Orchid XP v8 <voi### [at] devnull> wrote:
> >>>> It still somewhat blows my mind that you could do anything useful with
> >>>> so little memory. Presumably for processing large datasets, most of
> >>>> the data at any one time would be in secondary storage?
> >>> Large datasets then were also very tiny compared to large datasets of
> >>> today. :)
> >> Sure. But 1MB is such a tiny amount of memory, it could only hold a few
> >> thousand records (depending on their size). It would almost be faster to
> >> process them by hand then go to all the trouble of punching cards and
> >> feeding them through a computer. So it must have been possible to
> >> process larger datasets than that somehow.
> >
> > You never read the entire dataset into memory. You process it a record at a
> > time.
>
> Right. That's what I figured.
>
> > The only limit on file size is the media, not memory.  There is no difference in
> > memory consumption between processing 10 records or 10 million records.
>
> If you're trying to, say, sort data into ascending order, how do you do
> that? A bubble sort? (Requires two records in memory at once - but, more
> importantly, requires rewritable secondary storage.)

Right, you read as many records in as you have memory for, and use rewritable
secondary storage for work files.  Preferably disk, but if your not in a hurry
mag tape will work.  When VIPs would come around for a tour of our datacenter,
we would always start up a sort using tape for work files.  It was fun to watch
the sort program read, read backwards, rewind, and forward space the tapes.
Slow but fun to watch.

>
> --
> http://blog.orphi.me.uk/
> http://www.zazzle.com/MathematicalOrchid*

Isaac

Post a reply to this message