Origin of row/column based database software

From: +ACI-McFadden, Mike+ACI- <mmcfadden_at_cmh.edu>
Date: Wed Aug 18 14:48:30 2004

Bob asks about row/column based database software.
+ADw-snip+AD4-

Lots of mainframe database software existed but it was expensive and
took lots of CPU power and disk space. From my experience it all
depended if you purchased software or wrote your own.

At the University of Missouri in 1976 we made a database of food
poisoning victims.

I used one punch card for each patient which was one row.
Each column on the card was for each symptom or food consumed.
Columns 1-16 was usually last name
Column 17 might be Saturday morning vomiting
Column 18 might be Saturday afternoon vomiting
Column 19 might be Saturday evening vomiting

Column 20 might be Saturday morning diarrhea
Column 21 might be Saturday afternoon diarrhea
Column 22 might be Saturday evening diarrhea

Column 23 might be Saturday morning beer
Column 24 might be Saturday afternoon beer
Column 25 might be Saturday evening beer

We would usually look at a 3 day interval.
It usually takes 18 hours for food poisoning symptoms to start.

Some people are +ACI-immune/tough+ACI- and don't get sick.
Some people have symptoms from other sources+ADs- beer, other foods

If you listed a data deck of cards then row columns were how the data
was listed. We looked to trends and correlations. On one outbreak the
highest correlation was the potato chips but the chicken salad was the
actual cause. We used to get a lot of false positives for vomiting due
to the amount of beer consumed by students.

We also used SPSS on IBM 370/168 to analyze data also printed 2 X 2
tables of cross-correlations. The data was punched on an IBM 026 card
punch, later we got a 029.

We moved the analysis into FORTRAN programs on PDP-11/20 using DOS.

Later in 1979 I used BMDP and Datatrieve on a PDP-11/35 to computer
percentages of bacteria that consumed certain sugars.

The table looked like
               S L M C
               U A A A
               C C N S
               R T I I
               O O T N
               S S O E
               E E L

e.coli 75 5 5 10
s.aureus 5 1 1 95
P.aeruginose 10 50 50 1

The PDP-11/35 was later upgraded to a PDP-11/70. BMDP required overlays
and lots of common areas to fit in the limited memory we had. We had
320K words of memory for 6 users, the RSX-11M OS and all external device
pages. All of our data was on RK05's. The system was on one RK05, the
programs on another, one for scratch and another for data. We changed
the data packs a lot to load more data. Our first RP04 was setup by us
to look like lots of RK05's since we stored data and analyzed in those
increments. For the clinical machines the data was stored in a PROM
board on a PDP-11/04 that analyzed bacterial growth and identified the
bacteria. If I remember correctly the table was about 30 rows by 26
columns.

Minicomputers were an attempt to move out of the computer center. We
weren't really classed as computer programmers but R+ACY-D engineers. We
stored lots of data in RAD50 and bit flags to save space.
Received on Wed Aug 18 2004 - 14:48:30 BST