I came of age in the 1980s and worked on various minis and mainframes in high school and college (mostly DEC but also Univac and IBM). Even though I was an enthusiast then (as now), I never really knew the total amount of disk storage that the computers had. I know this can vary significantly by industry and location, but what would a typical disk allocation have been at a large university, on, say, a Decsystem-20 or IBM 4381 in the mid to late 80s (yeah, I know the DEC would have been considered obsolete by then, but there were still quite a few extant installations)? Assume it was their main computer and was used for student assignments as well as most administrative tasks. Not looking for exact numbers, more like an order of magnitude.
10
-
8VAX were very common in the late 1980s.– RonJohnCommented Sep 6 at 23:22
-
4I remember standing with a dozen or more geeks in the hall outside the C-MU computing center's main machine room, hoping to catch sight of a DEC RP20 that was delivered that day. Everyone wanted to be able to say that they had personally laid eyes on a one gigabyte hard drive. It felt like we were witnessing the dawn of a new age. No other threshold after that felt quite the same as that first gig.– Solomon SlowCommented Sep 7 at 1:26
-
1@RonJohn I've probably used more Vaxen than any other mini. But it was never the school's main machine.– SwechslerCommented Sep 7 at 2:03
-
2@SolomonSlow A geek never forgets his first gig.– Miss UnderstandsCommented Sep 7 at 7:04
-
@SolomonSlow - the RP20 was a two-spindle device, so actually two half-gig discs.– daveCommented Sep 7 at 18:10
|
Show 5 more comments
11 Answers
Marburg University in Germany has put the complete configuration of their IBM 4381 as of Dec 31, 1987 online. They list five "magnetic disk units" (Magnetplatten-Einheiten) of 2.52 GB each, totaling some 12.5 GB.
(This is part of the university computer center's quite extensive history pages, which are well worth reading if you're interested in that sort of thing, i.e.: the development of a university computer center from 1963 to 2013, with lots of technical and bureaucratic detail.)
-
Nice example of a low end CPU (43xx were low end, 4381 was IIRC less than 1 MIPS) with extreme disk sizes - then again, 1987 is quite late anyway.– RaffzahnCommented Sep 9 at 20:12
-
4The OP specifically asked about an "IBM 4381 in the mid to late 80s" at "a large university". Couldn't get closer than that. :-) Anyway, that's universities for you: lot's of users requiring lots of space, while getting only a small amount of computer time each. Commented Sep 9 at 20:44
-
And, btw, if you look into the report it says they only just installed the 4381 in 1987, replacing a 4361. Commented Sep 9 at 20:47
-
Which seems like a huge jump in computing power. 4361 was IIRC .5 to 1.5 MIPS, while 4381 was 2-5 (10 for dual CPU) according to IBM sales brochure.– RaffzahnCommented Sep 10 at 7:20
-
On the one hand, yes. On the other hand (and ignoring that MIPS isn't really that good measure of computing power across different architectures, and that you (should) buy a mainframe for throughput, not computing power in the first place): 3 MIPS and 16 MB of RAM is what a Sun 3/60 would have put on your desk at that time, so for a mainframe installation, it's a bit underwhelming. Commented Sep 10 at 18:12
In the early 80s, I was aware of a Burroughs mainframe running a building society with of the order of 3GB of back-end storage on a half dozen disk packs.
The installation was due to be upgraded from a "Medium System" (COBOL-based) to a "Large System" (ALGOL-based) so it's reasonable to assume that storage was increased dramatically at the same time.
It's also important to appreciate that a lot of activity was still basically batch-mode tape-to-tape, so an enterprise's total storage capacity might be substantially more than its disc capacity.
But I think it's reasonable to assume that site disc storage was probably less than 10 GB in most cases.
-
3I worked on IBM 360/370 systems throughout the '80s and never really thought about total disk capacity. What we did have to do a lot of was strategic and tactical planning related to tape libraries. I remember being audited because I was using "too much" physical space. After the audit it was determined I was using too little. We had tape and disk allotments for storage and memory allotments for our programs and I spent considerable time making sure everything fit in the available space. Most storage was off-line and disk was just for a few real-time databases and system storage.– Dave DCommented Sep 7 at 16:30
-
Our 1980s Univac 1100 was managed similarly. A lot of disk pack swapping went on. Users usually worked with round reel tapes, and, towards the end, tape cartridges. Commented Sep 8 at 22:12
-
@OrganicMarble That brings something else up. The original use of reel-to-reel tapes was record-oriented processing, but by the early 1980s some people were referring to "tape streamers" which (possibly) were oriented more towards high-speed sump/restore to secondary storage (i.e. disc). I /suspect/ that the move to cartridge storage effectively marked the point at which record-oriented tape processing was effectively obsolete. Commented Sep 9 at 7:39
-
Our use case was building a modified executable program which contained our updated software (this was a shuttle training simulator). This "training load" got stored on a round reel tape if it compiled and built cleanly. When we went in to use the simulator to test our changes, we gave the operator our tape #. They loaded it into the mainframe and "brought up the load" for us to use. When we changed to tape cartridges, the process for me as a user was essentially the same. Commented Sep 9 at 11:25
Variation in Use Case
The "main problem" for any answer might be that being a mainframe already spans several magnitudes. For IBM/360ish machines that already started with a 1:20 spread between small and large. By 1980 IBM and most compatible manufacturers offered a 1:100 up to 1:500 spread between smallest and largest CPU. All of those are mainframes used for tasks and businesses of different size. There was never any overall 'typical' installation; at best, you could say that something was typical for a certain company/department size.
And it worked the very same way for disk storage giving an even greater spread. It's all about the task. A high-end system with just 4-6 drives used for online transaction handling was as common as some mid-range with huge disk farms.
It's similar to how today's 'cloud' systems are available in more storage- or more communication- or computation-orientated setups. Mainframes have been there 40 years before: configuration is all about the task, and there is no typical task.
As an example, in 1980 we ran an online ERP system handling 300 concurrent users (*1) in a department managing a workforce of ~1200. A setup of two CPU at 0.9 MIPS (one hot, one standby) with 14 disk drives, 5 tapes and a lot of communication/terminal equipment. Of those 14 drives only 4, with a total of about 600 MiB, were used for production. The remaining drives were usually empty, used for daily copy (backup) and/or assigned to the second CPU to act as development system.
So that's an upper-end ERP (going by users and transactions) with just 0.6 GiB assigned.
The same goes of course for universities where different faculties had different use cases/requirements. A business faculty would not need much space (after all, most of their stuff is still today organized in 80 character punch card records :)) while in scientific more - but especially, fast-access - disk space would be needed.
With such a great range, naming a 'typical' scenario can't be meaninfully done. One can only give examples - like Michael Graf's Marburg in 1988 - but they will always only mark a single otherwise unrelated data point.
Variation Over Time
Equally problematic is asking for the 1980s as a whole, as that was a time where affordable computing power and storage space exploded. Thus variation was not only 1:100 between small and large installations, but in addition a spread of at least 1:20 between a system installed in 1980 vs. one bought in 1989.
The low/mid range 43xx is a good example as the 1980s was exactly their era: introduced in 1979 with the 4331 (*2), withdrawn in 1992 with the 4381 (last top model introduced in 1988).
- A base 1980 4331 offered:
- 0.2 to 0.5 MIPS (1 MHz clock),
- maxed out at 8 MiB RAM and
- used 60 MiB Fixed disks (max 16 thereof),
- while a well fitted 1988 4881 had:
- Single CPU at 5.5 MIPS, or
- Dual CPU at 10 MIPS (20 MHz clock)
- 64 MiB RAM
- use 3380 drives with 7.5 GiB each (IIRC up to 2x4)
So the low end 43xx grew in those 10 years:
- 20 fold in computing power and
- 120 fold in storage per drive.
Thus even the smallest 1989 mainframe storage setup (only one disk) will outclass a maxed out 1980 one 10 times. Even when ignoring that universities were able to buy many times larger systems, the spread over time alone makes it impossible to define 'typical' over such a long time span as well.
*1 - 300 concurrent users would be a huge installation for a university in the 1980s. At least during the first half.
*2 - First Computer to use 64 Ki RAM chips.
-
I've already gotten several good answers. It's looking like single to double digit GB capacity would have been pretty common in larger installations. Commented Sep 8 at 19:45
-
@Swechsler Except there hasn't been no 'common' or 'average' size. Being an FSE in 1980 I witnessed low end 4331 using 12-16 drives with a total of 8+ GiB, while many High end 308x barely featured 2 or 3 GiB. Not to mention older installations using less new hardware but lots thereof - like 40+ disk drives 288 MiB each for a single 1 MIPS machine (mid range at the time). There was no generic average. If at all, one could define an average according to business type. So Banks usually had huge disk farms, while Accounting had way less but faster CPUs.– RaffzahnCommented Sep 9 at 20:10
-
-
@Swechsler The very same goes for universities. Especially major having usually more than one mainframe. It all came down to college and task. A business faculty would usually have rather less disks (and space), whereas research orientated afford larger ones. This even goes for disk type and/or backup, as the later also need to exchange large data sets between sessions, so early on smaller but removable stacks were used instead of hard disks, later even robot archives were added. Fast exchange of data was baseline for research - as systems were shared.– RaffzahnCommented Sep 10 at 7:10
-
speaking of 300 users, I recall a 4381 mainframe and getting a response from it seemed to take ages during uni registration, and there were perhaps 300 or more users at this peak time.– MichaelCommented Sep 10 at 14:29
Well, no-one really thought of a -20 as a 'mainframe' - that was an IBM word.
The late 1970s, early 1980s disk in DEC was an RP06 drive, 176MB. Maybe two or three of those. These were removable-pack drives, common at the time; for the DEC-20 they connected via an RH20 Massbus controller.
The 1980s saw an expansion in capacity. The next large removable disk was the RM05, 256MB capacity. A confusingly-named RP07 (we all supposed RP meant removable-pack) was a 512MB non-removable disk.
This web page says Columbia's DEC-20s had 4 RP06s, of which 1 was available for private packs.
On PDP-11s we used to count disk size in 'blocks' of 512 bytes. DEC-20 people used 'pages' which might have been 256 36-bit words.
-
1FWIW, more than one online resource (including Wikipedia...yeah, I know, Wikipedia) lists the DEC-20 as a mainframe. AFAIK, big iron from other manufacturers (such as Burroughs and DG) were also called mainframes. I'm not claiming to be an expert, but I don't think it was strictly an IBM term (at least for casual use; it's possible those other brands didn't officially use the word) Commented Sep 7 at 14:09
-
1I suppose the DEC-20 was called a mainframe in periodicals. I have a bunch of old "Radio-Elektronica" magazines (Dutch) from the 1970's, and in one of them the DEC-20 is introduced, with the designation "mainframe".– chthonCommented Sep 7 at 15:27
-
1I used TOPS-10 at university, and later worked at DEC for decades, though not on 36-bit machines. No-one called the DEC-10 or DEC-20 a mainframe in my experience. In fact I worked for a time in connecting DEC machines to mainframes, by which we meant IBM S/370 and its offspring.– daveCommented Sep 7 at 16:23
-
2Digital themselves called their high-end 36-bit machines "mainframes" – see their 1981 "DECsystem-10 Technical Summary" (PDF page 63) which says "The KS10 has changed mainframe computing. People constrained by budgets to settle for something less than a mainframe can now install a KS10-based DECSYSTEM-2020..." bitsavers.org/pdf/dec/pdp10/TOPS10_softwareNotebooks/vol01/… Commented Sep 8 at 3:06
-
1My recollection is that PDP-8 and PDP-11 were considered minicomputers, while PDP-10 and PDP-20 were mainframes.– BarmarCommented Sep 8 at 3:16
This is just past the cut-off of "the 1980's" as this was from an installation dated around the opening of the new computer lab system as November 26, 1990. This was for an IBM 3090-170J but it had just replaced a 4381 system and presumably had a bit more storage so might be an interesting data point.
According to the documentation, the 3380 disks were 16 inch diameter aluminum platters, with a volume of platters containing 630MB or 1.26GB each, and contained a total of 23.4GB, while the 3390 disks totaled 45GB. It's possible that maybe the 3380 represent the original storage of the older system while the 3390 disks are the upgraded amount, but that is just speculation on my part.
This is for a mid-west state university campus.
A typical Soviet BESM-6 mainframe installation (single processor) in the mid to late 80s would likely have up to 8 units of 7.25 Mb IBM 2311-compatible drives, and up to 8 units of 29 Mb IBM 2312-compatible ones, totaling 290 Mb, being in the process of switching from the former to the latter. There existed a multiplexing controller allowing to have up to 64 drives (thus, theoretically, up to 1.8 Gb of storage), as it is supported by the OS, but whether it was used to its full capacity at the installation where it was designed, is unknown.
In 1984-85 my school had a VAX 11-780(?), dedicated to the freshman class, which was 149 students, who each had a disk quota of 200 blocks. That's a mere 15 MB of storage for the frosh. Obviously, the system also had the OS, software packages, data that didn't count against individual quotas (like email?), and a page file. So I'd give an order of magnitude estimate of 100 MB total.
The upper classes, faculty, and administration had accounts on a VAX 8600 with at least an order of magnitude more storage. I'd guess 4 GB. There were other minis around, especially in the computer science department, but the 8600 was the workhorse for the school.
Our university had two mainframes, one for the administration, and one for computer science. They each had two fastrand II drum units that stored 88 megabytes each.
My wristwatch has more "core" memory than the Univac 1108 had mass storage.
Its cpu is 900 times faster.
-
Surely true, but the 1108 was a 1960s system -- even if a cash-strapped university was still running an 1108 service decades later -- which is not the timeframe of the OP's question. But if we're boasting "mine is smaller than yours", the undergrad multiaccess system at my university had a single 4 million word (48 bits/word, so 24MB in today's money) disc store: KDF9, a 1960s design still in use in the 1970s.– daveCommented Sep 8 at 16:51
-
@dave the OP seems willing to accept answers about aged but extant installations even those that are obsolete, though his example of a Decsystem-20 seems to predate the 1980s by only three years so possibly his definition of obsolete differs from that of you or I.– TommyCommented Sep 9 at 16:50
-
2When I was born, there was 48 KB of memory in the world. 12KB of that was in ENIAC. The first computer I could actually be in the room with was an HP in the NPL in 1973. That had 16 16-bit values as memory (32 bytes). Everything else was read off and on the drum. Commented Sep 12 at 13:16
My university had a CDC Cyber 170 mainframe for a short time. I started in Fall of 1984, and the Freshmen programming classes were done on it. The university acquired it in 1981 and it was replaced with a high-end VAX around 1987. Here's a picture from the 'User's Guide for the Cyber 170 Computer', 12th edition, dated Sept. 1982, published by the university. We had to buy a copy and I still have mine. The next pages go on to describe Time-sharing Terminals, Remote Batch Terminals, Dial-up Telephone Lines, the NOS operating system, and the various compilers (Fortran 4, Fortran 5, COBOL, BASIC) available.
The CDC 'washing machine' drive was common about 1985. This held 300 MB in a demountable platter set in a blue plastic case, that pulled out upwards. We typically ran with two drives - one for the system disc and one for the images and data. Backups were run overnight with a system disc, and a backup disc. Backups were also written to half-inch tape.
A larger systems might have many of these drives. 3 GB would be ten of these.
Reading through the other answers, it seems like the college I went to was somewhat unusual in this respect.
In 1982, they had an (already fairly old) CDC Cyber 170 (model 720--pretty much the bottom end of CDC's range).
Despite the low-end CPU, they had 3 1.8 gigabyte disk drives (plus, a couple of tape drives).
Trying to get meaningful sizes of things "from the inside" was a bit strange though--Control Data's operating systems didn't use such primitive measurements as "kilobytes" or "seconds of CPU time". Things were measured in "primary resource units" (PRUs) and "System Resource Units" (SRUs). Although you could compare those relative to other things on the same machine, I'm not sure anybody was entirely sure how to convert either one to units the outside world would understand.
-
1PRUs... SRUs... 14 kiloquad interface modules, with the core element based on an FTL nanoprocessor with 25 bilateral kelilactirals...– MichaelCommented Sep 13 at 7:49