I am attempting to make sense of all the claims and counterclaims of y2k.Would anyone answer the following for me?
1Does anyone claim that residual failure does not exist?
2Can anyone prove that it occurs less than 1% of the time?
after any major maintenence change to a system (which Y2K most
certainly is) there is always a residual rate of failure as a result of the changes
themselves, even when the changes are properly "tested". The failures manifest
themselves when the system is placed back into the real world of "production", as
opposed to the artificial world of "testing". They happen because maintenence
programmers customarily test only the immediate effects of their changes. There is
neither the time nor the money nor often even the ability to test the entire
consequences of a particular change to a system. The residual failures typically arise
elsewhere in the system, at some point unrelated to the change itself and completely
unanticipated by the programmer.
This last is why residual failures are so hard to identify and correct. Often, we can't
even tell for certain whether a particular failure really is the result of a recent system
change or not. In turn, this is why a good system administrator would never return
two or more systems to "production" at the same time. Not only is the risk of failure
almost doubled, but there is also a small chance of both systems failing
simultaneously. For Y2K, the problem is greatly compounded by the fact that,
essentially, we will be placing all of our corrected systems back into "production" at
roughly the same time. We can even calculate the magnitude of the residual failures,
to a first approximation.
The actual rate of residual failure depends on a number of factors, but mostly on
the size of the system and the scope of the changes. Under average conditions,
modest changes to a moderately sized system, the rate would be about 7%. The
scope of Y2K changes is, of course, much more extensive than this and many of
the systems are extremely large, so the residual failure rate is also likely to be higher.
Nevertheless, for the sake of argument, let us again assume an overly optimistic
residual failure rate of only 5% for Y2K related changes. But this is only for one
system. For a business with multiple systems (which they all have) the chance of a
system failure can be computed as:1-(1-f)**n, where "f" is the failure rate and "n" is
the number of systems.
An average small business would have perhaps 5 systems so, assuming a residual
rate of 5%, they have about a 23% chance of at least one system failure
(1-(1-.05)**5 = 0.226). A medium size business would typically have about 25
systems and, therefore, a 72% chance of a failure (1-(1-.05)**25 = 0.723). A large
business with 100 or more systems would have a 99% chance of a failure
(1-(1-.05)**100 = 0.994). This is EVEN IF ALL OF THE SYSTEMS ARE
FIXED! Of course, many of these failures will be relatively easy to fix, but others
will require an effort beyond the capabilities of the business and they will not be
fixed before the business itself fails (this is particularly true for small and medium
businesses using packaged software). In addition, the great majority of these failures
will have at least some domino effect on related customers and vendors. To make it
even worse, virtually everybody will be facing these problems at about the same
time, leading to a chaos in which actually fixing the problems becomes almost
impossible. At the very minimum this will lead to an economic disaster, JUST
FROM THE ACT OF FIXING THE SYSTEMS THEMSELVES, without even
taking into account the effect of the unfixed systems, of embedded systems or of an
already declining global economy.
In reality, of course, the situation is much worse than this, and the residual failure
rate will be much, much higher. Just how much worse is anybody's guess since we
have, as yet, insufficient historical data of actual Y2K failures. One thing I can state,
categorically, is that a "bump in the road" is not even on the scale of possibility. As
we have seen above, the best case end of the scale really begins with a global
economic disaster and even then assumes that all systems are fixed on time and that
there are no outside factors such as a global recession. Clearly this, too, is an
untenable position.
The above was written by "Infomagic" and released by Cory Hamasaki in
his weather report 103. |
