--------- Forwarded message ----------
Date: Thu, 16 Apr 1998 15:14:06 -0400
From: [EMAIL PROTECTED]
Reply-To: [EMAIL PROTECTED]
To: [EMAIL PROTECTED]
Subject: Year 2000
A few months ago, there were many notes regarding the impending doom
associated with the possible Year 2000 computer glitches. I was the
"naysayer", saying it was "hogwash". Having nearly completed another
conversion/fix I must change part of my opinion since my opinion is now
factually based and have seen a possible outcome of not addressing the
problem. I still however, having experienced what it takes to investigate
and fix such problems, maintain that the process is relatively easy
provided those working on it "know what they are doing". The part of my
opinion that is changed is the possible disaster that could result. The
only thing we fixed was that people wouldn't get free utilities because of
an inability to 'post' customer bills . Now apply this to the angry
ex-steelworkers and laid off union people in CB, not receiving their
welfare and UIC cheque's....I'll face the New York crowd anyday instead of
this mob!
We have just finished a Y2K project. Myself and 2 other gentlemen have
finished our investigations. We looked at 3 legacy systems, a 4 terabyte
warehouse, 14 Datamarts (2 gig +), 54 applications and all the copybooks,
loadscripts, etc associated with the DBMS's. The legacy systems had 37 non
compliant (NC) date columns, the warehouse has 16 NC columns, and the
marts had 103 NC columns (duplicates). The applications use views instead
of the base tables. The applications run against the Teradata warehouse and
the Oracle Marts. Any date column defined in the Teradata warehouse as
"CHAR" that was non compliant resulted in the applications having to accept
this default definition thus making some date attributes in the
applications non compliant. The Oracle date columns are defined as
"Timestamp" allowing the application programmers to format the dates making
them compliant. Essentially the legacy system dates had to be redefined,
the warehouse definitions changed to Date versus Char , and the
applications that used teradata had to be updated. The applications in
question were:
VB frontends, with Crystal reports
Business Objects
Andyne GQL (kingston, Ont)
SPSS
The total time to do the investigation was 420 man hours. It took us
another 300 hours to make the code changes. Presently we have built 2 E3000
sun servers, 1 E 6000, and 2 Compag Proliant 3000's, have use of an E10000
and using 14 Dell Optiplex GXA pentium II test boxes. We have just finished
resetting all the system and operating times, have loaded a subset of
current data as well as test data. To set up this test environment has
taken us another 700 hours. Tomorrow is Dec 31, 1999 and Monday April 20,
1998 will appear to be January 3, 2000. Should everything go as planned we
will have "demystified" the Year 2000 mystery, at least for one company.
The next step will be to load more test data , reset the clocks, etc and
see if the "leap year" is handled as well. Our wages were miniscule in
comparison to the hardware expenditures....$374,000. Essentially the full
cost was a $1/2 million. For those of you with concerns and wondering when
the gov't of NS is going to address this problem, you "best dust off your
lobbying skills" because unless the gov't starts seriously addressing this
problem....welfare cheque's, grant cheque's, provincial employee cheque's,
etc, ain't going to arrive. The systems controlled by the NS gov't are
probably a heck of a lot larger than what we looked at and would
"guesstimate" that you are looking at a task approximately 20 times
greater, if not more. By the time we have tested and the actual production
changes are made, total time passed will have been 6-7 months. The NS gov't
if they haven't started yet, have 19 months to fix what I would guess to be
some monstrous systems.
Those of you that personally know the gov't ministers, premier, top
bureaucrats, etc may want to try to convince them of the need to allot
money to investigating these problems. There may very well be no problems
but I'd hate to find out on Jan 1, 2000 that there are indeed problems.
Statistics in the past have indicated that each computer replaced 20
employees....with year 2000 the gov't will have to find 20 people and train
them to replace each computer....here's one solution for the high
unemployment.
The point I'm making is that many seem to be "running scared" of the Year
2000, predicting Armageddon, etc. and investigating this stuff is
relatively easy. The tough part is getting started...someone/dept willing
to "cough up the $$$". I will keep you posted as to our test results next
week and as to the results of the "leap year" test. If all our testing goes
well we'll have disappointed many folks 'cause their January year 2000
utility bills will arrive as scheduled. Our total time from Start to setup
of the Test environment....4 months. Our estimated test period...2 months.
Essentially 6 months to find every non-compliant
column/field/attribute/row/ tuple, etc. Once the testing period is
complete, I'll let you know if "I'm blowing smoke" or if this is as easy
(although expensive) as I say it is.
Greg
