Information Technology Reference
In-Depth Information
Lest you think that a concern regarding complexity is only the
oretical in nature, let's consider a few illustrations related to the
space shuttle. We have observed that computational and logical
complexity can require considerable coordination and oversight,
raise extensive hardware demands, and provide many opportuni
ties for error. The following notes apply these observations to ac
tual circumstances:
In 1989, the onboard software for the space shuttle achieved
what was called an “exemplary” error rate of only 0.1 errors per
thousand lines. To achieve this “exemplary” level of correctness,
NASA spent about $500 million to develop the overall soft
ware—about $1000 per line. Even with the 0.1 error rate, how
ever, a program of 500,000 lines still contains about 50 errors.
In 1981, one of these errors caused the first shuttle orbital flight
to be delayed. Part of the reason for the cost of the shuttle soft
ware was that between 1981 and 1985, reports indicated that
some 4000 changes were made to the software. This would
translate to roughly 1000 per year, or perhaps 3 per day, includ
ing holidays and weekends. Each change cost NASA money.
(From p. 112 of
The Limits of Computing
. From Edward J.
Joyce, “Is ErrorFree Software Achievable?”
Datamation
,
February 15, 1989, pp. 53, 56.)
“The Space Shuttle Ground Processing System, with over
500,000 lines of code, is one of the largest realtime systems ever
developed. The stable release version underwent 2177 hours of
simulation testing and . . . . 280 hours of actual use during the
third shuttle mission.” Errors detected during both testing and
the actual mission were classified as “critical,” “major,” or “mi
nor,” and the occurrences of these errors is shown here:
Critical
Major
Minor
Testing
3
76
128
Mission
1
3
20
Thus, although many errors were caught during extensive
testing and simulation, a fair number of problems were not
encountered until the mission itself.
(From p. 108 of
The Limits of Computing
. From Mira,
“Software Reliability Analysis,”
IBM System Journal
, Volume
22, Number 3, 1983. Quoted in
Software Engineering Notes
,
Volume 11, Volume 5, October 1989, p. 16.)
