Supercomputing in 1983

January 23, 2013 1:18 PM

"The late Dr. John Fletcher describes, in 1983, the history of the Lawrence Livermore Laboratory's Octopus network." This video describes the history of the laboratory's large computers, from the Univac 1 to the Cray XMP, the evolution of mass storage, high speed printers, the Octopus networking facilities, and the emergence of minicomputers.
posted by FuturisticDragon (16 comments total)

Here is a 1972 article based on a talk given in 1970(!) about the Lawrence Radiation Lab Octopus Network. At that time it consisted of 1 PDP-6, several PDP-8s and over 300 TTYs. Each PDP-8 supported 128 TTYs.
posted by Ad hominem at 1:40 PM on January 23 [2 favorites]

And here is a link (subscription required) to an ACM paper from 1968 describing LRLTRAN, a custom version of FORTRAN designed for networked computers.
posted by FuturisticDragon at 1:51 PM on January 23 [2 favorites]

So what you're saying is that he's ...Dr. Octopus?
posted by leotrotsky at 1:59 PM on January 23 [1 favorite]

This is so great! Thanks, FuturisticDragon. "As you can perhaps hear from the background noise, we are now located in one of the two major computer centers at the Livermore Laboratory..." Note that that machine he's pointing to at the beginning, their newly-acquired bleeding-edge Cray X-MP (which had debuted just that year) was a fifteen million dollar machine. He brags about it having "a cycle time of 8 nanoseconds" and "a memory of 250 million bits" – which means that it had a speed of almost 125 megahertz and a memory of almost 30 megabytes. I'm pretty sure that home computers exceeded those specs within a decade.
posted by koeselitz at 2:09 PM on January 23 [2 favorites]

Wow,Seems like LRLTRAN is about as obscure as it gets. This PDF talks about implementing TCP/IP to run on CTSS (Cray Time Sharing System, not Compatible Time Sharing System) which was written in LRLTRAN to. All so they could hook up to NSFNET.

I found a word doc concerning a KERMIT implementation for CTSS written in LRLTRAN but reading about a dead file transfer protocol written in an obscure language is too boring even for me.

That being said. I give it 20 minutes before someone who worked with LRLTRAN or CTSS shows up to school me.
posted by Ad hominem at 2:20 PM on January 23 [4 favorites]

FuturisticDragon: “And here is a link (subscription required) to an ACM paper from 1968 describing LRLTRAN, a custom version of FORTRAN designed for networked computers.”

Awesome. Note also that, by the time this documentary was being made, Lawrence Livermore National Lab had moved on from their original Livermore Time Sharing System (LTSS) to CTSS, the "Cray Time Sharing System," on which they collaborated with Los Alamos National Labs. Here's an original Los Alamos CTSS programmer's manual from 1982.
posted by koeselitz at 2:21 PM on January 23 [2 favorites]

I was just thinking today that it would be interesting if you could calculate, for the average size of the hard drive of a retail computer in any given year these days, in what year in the 20th century said hard drive's capacity would be equivalent in size to all the disk storage in the entire world.
posted by XMLicious at 2:25 PM on January 23 [1 favorite]

"Looking even further into the future, say 30 years, it is difficult for us or anyone else to foresee what that world will be like, just as it would have been difficult for those people 30 years ago working on the UNIVAC I to foresee what our world is like here today."

Assuming for a moment that he's talking about "the world" of computing, they actually saw it pretty clearly.

In 1983, PARC had produced a familiar looking GUI, and a Macintosh with usage incredibly similar to mine was on the verge of release. A TCP/IP-based Internet was already in use. Cell phones existed. The idea for tablet PCs and mobile computing was already a decade old. Even hypertext was well understood.
posted by rlk at 2:28 PM on January 23

I'm impressed with The Radiation Printer.
posted by mrbill at 3:26 PM on January 23 [1 favorite]

The IBM Photostore is something I hadn't ever run across before. According to Wikipedia, it was a commercial dead-end and only a few of them were ever manufactured. But it has the legitimate claim to fame of being the first device to store a terabit (about 160 GB) of data, and did so quite significantly before magnetic storage of the same amount of data in a similar volume was practical.

And I also thought it was interesting that (during the segment on output devices), the narrator mentioned that there were 3 million pages per month of printed output produced, but 10 million pages of microfilm -- that was really the dominant output medium, not paper.

It's strange how photographic storage, e.g. microfilm and microfiche, was so firmly in the minds of most visionaries (e.g. Vannevar Bush in his famous 1945 essay "As We May Think") when it came to data storage up through the 80s, but has been so quickly and apparently thoroughly forgotten.
posted by Kadin2048 at 4:31 PM on January 23 [2 favorites]

Kadin2048: “It's strange how photographic storage, e.g. microfilm and microfiche, was so firmly in the minds of most visionaries (e.g. Vannevar Bush in his famous 1945 essay "As We May Think") when it came to data storage up through the 80s, but has been so quickly and apparently thoroughly forgotten.”

I was struck by that device, too – and by how obvious it seemed to me immediately that it couldn't possibly be sustainable, at least as far as I can imagine. I'm guessing that people thought of photographs as the most high-resolution and instant-seeming technology at the time, but the video says something about how the drive 'contains a full working chemical lab for developing its film' - sheesh.
posted by koeselitz at 5:49 PM on January 23

Remember, Ad hominem, The List is:

1. Hello World
2. Kermit
3. Unzip

posted by NortonDC at 7:44 PM on January 23

the drive 'contains a full working chemical lab for developing its film' - sheesh.

Yeah, although that's not really all that bad -- I've maintained a film processing system and although they require topping-off of chemicals and water they're less fussy than some humidity-sensitive early tape storage devices.

What's really impressive about those things was the throughput: (from WP)

Speed of the system was fairly good, writing at about 500 kbit/s, and reading at about 2.5 Mbit/s. Cells were moved between the Files and Readers using a pneumatic tube system similar to those used to move documents around in some stores and hospitals. The system could keep up to 13 cells "in flight" around the system in order to minimize delays.

That's pretty good for the time, particularly considering that it basically used all off-the-shelf technology (black and white film, CRT technology to write the data, manchester encoding, etc.).
posted by Kadin2048 at 8:25 PM on January 23

I'm pretty sure that home computers exceeded those specs within a decade.

Which has precisely nothing to do with the comparative performance of the XM/P-48 and the decade later PC.
posted by kjs3 at 9:36 PM on January 23

Memories. I did time around '86 on this class of machine...XM/P-48, CDC Cyber 180/990 and 205, ETA machine. Fascinating to profile performance on different workloads between the Cray vector register approach and the CDC 205 memory-based vector approach. Good times.
posted by kjs3 at 9:40 PM on January 23

Lawrence Livermore National Laboratory: nothing meets our needs.
posted by SkinnerSan at 9:41 PM on January 23 [2 favorites]