IO call reliability

About fifteen years ago, I read an article that suggested one CPU
error per 1000 hours of computation. Nowadays CPUs are faster, but
they are also tested more extensively, so the value is probably
still as pertinent now as it was then.

This is why systems like the Space Shuttle have triple-redundant (or
is it quintuple-redundant?) computer systems, which all run in
parallel and vote on the result. Not only are there 5 computers, but
they're of two different designs, so they're unlikely to suffer from
a common design defect that causes them all to get the same wrong answer.

Having 5 CPUs that each have one error every 1,000 errors produces a
system with one error every 1e9 hours, which is over 100,000 years.
The homo sapiens species hasn't been around for that long.

The software in the spaceshuttle is well designed and I think there's
only one copy.

No. It is as M. Margolin said. (Although it is the software, not the
hardware, that is designed indepently. All of the GPCs are the same.)
There are two MMUs in the Shuttle, MMU 1 and MMU 2. They both contain
copies of PASS, and any of the five GPCs can be IPLed from either MMU.
In normal operation for critical mission phases, four of the five GPCs
run PASS, and the fifth runs BFS, which is a separate software package
to PASS, written by a different company to the writer of PASS.

In sci.space.tech Jonathan de Boyne Pollard wrote:
[[about the onboard computer systems in the US space shuttle]]
No. It is as M. Margolin said. (Although it is the software, not the
hardware, that is designed indepently. All of the GPCs are the same.)
There are two MMUs in the Shuttle, MMU 1 and MMU 2. They both contain
copies of PASS, and any of the five GPCs can be IPLed from either MMU.
In normal operation for critical mission phases, four of the five GPCs
run PASS, and the fifth runs BFS, which is a separate software package
to PASS, written by a different company to the writer of PASS.

There's an extended interview with some of the key software architects,
in which they discuss the software design (and also a bit about the
hardware), in
Communications of the ACM
Volume 27, number 9 (Sept 1984)
pages 872-900
this is online at
http://www.klabs.org/DEI/Processor/s...ter_system.pdf

--
-- "Jonathan Thornburg [remove -animal to reply]"
Dept of Astronomy, Indiana University, Bloomington, Indiana, USA
"Space travel is utter bilge" -- common misquote of UK Astronomer Royal
Richard Woolley's remarks of 1956
"All this writing about space travel is utter bilge. To go to the
moon would cost as much as a major war." -- what he actually said