Gene Amdahl: The Thomas Edison of Computing
Here’s a short computer-history puzzle, complete with five hints.
Name a successful computer engineer some call the Thomas Edison of computing.
Hint #1 What character in the history of computing built a reputation working for a large computer company, then created a startup that had more than 6,000 employees (2,000 employees more than Twitter has today) — and went on designing computers for decades?
Hint #2 He’s an engineer / entrepreneur who has a fundamental principle of computation named after him.
(Generations who recognize the names of few computer luminaries beyond Steve Jobs, Bill Gates, and Mark Zuckerberg are probably drawing a blank.)
Hint #3 This engineer received an undergraduate degree from South Dakota State University.
Hint #4 He was an entrepreneur who made a name for himself in the mainframe world.
Still no luck? Ready for your final hint?
Hint #5 FUD.
“FUD?” you ask? Stay tuned. But by now you should have guessed that we’re speaking of Gene Amdahl, whose computing career spanned nearly six decades. He died only recently, in November of 2015.
Fifty Years of Design Longevity — and Earnings
Just how important was Amdahl to the mainframe?
In a New York Times obituary for Amdahl, Stanford scientist and former collaborator Michael J. Flynn tried to explain just how extraordinary the IBM 360-series design had been:
“[It] set the design philosophy for computers for the next 50 years, and to this day it’s still out there, which is incredible. This same instruction set is still bringing in billions of dollars for IBM.”
A Mainframe Bio
Gene Amdahl worked on IBM’s designs for the 704, 709, and 7030. He left IBM in 1955, but returned in 1960 — a propitious re-hire for which one hopes a recruiter at IBM was properly rewarded.
What came next for Amdahl was key work on the team producing what became the IBM System/360 architecture. How important was the System/360?
“No new product offering has had greater impact on the computer industry than the IBM System/360,” said the authors of IBM’s 360 and Early 370 Systems, writing in 2003. “IBM’s 360 and Early 370 Systems describes the creation of this remarkable system and the developments it spawned, including its successor, System/370.” A decade later, that assessment still stands.
Once again, IBM could not keep this creative entrepreneur on its team. In 1970, he left New York for Sunnyvale, California, the year before it was first dubbed “Silicon Valley.” There, with help from Fujitsu, he set up Amdahl Corporation.
The Amdahl series computers were the first to use VLSI. When the Amdahl 470/V6 was released, it was priced around the same ($3.5M) as IBM’s roughly equivalent machine, the IBM 370/165, but it was a quarter the size and four times as powerful. Amdahl Corporation was later acquired by Fujitsu Ltd, who was a major investor all along, though it no longer operates as a separate subsidiary.
A Brief History of Amdahl’s Law
Syncsort is one of the oldest existing software companies. It was born in the mainframe era (and still thrives there), and by the time Syncsort was founded, in 1968, Gene Myron Amdahl had already presented his famous edict about parallel computing that still bears his name.
Parallelism has seen a resurgence of sorts with Big Data, Hadoop, and Google’s large-scale use of low-cost commodity hardware. And IBM, Syncsort, and the rest of the mainframe ecosystem have embraced this latest iteration of parallelism. Seasoned computer scientists know it is no apparition but more of what Amdahl foresaw decades before.
In that 1967 formulation, presented at the AFIPS Spring Joint Computer Conference, Amdahl presented an argument that he believed places a limit on the ability to speed up program execution using a parallel computer design. He postulated that a relatively few instructions that need to be executed in sequence (not in parallel) are a constraint on the speed to be gained by adding more processors. That “law” has occasionally been challenged, but it has often been used to estimate improvements to be gained by parallel processing.
John Gustafson, when he was working at Sandia, offered his eponymous corollary. More recently at AMD, he claimed that program size is not constant, and that there are circumstances under which Amdahl’s law is less predictive.
The problem continues to spawn further study. Xian-He Sun, chair and professor of the Department of Computer Science at the Illinois Institute of Technology, established what in 1990 he and Lionel Ni call Sun-Ni’s law, which suggests that memory constraints should be identified as important — hence the shorthand, “memory-bounded speedup.”
In a paper titled “Breaking the Law,” Thomas Lippert, director of the JA1/4lich Supercomputing Centre at JA1/4lich, Germany, suggested in 2013 that experiments from the EU Dynamical Exascale Entry Platform (DEEP) might present a challenge to the law. Even so, Lippert told CIO that he was by no means deprecating the power of Amdahl’s original formulation.
“On the contrary, I think, we are not taking [Amdahl’s law] serious enough. It is simply obvious that we should adapt the right piece of hardware to the corresponding concurrency.”
That Amdahl’s original law identified issues that still generate conversation in this fast-changing field speaks to the engineer’s prescience in that 1967 problem assessment.
And Finally, FUD
Today’s giant computer enterprises are known for keeping their plans close to the vest. Anticipating the next move by Amazon Web Services or Google gives the best industry analysts fits. What many may not understand is that this obfuscation is not a new tactic.
FUD, in its most widely understood current meaning, was employed by Gene Amdahl in 1975 to refer to sales tactics used by a dominant industry leader to play it safe with the dominant player’s offerings rather than taking a chance on a competitor. “FUD,” he said, “is the fear, uncertainty, and doubt that IBM sales people instill in the minds of potential customers who might be considering Amdahl products.”
In a country that happily sings the praises of its most successful citizens, Gene Amdahl ought to be remembered not only for his work with the IBM 701, 704, 709, 7030, 360, and Amdahl 470V/6, but also for his contributions at Trilogy Systems, Elxsi, Andor International, Commercial Data Services, and Xbridge. These latter companies have faded into computer history, notable only because they were “failures.”
About Amdahl’s legacy there is no FUD. He will be remembered not for wealth or notoriety, but for a lifetime of perseverance through success and failure, in a league with Thomas Edison and Mark Twain, to name two others who failed and sometimes succeeded.
Amdahl was interviewed by Tekla Perry @TeklaPerry for IEEE Spectrum in 1982, and she highlighted these Amdahl remarks in her remembrance in the same publication 33 years later:
“Engineering is satisfying people’s needs, and people’s needs aren’t all technical. Engineering is really a social undertaking.”
It’s something Edison might have said.