had been created by 1990: Simone, Modula, CSP/k, CCNPascal, PLY, Pascal Plus, Mesa, SB-Mod, Concurrent Euclid, Pascalc, Concurrent C, Emerald, Real-time Euclid, Pascal-FC, Turing Plus, Predula.
Concurrent Pascal was the first concurrent programming language: the first language developed specifically for concurrent programming, and more importantly, the first language to demonstrate that it was possible to incorporate secure, high-level facilities for concurrency, where the system could guarantee that processes access disjoint sets of variables only and do not interfere with each other in time dependent ways. Hoare described it as "an outstanding example of the best of academic research in this area."
Source and portable code for Concurrent Pascal and the Solo operating system were distributed to at least 75 companies and 100 universities in 21 countries, resulting in its widespread adoption, porting and adaptation in both industry and academia. Greg Andrews observed that Concurrent Pascal and monitors "greatly influenced most subsequent concurrent language proposals."
Brinch Hansen's The Architecture of Concurrent Programs was the first book on concurrent programming, and was eventually published in three languages (English, Japanese, and German). More than a dozen years after its publication, P. J. Plauger observed:
Sure, improvements have been made in the past dozen years. We have better synchronization algorithms and fancier (if not necessarily better) languages with concurrency control. But you won't find a better overview of concurrent programming than this book. At least I haven't found one.
In 2005, ACM members voted The Architecture of Concurrent Programs a top 25 classic book, ranking it 18th in the survey, and appearing on 8% of ballots. It is now available from the ACM Digital Library.
Remote procedure calls used in modern operating systems trace their roots back to the RC 4000 multiprogramming system, which used a request-response communication protocol for process synchronization.
In his 1978 paper on distributed processes, Brinch Hansen proposed a language for distributed computing based on external requests consisting of procedure calls between processes. This later became known in an operating system context as the remote procedure call.
Edited: 2021-06-18 18:37:03