Fault Tolerance in Tandem Computer Systems (1986) [pdf]

I'm not sure if you're being serious, or delivering scathing sarcasm.

> "Processes are the software analog of processors." Yes of course.

Maybe this analogy worked in 1986 when hardware was a lot less reliable, but I don't think it goes very far these days: processes die all the time (normal termination, crash, get killed). When was the last time a processor or core died on you? In fact, according to this analogy most of us are running MS-DOS equivalent of systems since if your processor or core dies, your machine dies. And I don't see this changing any time soon.

Things are different when "mission-critical" is the highest priority.

Before PC's really got popular, I went out to a petroleum pipeline installation one time where they were just unboxing one of these Tandem rigs.

They were going to use it to further automate, control, and account for transfers like some of the other oil companies were doing with their mainframes. As part of an expected technology advance at the time.

Here it was not just a CRT terminal and a printer, but the whole thing right there in a fairly hazardous location in the blockhouse office where contractors would do their hand calculations.

I was there to take readings on the mechanical totalizers, especially on the piping section we had independently calibrated, and bring samples back to the lab for precision viscosity and density determination to more decimal places than available elsewhere. Along with all kinds of other routine and research parameters.

Turns out I was the pioneer in digital densitometry among the multinational contractors. That's another story altogether but within a decade they all had it and I was in more demand after the niche had grown than it was when I owned the niche. People still never want me to stop.

Anyway, I had a pretty good handle on floating-point error and was doing my part to reign it in with improvements in physical measurement.

It didn't take long to realize that my Atari would be basically capable of handling all of the things they were going to use the Tandem for.

The shortcomings would be the redundancy/reliability and Atari just couldn't count that high :)

When you're moving large numbers of barrels the numbers go through the roof when you convert to liters or even worse, some currencies.

If the figures didn't agree very well with manual calculation using 16-digit calculators, some big shot may very well hit the roof.

I would have had to hook two Ataris together and try to get more precision somehow at the same time as try some redundant reliability. Never did.

Although within a couple years I did hook up two TRS-80s together and they were quite adversarial . . .

I worked for them briefly before University, I was essentially the sole report for a handful of senior people who were being dropped in to deliver "Project Genesis". They all told me this was the best place I'd ever work and so far (30+ years later) they're right.

That's also the first place I had something resembling routine access to the Internet, and coincidentally Tim's stupid hypermedia technology (the World Wide Web) was just taking off too so I got to see that brief period when Netscape Navigator is exciting and new.

Apparently Jim Treybig (Tandem founder) was big on people. He came from HP, and brought over the HP (at the time) way. I think a lot of employees even got 6 week paid sabbaticals every 4 years.

I started there as an intern in 1982, and it was my first job out of college. Tandem was Kleiner Perkins's breakthrough: it was the fastest-growing public company in America, and the growth was anticipated and carefully planned for. There were brilliant people everywhere: Jim Gray's office was on the other side of the wall from ours, and Tom Van Vleck, who wrote the first email program for CTSS at MIT, was down the hall. It was the closest to a Manhattan Project I had in my career, and pure luck that I was there.

I can't find the link, but I read a similar story about Tandem - removing parts in front of customers and it kept going.

The funny bit was that it was configured to dial-up to the Global Support Center, when something went wrong, so after all of these hardware fail/component removed errors happened, a technician was sent over with replacement parts that weren't needed.

Yes. Proprietary CISC chips->MIPS->Itanium and ported to x86-64 around 10 years ago. Apparently is or will soon be offered "in the cloud" as a "NonStop as a Service".

It seems like overkill for just FTP. Although I believe that Tandem also was (is?) fairly common in running telco stuff, plus billing.

Telcos I've touched have used Tandems for E911, SMS processing, Device authorization, and automated tariff negotiation between CLEC/ILEC, but FTP server is a new one!

They still kept their software redundancy, but when Tandem switched from their own custom CPUs to MIPS, they used pairs of lock-stepped CPUs like Stratus.

When they still had their own multi-board proprietary processors, Tandem also used lock-stepped microprocessors in their maintenance and diagnostic subsytem and some of their communications controllers.

I also gather Stratus is still around running on x86-64.