I couldn't find this information on Google, but why is the
visual representation of a database (in presentations and
graphical descriptions) always a can or a cylinder? Is it
because hard drives contain cylinders or is it because that
is what hard drives used to look like before the rectangular
ones appeared?

Thanks,
Josh

On Jun 14, 3:06*pm, howfie <yaday... (AT) yabloo (DOT) com> wrote:
Yes it harks back to storage being disc storage (hard drives).

Ed

howfie <yadayada (AT) yabloo (DOT) com> wrote:

Correct. See http://en.wikipedia.org/wiki/Early_I...orage#IBM_3330 for a
picture of such. And that is basically what you would see inside a modern hard
drive. Mind you that one is 200 Mb. I've seen others in 1979 which were mounted in
a machine slightly narrower than a washing machien and were 10 Mb.

Tony
--
Tony Toews, Microsoft Access MVP
Tony's Main MS Access pages - http://www.granite.ab.ca/accsmstr.htm
Tony's Microsoft Access Blog - http://msmvps.com/blogs/access/
Granite Fleet Manager http://www.granitefleet.com/

"Tony Toews [MVP]" <ttoews (AT) telusplanet (DOT) net> wrote

Also see the wikipedia article on drum memory.
http://en.wikipedia.org/wiki/Magnetic_drum

The picture of the drum in the article really looks like a can. Drums fell
out of favor when compared to discs, because moveable heads made for much
cheaper construction.
In addition, stacking disk on top of each other in cylinder fashion allowed
you to pack more media onto the same square footage of floor space. The
great timesharing computers of the 1964-1982 time frame all used discs of
various sizes and capacities, as did the giant database mainframes of the
era. The early computers in the 1950s tended to use drums for main memory,
before core memory gained widespread use. See the article on the IBM 650.
http://en.wikipedia.org/wiki/IBM_650

"Walter Mitty" <wamitty (AT) verizon (DOT) net> wrote in
news:wDpZl.1097$P5.777 (AT) nwrddc02 (DOT) gnilink.net:

sure before I said anything to my database students or add
it to my powerpoints :-).

Josh

howfie wrote:
I must say I thought it was one of the greatest questions I've ever seen
here. And half-thought it was a send-up until Walter replied! If this
is for real, maybe you should point out to the kids that the cylinders
were and are 'virtual', get them used to what that term can mean.


Once, I was given a 2321 (if I recall the model), sometimes called a
spaghetti picker. Real Buck Rogers concept, half disk and half tape
(with an astounding 500 msec access/latency time) but with serious
engineering, like a three-horsepower motor labelled "lubricate every
twenty years", no guff. Serious Nasa-style hydraulics, twenty gallon
reservoir and compressed nitrogen. I couldn't afford to hook it up. I
knew a really smart programmer but much younger who asked me what a
mainframe disk looked like. I told him the box that held seven or eight
of them was taller and longer than him so the 'Customer Engineer' (who
was a regular presence at all big IBM installations right up to about
twenty years ago) could get his head inside. Unfortunately it seems one
CE was decapitated which resulted in an EC, so-called "engineering
change".


When I thought again about his question afterwards I remembered how much
closer programming was to manual labour I was envious of that kid
because he had saved himself a lot of wasted years and I wished I had
been born later. On some smaller machines even in the 1960's, people
used tape drives for working storage to sort. These days, for myself, I
just use solid state stuff.


In those days a macho term among assembler programmers was
'bare-metal-programming', which the typical disks encouraged with the
'Count-Key-Data' instructions that were wired into their controllers.
Thankfully, modern disks have abstracted out that physical nonsense.
The same people considered it important to flip three thousand cards
without spilling them and looked down their noses at any object deck
that was thinner than three or four inches.


Then there's memory capacity. Did you know that even into the 1990's a
typical PARS/ACP flight reservation system for a medium-sized airline
(say 100-300 planes) occupied only about 300MB disk space? I have to
wonder how things would be now if today's main memory were available
then, a lot of decisions would have been different. Of course we know
what really happened to that memory - as a friend who was an IBM
salesman in the 1960's was told when some 512KB memory cpu's came out:
"your customers can't use it, we need most of it for OS/360'. In those
days the famous IBM motto was "THINK" and they used to give out desk
plaques to remind you to do that. I saw one that had been graffitee'd
to say "THINK OR THWIM".
..

"paul c" <toledobythesea (AT) oohay (DOT) ac> wrote

Don't get me started down memory lane.

Hi Paul,

Thanks for the info. With pictures like this,
http://www.columbia.edu/acis/history/cell2.jpg,
I guess it's easy to see why it's always a can or a cylinder.

Could you discuss a little more about your 'cylinders are virtual'
comment? I'm not really a hardware guy and this question is actually for
a simple introductory relational database course that I teach which
exclusively focuses on 'the query' and SQL. In the first lecture though,
I really like to talk about the history of databases and other little
tidbits of potentially useful (or useless) information.

Thanks!

paul c <toledobythesea (AT) oohay (DOT) ac> wrote in news:0IC_l.31406$Db2.24878
@edtnps83:

"howfie" <yadayada (AT) yabloo (DOT) com> wrote

I'm not sure what "cylinders are virtual" meant. I await Paul's response.

The term "cylinders" does come up repeatedly when discussing stacked disk
platters at the physical level.

When a spindle holds a stack of disk platters, there is one head for each
surface that's actually used. These heads generally move together as a
unit, in what's called a "seek" operation.
Once a seek is done, each head can read/write one track worth of data. If
you put all these tracks together, they make up a "cylinder". This is the
total amount of data that can be read or written before another seek has to
be performed. Typically, the disk unit reads or writes a small percentage
of the cylinder before performing the next seek, but that depnds on the
usage pattern.

When disk geometry is viewed this way, a track can be seen as the
intersection of a cylinder and a surface.

The file system of the operating system can be built so as to attempt to
optimize the clustering of data onto a single cylinder if data is
anticipated to be used at the same time. This can reduce the total amount
of time the disk unit spends seeking, and materially speed up the disk's
throughput. The files that contain database data are sometimes allocated to
tables in a way that attempts to anticipate usage patterns. This depends on
the interplay between the DBMS and the OS.

Walter Mitty wrote:
When they first made rotating disks, in the late 1950's I guess it was
(before this there were gizmos called drums), they resembled an LP
record in size and were stacked ten or twenty together with space
between each layer. Unlike audio records, the didn't have a spiral
groove. A servo mechanism of some sort held little inductive pickups on
very thin 'arms' which were held away from the surface by an air cushion
produced by the rotation of the disk. (I'm not an engineer so I may
have distorted some of the finer points of the mechanism.)


The electronics that interpreted and drove the pickup 'heads' had
discrete steps, as the head moved from the perimeter to the centre of
the disk, each step corresponded to one of several dozens or several
hundred (depending on the manufacturing precision and electrico-magnetic
sensitiivity of the pickup heads and the iron-oxide coating on the disk
platter's surface) of 'virtual' concentric rings. Unlike an audio
record, there were no physical grooves, rings or 'tracks' in the disk
material, which is why I referred to them as 'virtual'. One could
think of the ten or twelve or twenty tracks as a logical arrangement of
the horizontal platters on the vertical plane.


Also, the tiniest speck of dust in the assembly could wreck it because
it interferred with the air flow and could cause the delicate head arms
to bounce and on their return to scratch the oxide surface. This was
called a head crash. Head crashes could propagate, because the head
assembly could go out of adjustment after a bounce. When disks were
miniturized for personal computers and certain economies eliminated
certain features such as head retraction on power loss, it was some
years before manufacturers added circuits to retract the heads when
power was shut off and many small disk drives went into the garbage as a
resullt.. I remember after one big expensive head crash propagation on
a bank of removeable disks, cigarette smoking was forbidden in the
machine room.


But the big implication of all this is the time required for head
movement. A smaller implication was the 'rotational delay' or 'latency'
for the disk to rotate. Unlike other computer developments, disk
manufacture hit physical performance limits sooner. Even the most
expensive mainframe diisks never got better than about 100 head
movements per second and the time to read a track was about a tenth of
that. By the mid 1980's it was unusual for a real system (as opposed
to a staged TPF setup) to beat about 100 'trivial'/'TPF' transactions
per second and it's not hard to see why rotating disks with moveable
heads were one big limiting factor. .I believe that today those numbers
have no improved much although the latency delays are sometimes
disguised by the use of special memory in disk controllers which is
sometimes called cache.


On some very custom multi-user systems, such as the PARS/ACF airline
systems, one technique to deal with seek time and rotational delay was
redundancy, replicating the same records across multiple disks. The
disk controllers were themselves programmable, so a seek order could be
sent by the central cpu to the controller, identifying a particular
physical disk, if the controller didn't respond immediately, presumably
because another 'task' was accessing that disk, the driving program had
tables to direct the seek order to another disk which cloned the first
one. Those disks were quite small in capacity, the 2311's of the early
1960's were about 10 Mega-bytes in capacity, the 2314's of the
mid-1960's bumped that to about 30 Mbytes (which was rather convenient,
since that was the capacity of a typical 2400 foot 9 track tape reel)
and by the mid-1980's a typical disk was about 300 Mbytes, so a bank
might approach a couple of Giga-bytes in capacity. They were configured
in bank of eight or so, purchase prices had fallen to about $100 grand
per bank, plus a grand or two per month for maintainance/warranty.


Looks to me that it's only a matter of time before Solid State Disks
take over.