DEC: The best of systems, the worst of systems

My 1st real job in computer operations introduced me to VAX/VMS on an 11/780

I later moved to programming and spent 20+ years on VMS systems VAX and Alpha.

The VAX architecture was for its time the best by a long way, easy to use, easy to debug, and probably still the most stable and reliable OS you can buy. There may have been some instructions that were hardly ever used, but DEC did the right thing when it came to machine checks.

Every (and I mean EVERY) instruction possible on a VAX was ether correct and verified or it was guaranteed to throw a hardware exception. This was also true of the Alpha, so unlike the early SPARC, MIPS and other processors you couldn't crash a system from user mode just by creating duff machine code.

As a result DEC were able to both add and remove instructions in later hardware and just add code to the exception handlers.

For example the MOVC3 and MOVC5 and other instructions were not in the MicroVAXs hardware, they became functions in the OS that were called when the CPU throw an exception.

Likewise the later Alpha CPUs had new instructions for accessing single bytes in memory, this improved performance of later Alphas a lot, but running that new code on older Alpha would run (slower due to the exception handling).

Likewise VMS was (and is) source code compatible with all previous versions, and image compatible within a hardware platform, so I can run a program compiled 30 years ago on an 11/780 and it would still run on the MicroVAX I have at home or a SIMH emulator today.

Somewhat different to the crap OS's we have today with Windows or this weeks release of Linux not being compatible with a year old application without a rebuild.

I'm certain anyone who truely got to know and understand VMS misses it.

Ahh the memories

PDP11/45.

80Kwords of RAM - 4Kw Shared with a PDP 11/05

TWO Unibus Repeaters

Two RK05 Disks

I TU45 Magtape

All sorts of A-to-D & D-to-A plus digital I/O stuff. Including a programmable clock

The software was a mixture of fortran & Macro-11. It ran in a 20mS cycle controlled by the clock. The applicatino was driving the Avionics System for the RAF Harrier. This was 1975 at Hawker Siddley, Dunsfold (Where Top Gear comes from).

Man that system was good. The 11/45 was about 250KHz..... Kilo Hertz. not MHz or GHz.

Now for my confession. I worked for DEC from 1979 to 1999. I was the systems Engineer that took the RL02 drive off the top of the VAX 11/730 and put a Tape Drive in. We called it the Compact Vax.

Now I'm off down the pub to toast the great team of people I worked with at DEC CSS in Reading.

DECsystem-10

jrst @ .

Happy days clocking up 'resource units' used in the dark hours of early mornings.

However, as to its OS (TOPS-10 presumably) being rock-solid - Not when it came to a batch file which assigned too many drive letters, popped the kernel stack and brought the whole system to a shuddering halt.

Alpha

"Don't get me started on the Alpha…"

How dare you trash the Alpha. Heathen!

The Alphas might have been power hungry heat pumping beats, but they screamed performance wise. Not only were they fast, but their design was a HELL of a lot cleaner than anything Intel ever designed. They were clean, fast, and just all around awesome.

At least when buying Alpha gear from DEC, you knew you were buying a machine that could literally get the crap kicked out of it on a continual basis and not even flinch. I wish I could say that about our Sun gear!

MIPS-based DECstations/DECservers

Some of the MIPS Rx000-based DECstations/servers were quite impressive machines and ran Ultrix, a fairly vanilla BSD Unix variant. The only problem was that they were introduced late in the game, and DEC salespeople would rather shoot themselves than sell you one. I remember explicitly asking for a DECstation 5000 quote, and then getting a visit from the DEC rep pushing VAXstations.

The Good, the Bad and the Baroque

VAX-11/780 was my first taste of DEC, but in truth I had more fondness for the dinky little PDP-11/03 that bootstrapped it. Some years later I worked with VAX 6xxx's, and they gave so little trouble I took them for granted. Still later I worked with Alphas, they were quick, they were reliable (running NT 3.51 or DigitalUNIX) and they were a joy to cut code on. It took me less than an afternoon to grok the disassembled output of some broken C code having never seen an instruction set manual. My all time favourite ISAs are Alpha, T222/T800, 6809, and VAX-11 for the sheer kitsch of it (the -11 bit isn't just there for show folks) ... PDP-11s are fantastic boxes too, but I never learnt to enjoy the PDP-endian thing.

I find it odd that DEC don't get more credit for what they achieved with silicon down the years (as a low-volume niche player they had the odds firmly stacked against them). Time and time again they pushed the back boundaries with the -11, the uVAX and the Alpha. What would you rather have to cut code on 6MHz 80286 or a 15MHz J-11 ? Then years later you had the choice of a 333MHz PPro or a 600MHz 21164.

Yay of Alpha

Another +1 for the Alpha here -- I was a placement student when I first came across DEC kit in the form of a DEC 3000 supporting a not insignificant development team. In a world of 32-bit, 64-bit seemed like the future :-) I now have a DEC 3000 acting as a monitor stand on my desk, but keep wondering what I could use it for if I powered it back up again...

Bring back "This Old Box"

And show us some of these machines in detail!

Dec started Me out in Computers

Spent the Dinnertimes at Tech College playing Lunar Lander on a Dec PDP ??

with RK08 disks GT40?? Vector terminal

I Went From this is GREAT (How do I Build One)

Then to the 77-68 Then Acorn Atom BBC A/B Amiga, I all ways Wanted a Alpha

By The Way You COULD Screen Edit On A PR1me

We Had Dacoll Terminals with a Custom ROM it could display Teletext also

(Any Colour as long as its Green ) & Software on the Prime to Do Full Screen Editing...

You could recognise a DEC engineer by..

..the lacerated fingers.

Thing I instantly recall about DEC PCs was the razor-sharp, unfinished metal edges inside. Very nasty to work on, especially as they seemed to go for ultra-compact layouts which meant digging-in tight corners to fit HDs etc..

Only ever pulled one PDP apart, and I don't recall it having the same 'sharps' problem as the PCs.

(There isn't a thumb dripping red stuff icon so..)

You forgot the biggest failure of them all

The PDP-14.

This was meant to be an "industrial" controller (nowadays called a PLC) that one programmed in pure Boolean form, no "language" as we understand it was available. Once you had your program all designed, you sent the design to DEC and they would create a core memory board with the Boolean logic implemented by weaving wires through the cores. The resulting board was shipped back and you put it in your PDP-14, turned on the power and prayed you hadn't made any mistakes in the Boolean that would result in a machine malfunction. If you had goofed, you had to go through your Boolean code again, send the corrected logic back to DEC and wait for a replacement board. Meanwhile, your production machinery was down and you weren't making any money.

We had one (and only one) machine that had a PDP--14 that had gotten programmed correctly so nobody in Manufacturing Engineering was allowed to request any changes to the machine.

PDP-14's didn't last long being beaten in the market by Modicon 084's and Allen-Bradley PLC-2's that used standard computer memory and were easily reprogrammed at the machine with a "programming panel" that translated "ladder diagrams" into internal machine code.

----------

I know of one Data General Nova at the University of Michigan that became so hated by both the sysadmin and users, they took up a collection, rented a backhoe which they used to dig a six-foot deep hole in the lawn next to the building, then tossed the Nova from a third-floor window into the hole, back-filled it and tamped it down hard.

-----------

Not all of us old timers had BSCE's. My degree was in Physics, but I still groked how a computer operated almost at an organic level. I often could literally "see" the code working and the bits moving about.

VAX does more than clean your carpets..

Vax written off in just a sentence! Sacrilegious! As recently as five years ago I worked on 3 Vax systems, two of which were running a financial and a safety critical system. Reliable as hell, but any graphic screen changes (where supported) were, well, just a bit slow.

'Rubber banding' was basically clicking the mouse button (and they were awful mice) and then dragging to select, but nothing would actually happen, not even a dotted outline. Returning after a cup of tea would have resulted in the 'selection' of the desired icons etc. However if you'd have let the button go at the wrong time, and had not selected what needed to move/change etc. it was time to click, drag, and have another cup of tea...

Anyone for Chess?

Anyone remember playing Chess on a HSC50 cluster controller?

Mike

VAX - Just how wrong can you be

Interesting that you should consider VAX not Gordon's best work, and lacks elegance. Well as somebody who wrote device drivers for DEC in the first days of the VAX and for 10 years on PDP 11 I for one consider the VAX some of Gordon's best work.

While the PDP had an elegant instruction set you still wrote code for the hardware.

The major step forward on the VAX was that the hardware was designed for the operating system. Take a look at how the Exchange/Jump instructions work and you will find hardware that automatically dumped process context onto the stack or the process header. Suddenly when you called a subroutine/function the interface was through a hardware instruction that put the calling parameters into a frame on the call stack. The hardware was so integrated with the software that when we upgraded from VMS 1.0 to 2.0 we changed all the CPU boards.

Next lets look at Virtual memory, in 1978 we were in an office with 8 users running all the packet network development code, with one 11/780 which had 256K of memory. It was still quicker than we got from our 11/70 running RSX-11M+.

The VAX took DEC from a small computer company to a monolith that didn't know how to spend its money. There are many reasons to show why DEC got it wrong, but VAX/VMS was not one of them.

Though the PDP 11 was my first love after programming a CDC 6600, VAX/VMS was spectacular with an instruction set that made high level languages obsolete. It was only later when we returned to MIPS/INTEL/RISC/ALPHA hardware, that we lost the wonders of the 11/780 instruction set.

Happy Days

PDP-11/34 seemed so high tech in 1988, yet I suspected I had more power in my Atari ST at home? Had to use one to download data from flight data recorders on GR5 Harriers!

Plug in the "Black Box" which is actually bright orange, (only goes black when it gets burnt, but it is fireproof) select the data sources you wanted, get a print out on some weird multi trace wet plotter, then hang the traces to dry or all you got was a huge multi-coloured smudge.

wow - el reg readership age demographics

never realised there were so many old timers on here....

Sorry, I said ' I NEVER REALISED THERE WERE SO MANY OLD TIMERS ON HERE'

PDP 8/S

Having sampled the wonders of punch-card batch-programming in Fortran on the UNIVAC-1107 in the mid-'60s and with a subsequent stint in navy communications and cryptography - which were just going digital at the time, I returned to grad school to find a new PDP-8S being used to replace electromechanical logic-programming devices for data acquisition and control of behavior and physiology experiments. Data analysis was done on an interactive IBM-360 in various languages.

Say what you will about stodgy serial data-buses and scant core memory, as a student, the experience of "toggling in" the "boot-loader" from the front-panel toggle switches (in octal), subsequent control programs from paper-tape - bringing a heap of dead digital circuits and core-memory to organized life one step at a time, mastering my first assembly language, writing interrupt service routines, and creating critical programs on which expensive research depended, provided invaluable insights into the workings of and appreciation for subsequent micro-processor systems with all the basics automated and choices of low, medium and high-level languages.

In the almost 40 years since, I've seen many changes, stayed with most of them to produce ever more capable data acquisition, control and analysis systems, and developed a hypothesis about programming and programmers that I've yet to see violated.

That is, The "level" of programming (assembly, low, medium, high or specialized apps) which one masters first and in which one produces the first complete project, sets the level below which an individual rarely, if ever, masters or succeeds. If starting with a medium level, higher levels will be easy to appreciate, grasp and utilize, but lower levels will frustrate and be abandoned. Unfortunately, most programmers today started with medium- or high-level programming and have little feel for what's going on "under the hood" of their compilers or libraries, much less, the operating system controlling their apps. They argue "why should we?" at which point I find something more productive to do than argue with someone who doesn't understand the concept of actualizing the full potential of a given amount of compute power. It might be claimed that it's all relative to the resources available and the incentives, but that misses my point.

Favorite memory time now,

2am, in a computer room, with a Cray Engineer wondering why our Vax would not connect to the brand spanking new Cray XMP. Out with the pulse oscilloscope and digital signal monitor, crocodile clips on the massbus, and HEY - we found two pins bent on the mass bus plug so that touched.

Job done and a very happy bunch of seismic engineers next morning.

I loved that job.

"The Alphas might have been power hungry heat pumping beats"

In their day, yes, though in their day they were designed for fanless heat sinks because fans were considered too unreliable.

These days, routine desktop and server chips use almost as much power as the early Alphas and folks are perfectly happy to rely on fansinks (150W max for Alpha 21064, 130W for current x86).

"clean, fast, and just all around awesome."

That they were. Not just in the nice clean architecture and instruction set, but some of the system design features such as integrated memory controllers and a high speed bus for glueless SMP etc. Even Intel eventually caught on to those ideas, once they realised that IA64 was on the road to nowhere.

@Steve Davies re CSS

You and I will have to disagree about the importance and significance of CSS. Important, yes. Positive contributor? Far from clear.

In my experience initially as a customer and later as a DEC employee (though not in CSS), CSS entirely deserved their popular nickname of Cowboy Special Systems. Some of the products (such as the Compact VAX) did indeed address an interesting and valuable niche, but mostly CSS in the field seemed to be about being creative with DEC internal accounting so that two sales teams (CSS and the customer's own account team) rather than one could get credit for the same sales.

"All sorts of A-to-D & D-to-A plus digital I/O stuff."

DEC sold some handy stuff much of which was rebadged Data Translation cards (at least on the Qbus kit I used). Nothing necessarily wrong with that except unlucky DEC field service engineers typically didn't have a clue about "calibrating an A-to-D". Understandable but unfortunate.

"Including a programmable clock"

KWV11 was the Qbus version. Was it another Data Translation card? I can't remember.

National Instruments have long since taken over that test and measurement computer market. Their prices in comparison with the alternatives make DEC vs its alternatives back then look quite affordable; how on earth do NI manage to do that and stay in business (other than by lots and lots of advertising and marketing, eg Processor Brian Cox as keynote speaker at their most recent UK user group )?

And then there was the CSS Qbus graphics boards. Even in the days when workstations were taking off (after even DEC field management realised workstations were going places), there were CSS folks trying to extend the outdated VSVxx line. Madness.

DECdirect UK?

In the former Reading HQ (now demolished), not far from CSS, but a whole lot smaller and more productive, lived DECdirect UK. In the early 1990s their performance in terms of product awareness, demand creation, sales support and product delivery put most of the official box-shifters er sorry "channel partners" to shame, which is why DECdirect UK had to be abolished. DECdirect UK even produced useful glossy but not content free catalogues (later for UK+Europe), with information and prices for most of the commonly ordered stuff (software as well as hardware). They also managed together with the factories (primarily Ayr) to get commonly ordered stuff set up to do 5 day turnround from order to shipment, which was unheard of back then, and is still not entirely routine these days.

Tell that to the young people of today and ... you know the rest.

Oh happy days.

I can't understand how anyone can denegrate VAX.

First encountered DEC kit in the form of a 2060 running TOPS-20 at the Poly of Wales (now the University of Glamorgan), programming in Pascal and COBOL using LA-36 teletypes at first and then VT52's. Managed to cajole the ops staff in to giving me a login on the 11-750. Spent a year working for DEC at DECPark, Reading as part of my BSc course where I got to use various VAXen, Pro-350's and -380's, the odd Rainbow. Fell in love with VMS at version 3.7. Sponsored by DEC for my final year project doing data warehousing in Rdb/VMS then went back to work for them after graduating. Moved to a software house developing on a MicroVAX 2000 and then a 3100, before contracting for them at various times. Messed around with LAVC and VAXstations (kinda had a love/hate relationship with DECwindows).

For me VAX/VMS and the layered products were the business, providing superb development and production environments. Alpha could have taken up that mantle with better management. Seeing the emasculation of DEC as products and divisions were sold off, then the Compaq takeover and the subsequent HP takeover leaves me feeling sad still, even though I've not touched a VMS machine for 7 or 8 years.

Another former DEC employee

DEC was my first job out of school in the '70's. I had interviewed with IBM (amongst others) but was turned down for my grades being too low. DEC hired me because I had a bunch of real-world design experience while I was in college. I'd also been responsible for the maintenance of the Chem dept's PDP-8/i and was an operator for the engineering department's PDP-8/e.

My first job was working on the memory and memory management for the 11/44. One day I was cussing out the board, looking at timing issues, when I hear this guy behind me say, "What's wrong?" Without looking up, I said, "Oh this damn memory timing is off," while turning around, and there's Ken Olsen and Gordon Bell, doing their walk-around the Mill labs. They were then doing the "oh have you checked XXX?" and "let's hook another scope up to the row lines and see the timing there." I've never worked in another large company where the CEO was grabbing test leads and looking at debugging information!

As I said, I worked primarily on the 44. I also worked a little on the still-born 11/68 (11/60 upgrade) and the 11/74 (multiprocessor 11/70 system, basically 4 cooperative 11/70s). We used to love the 11/60 because you could rewrite the microcode and mess with people's heads! :) We also used to sit down the hall from the RSX guys, and would cadge source code so we could put hacks on our 11/70 development system to annoy our fellow engineers. Good times.

One last story. While on the 11/44 project, the lead came to me and said, "You're the young kid, you know hardware and software. We have these oddballs coming in to test their operating system on the platform. If that goes well, they'll buy these." The "oddballs" were people from Bell Labs, running Unix on the 44. I learned a lot about Unix working with them; I've been parlaying that knowledge into work for the past 30-something years!

I was supposed to go work on what became the VAX 11-730, but I was young and left the company to do other things. I ended up back there 20 or so years later as a consultant but that's a story for another day. I still have all my PDP-11 manuals and some of the software/OS "helper cards".

36 bit mainframe?

I worked a bit wit a PDP running VMS. Nice business OS as I recall. I'm told a lot of banks ran VAXen clusters because it gave them affordable reliability. Perhaps they still do.

I cannot get my head *why* anyone would think a data word that was *not* a power of 2 would be a good idea.

That said I've some experience of the boxes of the Harris Corp. 24 bits with an assembler (IIRC) whose mnemonics changed depending what part of the word you're addressing and for which the preferred programming language in terms of functions to access the OS was FORTRAN.

They're meant to have been quite good for running geological survey imaging systems and the DOGS CAD package. Rarely has something being *so* well named.

Fair Well Ken, although at least some of the VMS architecture lives on under the covers of NT and its successors.

Real-time performance feedback

You knew when a VAX/VMS was running slow with the delay echoing characters back on a vt-100 terminal, with a 3270 terminal you had to wait until you filled a whole page of code before you got that feedback.

The greatest contribution though has to be providing a platform for Oracle.. with DB2 we used cost-based optimiser application plans, but with Oracle’s rule-based optimiser you worked-out the plan yourself and coded the SQL accordingly.. getting instant feedback if the query was running slow from all the other users.

Without DEC, we’d all still be time-sharing computers in the cloud from terminals with some kind of SGML markup..

VMS roxxor

DEC and its descendants have been paying my wages for the past 25 years, as a VAX and then Alpha BOFH, and any OS I've been press-ganged into attempting to drive other than VMS has been a pale imitation of the real thing. It Just Works.

Also, I loathe Un*x from the depths of what passes for my soul

$ SUBMIT SYS$MANAGER:AGEN_FEEDBACK.COM/LOG/KEEP/NOPRINT/QUE=CORRIB$BATCH

I don't remember the model numbers of all the VAXen I've worked on, from '85 to ES40s in 2009, but I absolutely loved VMS. None of those machines ever let me down and DEC's field service was fantastic. I still refer to babies' changing bags as Field Support Kits.

My modest claim to fame was to hook up the Call Detail Recorder printer output from a Nortel SL1/st PABX (another box I loved) to a terminal port on a DECserver 200, to capture all the call details. I wrote a billing system to charge all the phone calls to the proper department's budget, instead of the IT budget. It was a part-time project, what would now be called a 20% effort. But it was a great success and reduced our annual increase in phone bills from around 40% to 5% or so.

3,300+ lines of DCL. Took about half an hour to run, once a week. Worked perfectly. It was right at the limits of what one DCL program could do.

I also had some small batch jobs which checked through backup log files and emailed me a report. One of those ran automatically every night for 8 years without modification. And a bunch of batch jobs which purged records and compacted files in a TOLAS ERP system on a 13 week cycle. Again, all automatic, and ran for several years more or less by itself.

First thing I always did was to remove BYPASS from the privs on the SYSTEM account.

Legend has it that there was a VAXcluster somewhere in DEC Engineering in MA with 200 nodes in it. IIRC, the officially supported figure was 32.

VMS, virus free since 1978. Nothing to match it for reliability. I miss it immensely.