@@joeofloath Sure, but would you really want to ship a 25 year old card laden box clad in brittle plastics on a pallet via ground? Survey says pay for the next tier up if you want it to live.
Seeing how they've "restored" the Apollo Mission Control (it's mostly show, real consoles with modern hardware inside, like LCDs instead of CRTs) makes you wonder why they won't stick to this or at least part of it for historic reasons. Especially when they spent ages figuring out the correct type of carpet the room originally had... I mean, sure you can emulate the entire Apollo era complex at Houston on a desktop these days but that's not as cool as the real deal imho.
I remember like 10 years ago while working as a backup consultant, I went to work on backups for a government office they where in charge of public water, they had exactly this model and it was absolutely forbidden to get close to it, it was so old and nobody there really wanted to touch it, they had cones and “police” tape it was funny but it was pretty serious thing. They warned me that there was no spare parts and support for it and the app was a legacy one so they couldn’t just update it or migrate since the guys who developed it just disappeared many years ago. Funny days as a consultant.
heck... hp-ux was so damn backward compatible, as long as you'd have stayed on PA-RISC you could have run HP-UX 8 code on even the last series of HW and OS... always sad when fear and ignorance cross paths. They had people who knew, so they didn't bother to learn and then they're left ... putting cones around it. lol.
@@udirt To be fair though, I heard unix consultants charged out the ass so if they didn't have an inhouse guy know knew the thing. I have seen this fear from old PDP systems to even modern web servers.
It's only an issue if they're too cheap to find a decent developer or 2 and pay them a reasonable salary to migrate whatever it is to something modern.
The old "Let's hire an external consultant to write a bespoke business critical application and not demand the source code or comprehensive documentation because it costs extra." - classic! What could possibly go wrong?
I disassembled and scrapped about 40 of these suckers in 2005. GE donated them to the local refurbusher/e-waste place I volunteered at at the time. The CPUs are freaking huuuge
@@kedamono2562 obsolete tech becomes recycled or gets dumped in a landfill. This stuff has very limited uses in modern day. People out there who collect this junk and do anything meaningful with it are scarce and when they die the same thing will happen to their collections. Yeah you paid $400 for a outdated e machine but your next of kin is going to dump it to the highest bidder or throw it in the trash
perspective, a 1996 system in operation until 2008, that is 12 years. My home server is an i7-970. made in late 2010/early 2011 and is still running today and likely for some more time, also about 12 years old. No one would flinch about a i7-970. I recall working in the 2007~2012 range and losing my mind whenever I saw a min/late 90s server running something like NT and making it my mission to get rid of these relics. Today we have server 2012 boxes that are starting to be replaced and I feel like 2012 was just yesterday and I'm getting old.
The difference is the OS. A server from 96 isn't going to run an OS from 2008 very well, or if at all. Never mind the many hardware changes in those 12 years. Compared to now where an i7-970 will run pretty much whatever you want and if you need something hardware wise its board doesn't support, you can hop on over to ebay or where ever and most likely find exactly what you need.
Have to remember that hardware was improving at a much faster rate back then compared to today. A 10 year old 486DX in 2000 was no where near as capable as a second generation core i series workstation from 2010 is today. My main desktop is still a dual dual core Xeon 5150 with 32gb ram from 2006 with hardly any upgrades. I could certainly use an upgrade but it is handling Windows 10 and kubuntu with no issues so no big incentive to upgrade yet.
@@kathrynradonich3982 I agree and I wasnt trying to imply that these 90s systems were actually still good in late 2ks but more to reflect on how much has changed and accelerated since the 90s.
The main thing is single user Vs. 1000s. Just like my 1993 386 was still tolerable for games in 96, the i7 is still ok. But for a lot of users a desktop grade system lasts one generation, while servers need to be good for 2-3 even IF you start replacing after 3yrs. And what always pissed me off in the past: people never upgraded Ram at all, ignored broken raid bbus / never properly enabled caching and then they throw out systems that never ran well... Like Xeon 5500 series that could easily fit 96GB and they run one as their SBS server with like 8GB Ram.
Could they have also been used with those weird modems where you put the handset up to a speaker and it listens for the data as a series of tones? Like as a backup of a backup of a backup or for backwards compatibility with really old stuff that used that?
@@forgottencameras those modems with "acoustic couplers" (where you put the handset up to a speaker) also used the standard RS-232 ports, on this server it would have been plugged into the "external modem" connector.
I used to have one of these in my house. I got it from a school on the cheap, and used it to manage my file archive, DVR, and ran a BBS for Text Games and classic computer game files. Man, what a time.
@@ninja011 I do recall a BBS "Divided by Zero" being talked about in my company's break room at that time. those were the days I miss them. The young folks have high speed internet today and smart phones its great but they missed out on the BBS glory days!
I worked for Sun Microsystems in the early 90’s. This was a “competitor” to Sun. Your presentation was fun to listen to but contained several factual, how would you say, deviations from reality. But that’s understandable and I commend you on keeping this part of computer history alive. HP made incredibly reliable hardware. The fact you can turn it on in 2021 and get a prompt supports that statement. Hence the requirement for a single power supply. Electrical sources were less reliable than the machines themselves. Turning off a Unix based machine and then turning it on again at a later time was always a very stressful event. Disk corruption was not uncommon. Then you’d have to perform an fsck and hope that repair was possible. For this reason, you’d always want to “sync”, twice for good measure, to flush all the in-memory writes not committed to disk, prior to a “shutdown” or “halt”. On a server if this type you’d run a shutdown command that would send a message to each logged on user giving them ample warning to save their vi or ed sessions and log off. 1996 seems a long time ago but communications protocols were not very much different from what they are today. Client-Server computing was a thing. So you could be running something like Ingres (precursor to Postgres) serving multiple client programs. The multitude of serial ports could be used for modems or more likely “terminal servers” such as a Xyplex that gave you access to the machine via a telnet session. Running a bank of modems into a terminal server and then to the actual server was also common. To put things into perspective, I was running Perq computers with an operating system called Penix (no kidding) having a portrait graphical display (1024x768) and a bit pad (a deluxe optical mouse like device) with a whopping 10 mb Harddisk in 1983 or so while at McGill University. HP was rarely ahead of the technology curve when it came to workstations. But the hardware was solid. If ever you get another one of these, go ahead and open it up. There is nothing to break. By the way, the CLI on that machine was definitely “sh”. “sh” exists on all Linux machines to this day. RaspberryPi users would feel quite at home on that thing.
The 9000s were workstations as well as servers. The multimedia was used your application had a multimedia element. My fathers company used then for acoustic testing in the 90s.
@@jamesjross yes indeed. SGI had similar footprints. Imagine a workstation like this under your desk. Imagine 4 in close proximity. The noise must have been deafening.
@@AllanKobelansky it was in a room attached to a chamber of sorts.. It was 20 years ago so I can hardly remember. They had 4 Sun workstations for PCB designing - that was more like 92. I can't remember it being loud - I'm guessing the terminals where in a different room
I currently work at an HP campus and believe it or not these things are still in use. I find it entertaining to see all of the old 90s tech around that site
@@silversurfer420 and why would anyone need to change them? If the box is doing what it’s supposed to do, there is no need to replace it. I’ll wager that machine will continue to work till 2038.
I really hated having to deal with HP-UX. I especially hated dealing with HP's pre-boot environments, because it was always a fight to get them to pick up and use the boot media I wanted them to use.
Sold and supported the same. I came from the MPE side with HP3000’s. These were the smallest server system we made at the time. I did have a HP 712 desktop “workstation” on my desk and I really learned HP-UX that way…
And I'm pretty sure that this is a low estimate, because like today, servers used to have different, more complicated and higher quality memory than consumer or even workstation machines. I wouldn't be surprised if the actual number was much closer to half a million. Edit: We see the price in the video later on: 256MB = $12000. A full 4GB adjusted for inflation would then be at ~$335000. So not quite as bad as I thought, but still wow.
@@craigjensen6853 you could, yeah, but I'm Not Sure you'd want to buy a Chromebook. And for me that's Not even because they're that Bad. I wouldn't want to buy a Chromebook because their ads are so useless. "Tired of having to Charge your Laptop? Switch to Chromebook". That doesn't make any sense. As If the Chromebook ran on a Solar Panel or some other Kind of bullshit. No it doesn't, it's a regular dumb Laptop. You'd have to Charge it less often than say a ThinkPad with an AMD Ryzen, but the ad suggests you'd never have to Charge it. Same with the Updates. They try to get to the specific Kind of User Base that doesn't know how to disable automatic Windon't updates. "Are you tired of your Laptop rebooting while playing a Game to Install an Update? Switch to Chromebook. Because then you can choose when to Update". Like duh, I can do that with Windon't too, Not that I'd want to because I don't really use it, but it's Not that hard. Also this ad suggests you can Play Games on a Chromebook. You could If you were to Install Chrome OS on a powerful Laptop but a Chromebook itself is really only suitable for Things Like Solitaire or maybe among us.
To be clear, mainframes are used for a specific reason, such as reliability and scalability. The banking sector cannot afford to lose a single transaction, much less "move fast and break things."
100%, and back in the early 80's when I was still a toddler my mother being a single parent(my parents split when was 2 1/2) was an oncall 24/7 ATM/computer technician to make extra money for the bank she worked for(First Citizens Bank & Trust Co.), and even at that age I can remember her grabbing me out of bed in the middle of the night to make a call, and some of those systems she worked on at the time with 10-40MB HDDs the size of a small lawn mower engine where still in use well into the late 90's simply because of cost, and the fact they could not loose customer account data, where some customers had $100,000's, or even millions of dollars in them that had been with that bank for well over 1/2 a century at that point.
@@TomStorey96 I've been a long time sub of Dave, and saw that video, it's indeed crazy how big that HDD was, and to think these days we carry around 128GB MicroSD cards in our phones like it's nothing.
Also System-Z is a HUGELY profitable enterprise for IBM so they're going to continue making it. And they're actually doing a lot to support training "the next generation" of Mainframe operators (they even have a lab you can access for free if you're interested in it) It is pretty much guaranteed that you have interacted with a Mainframe today, they aren't going anywhere any time soon.
I ran these machines. The reason it does not have redundant power supplies is this is a lower end model, it likely didn't come out of a server room. That I/O panel gives that away, It does not have the color workstation options which was another market HP was after. A lot of these were used in industry and control rooms to run experiments and production equipment, if it were it would have the additional I/O boards and a large color monitor. The console connector (often this was duplicated on the front panel on rack mounted machines) was mostly used to monitor the boot up. These machines don't have BIOS so this connector was pretty important to keep them running. I used to access the console with my palm pilot to equipment like this. At the time this part of the market was pretty heavily competed by not just IBM and SUN but SGI, DEC, Data General, and a number of German and Japanese companies all had mobile ped models. A number of these machines were market specific but if you knew unix you could sit down at anyone of them and get some work done. I used to say that there were more similarities between any two varieties of unix than any two major versions of Windows.
They are absolutely ROCK SOLID servers still. I can't attest to their consumer market products but the two HPE personal servers I've dealt with the past 7 years have been on 24/7 with a minor hiccup of the integrated SD card reader.
Started my IT career as admin in '97 in Germany, and one of the first servers I was responsible for, was this HP9000 K class. It was the main server for the whole SW development team, later we got an add-on tower with 10 (iirc) 18GB SCSI harddrives, alone this did cost a fortune at that time :-) We decommissioned it in 2010 or so, maybe it is still somwhere in the basement, besides the MicroVAX :-)
Back in the day (1985 thru 2000), these systems were all over the place. They required real skills to implement and manage. Now everything runs on Linux servers and things are relatively simpler. The only thing complicated now is the Web UI layer. (Which hopefully at some point will go through it's simplification phase.)
HPUX, AIX, or even dual universe unix flavours - once you know the core, it's all good. Ported a radio dispatch system from NCR Tower onto HPUX back in this time - still not no DEC VAX environment but still, fun. Oh and early HPUX had an error code of 8008 - that ended up getting changed due to early PC mentalities in marketing overruling that showing on a display due to calculator word legacy.
@@mndlessdrwer It was always installed. If you ever saw it depended on the usage scenario - most devices probably ran just databases and server SW for rich clients. XWindows desktops for terminals or X emulators were seldom seen in the nineties.
I know for an absolute fact that as recently as 2018, SOME hospitals with pneumatic tube systems to transport blood samples and other sensitive items, were STILL using servers similar to this, sourcing spare parts from friggin nasa, just because the entire system was relying on some proprietary OS from 30 years ago that has no possible hope of ever being reverse engineered or ported into a modern OS. It was only after literally investing tens of millions of dollars to basically gut out the entire tube system and replace it that the old beast of a dinosaur finally came to a rest.
Are developers that can reverse engineer something really that hard to come by? I'm sure the tube thing isn't that complicated, and could've been reverse engineered for much less than "tens of millions" of dollars. I would've done it for $1M easily!
@@gorak9000 do you really want to be responsible for anything a life depends on? I decided when I was like 10 (started coding when I was 5) that I wouldn't ever take that responsibility personally.
@@forbiddenera The pneumatic tubes they're talking about transport samples from the hospital to the lab. If the tube breaks or stops working, the worst that happens is they take a new sample and someone walks it over to the lab instead of sending it down the tube system. It's not directly connected to the patient. It's the same kind of thing some stores use for moving money to the cash registers and back, and some banks use at their drive through tellers for deposits. Not sure if they're used anymore, but most Home Depots have them.
@@gorak9000 I suppose but extra time could still cause problems..also I assumed it might be used for other purposes in some hospitals like sending med ampules or something, I have no idea. Though, even if you're 100% right, hospital admin/insurance might disagree or require a certain level of service or guarantee or something. I dunno, I'm paranoid enough doing financial stuff lol
The lack of redundant power supply is most probably due to that this is a small workgroup server, the larger HP servers for datacenters had redundant power supply's...
I was system administrator for the a couple of K460 versions from 2000-2005. The chasis was the same size of the one in the video. That was 20 years ago, but I don't remember having a redundant power supply in that chasis. For sure the bigger ones did. We did have the HP AutoRAID 12h which had triple power supplies, 12 drives, and dual controllers.
@@timkr66 and redundant batteries (but i hated the autoraid). agreed that k-klass never had multiple PSU, but there was the fat brother in the 90s, i think T-class as SuperDome's business use case successor. never saw a T-class outside of history websites.
Nope, the K class just didn’t have redundant power full stop. It wasn’t a “small workgroup server” either. At the time of its introduction the only HP9000 systems that would expand more were the T Class, and I don’t think they had redundant power supplies either, although I think they did have more power resiliency than the K class.
Their rack-mount server power supplies were extremely loud and, well, less reliable than the massive single PSU. They were prone to letting out the magic smoke and causing problems. I suspect it's because they had to shrink everything down to make it fit and it couldn't keep up with the cooling demands of the components in the PSU. Still more reliable than some of the modern PSUs, but finding replacements without constantly renewing a service support contract with HP was always a bit of a nightmare.
Man, this brings back memories... When I started working, I had access to this beast, along Sun Solaris servers, Compaq TruUnix servers, Digital Unix, Red Hat 6 clients... Being at work looked like being on a spaceship 20 years ago :D
Same! My first job out of high-school back in 1999 was working for a company called DCG Inc and they specialized in Alpha Linux clusters. I built a lot of Alpha systems, the biggest one being like a 20 server cluster for Bloomberg! I even had my own Alpha workstation that I got to use at work for things like AIM and Email. Those were the days!
This system was designed by the General Systems Lab in Roseville, CA. I was the system design lead for the corresponding entry level platform, the D-Class systems and I adapted the four to six processor designs of the K-Class to the dual processor D-Class systems. At the time, PA-RISC/HP-UX was used both by the Systems Technology Division for enterprise computing and by the Workstation Division for engineering workstations. After the acquisition of Apollo, there was a more concerted effort to leverage ASICs and engineering between the divisions, and this is reflected some of the core I/O hardware, like mouse and keyboard, but it was limited in the K-Class. The D-Class actually shared much more with the corresponding generation of workstations, as we defined a common two socket CPU/Memory design that allowed us to share much more.
Kirk Bresniker is the real deal when it comes to these platforms and he is of course spot-on with all his comments. HP sold a ton of the D and K-class hardware at that time, targeted for the OLTP customer using big-metal databases, and Kirk was right there, in the center of the action. I'm glad to read his commentary and happier that he has time to look at TH-cam videos.
So how did this crime happen that the L2 Cache was an optional component most people would not afford? You were selling castrated servers! I got my first D-Class in late 2000s as payment for "i was never here" storage support at an insurance co, proving their EMC array was underperforming. I think I later sold it for a crate of lemonade, minus the cache module that went to my A180.
@@udirt Why is "you get what you pay for" a crime? And how is it different than paying more for a 5% resistor than a 20% resistor than came from the same line or an x86 CPU SKU with more cache than the cheaper SKU with less that came from the same wafer? The D-Class offered a 10X performance range over a series of in-package upgrades and that was really important to our customers since switching to a different vendor was a big investment. Whenever we introduced a new upgrade the biggest bump in sales was almost always for the prior top end because our customers always wanted one more easy performance upgrade, they were planning on their future success and we were helping them realize that future.
Thanks, @@curtthompson7279, I'm on a 665 day streak with the NY Times Crossword so every night I have between 5 and 60 minutes (depending on the puzzle) to also watch TH-cam while I solve and vids of old HP equipment are my hands down favorites!
I'm a field server engineer and I've worked on these twice. They are so foreign compared modern servers. I still have a couple customers that still use them for legacy data.
The CDE login screen on the "graphical interface" is so nostalgic for me. CDE was also the default window manager for Suns Solaris OS and I used that in college. Well, I didn't USE it because I hated it. I always switched to KDE. Yes people KDE existed in the the late 90s and it was so much more familiar and configurable then CDE. Also fun fact, XFCE one of the least resource intensive desktop environments currently for Linux. Started as a CDE clone. Then they got their act together and made it much better. ;) For those of you interested in what CDE was like: en.m.wikipedia.org/wiki/Common_Desktop_Environment
@@ps5hasnogames55 maybe up to 5 or 6 years ago but not so much these days, GNOME in my experience is waay more resource intensive and far less usable, a nice minimal KDE-Plasma install is closer to xfce and mate in resource use, and is by far the most consistent in look, feel and configurabiliy. Kmail does still suck though
@@mrmysteryguest Warning... long rant. "Minimal" KDE just doesn't exist. All of it ties into these useless bloat framworks and Qt is semi-proprietary trash. At least it looks nice and works somewhat like how a normal person expects a computer to work. Too bad KDE crashes on you if you blink at it wrong, and KWin is still a horrible window manager that doesn't work with mutliple monitors. Steam Deck customers are going to be in for a world of pain when they connect their shiny new hardware to an external monitor only for KWin to outright crash on them, or for their panels to go missing when they disconnect the external monitor. GNOME's idea of a "desktop" is some abstract bullshit that's unusable for an average user. You want desktop icons, like every other desktop has? Too hard to maintain in the file manager so we'll just remove it and force you to install an extension, and if you complain, well you just shouldn't use your computer that way because "minimalism is the future". Tell that to every normal Joe and Mary who just want to get their work done and puts their most-needed documents right on their desktop so it's just one click away. Their developers are arrogant prats (like every other project that Red Hat usurps while parading around that they support free software, lol whatever). While GNOME doesn't have a tonne of framworks, instead it relies on (and is the reason every other distribution is forced to include) subsystems like Systemd, Wayland and PipeWire that nobody asked for. Classic Red Hat forcing everyone to use Linux how they want you to, right down to the init system. Funny how Canonical gets hate for making Linux easy with Ubuntu, but Red Hat gets zero hate when they're the reason your desktop has graphical glitches and random crashes (GNOME on Wayland randomly booting you back to the login screen causing you to lose all your work with no explanation sound familiar?) and your audio doesn't work (PipeWire, just like Wayland, is incomplete, beta-quality software yet they're still trying to force it on everyone), and has loads of security bugs (systemd and all of its tentacles and everything it touches and tries to integrate... and runs all on PID 1 :wink:). Cinnamon is alright, they have the right idea of how to make a usable desktop and it was forked from GNOME 3 right at the start before Red Hat forced everyone to use insecure and buggy bloat software no one asked for like systemd, Wayland and PipeWire in the name of "innovation". But at the end of the day, it's still GNOME 3. Bloated. MATE and Xfce are lighter, but that's only because they're outdated. You show a desktop runinng MATE or Xfce to an average user, they'll laugh at it thinking it's from 2005. Of course there's still more like that one that Elementary OS has (wow it's just GNOME re-skinned to be even more iPad like... who cares) or that Pop OS Cosmic thing (it's literally just GNOME with more bloat extensions for window tiling, as if any normal computer user works like that).
@@ps5hasnogames55 for the last 17, almost 18 years I have been a Linux user, a distro Admin/Dev, minimalist, and for the last few years recently a more pragmatic user of whatever tool gets the job done. KDE-Plasma has for the past few years been more stable than any other environment I have tried/used, including the self styled wm only setups I used in the past and does not require a massive investment of time and effort to be usable. There is no such thing as a perfect distro or desktop environment, if there were, then there would not be so many available. Plasma Desktop has evolved and is evolving to be among the most usable and configurable environments, both on Linux and FreeBSD. The issue surrounding the licencing of qt may not become an issue as KDE has documented their plans should it become problematic. Although I currently use Kubuntu, I would whole heartedly recommend ArchLinux+PlasmaDesktop to anyone, even to adventurous folks not afraid to experiment that are new to Linux
This is beautiful. I worked on these and installed a few back in the day running HP/UX. All with FDDI. That card is single attach, by the way. Dual attach was also available. We also ran these in a cluster (actually active/passive pair), so the lack of a redundant power supply wasn't that big of a deal, since we would purchase two and have them failover. Such good times. I remember the failover code was version 1.0 and I worked with HP support to get it running. Very basic but worked very well. It included shared SCSI storage across two nodes, and the active system would seize r/w access with the backup system having r/o access. Also the days of yellow pages (precursor to NIS)... So much fun.
IBM are still selling AIX and IBMi systems in 2021, although they're rack mount only now. And you can still buy Solaris systems from Oracle. They're not dead yet!
@@dupajasio4801 Why do you say this? They still do things that other OS's can't. AIX is renowned for it's very long uptime, and the RAS features of Power systems are second to none!
@@petergathercole4565 Oh just bad experience with those systems. We used to run D class version of this server for our ERP. I still have to keep it's Oracle DB running on 2003 and XP client. Also we still run Teamcenter on Sun and IBMi for ERP. We are too small to properly manage these. So it's all voodoo kind of DR etc. Perhaps these systems are good for some apps but knowing Oracle and IBM they always are overpriced or are missing some features that are better supported or implemented in Linux or M$. It could be that our apps suck on those systems. Just my opinion. Regards
Being a massive HPPA fan, I still have PA-RISC systems here at home and work still running. My oldest is a G40 :) Build date is 1989. I've worked on and setup K-Class's back in the day. (And T's and D's and C's and snakes, and the instruments)... don't get me started.
I also had a few, have one demonstrating at my channel, the line I never quite understood is the E series, single cpu server not that much powerful than a workstation and I think at the time undercompetitive in performance vs. a Pentium or Alpha cpu, I had an E25 fully loaded and it was silly to run 11i on it, it was so slow for a mid 90's server.
My still working D230 is even bigger and very heavy. It was used for about 20 years, for administration of several network elements of a submarine cable, between Italy and US
@@benjames866 Yeah, I've a couple of the C-Classes in my collection, same CPU as this badboy but IIRC they're Uniprocessor machines. And I can confirm the multimedia ports are for exactly what they're for today.
I remember the electronic engineering students at uni in the late '80s early '90s used HP workstations with all the multimedia bells and whistles for designing ICs.
My first big IT job was to source a replacement for an HP 9000 G50. I bought a D270 with an external "Jamaica Box" SCSI array. It cost more than my house at the time.
A company I used to work for used this server to support their sales and inventory systems. They had around 100 users on WYSE terminals in the office and several at the warehouse.
This video resonates so hard with me. I worked as a digital analyst for a large financial services company for about 4 years. We had applications running code written in the 70s running on old IBM mainframes, and we used to joke that we probably had at least one random computer stuffed in a closet somewhere running a critical process.
Great video. I worked in an IT shop in the 90s that used pretty much the same K class machine. Will definitely take issue about businesses not needing uptime as much in the mid 90s. Even in 1996 we heavily relied on our Oracle ERP business system running on HP9000s to ship product and to get paid. I do remember CFO angst about the money to upgrade the memory!
Even the plebs (me) on the office floor got to be familiar with the character of these things, as all the problems would filter down to us. And management was always keen to brag their systems. Brings back happy memories.
I used to work with a HP9000. It was only turned off 1 time in the years I worked there and all I ever had to do with it was run the daily backups, It worked perfectly! We also still had the HP 3000 it replaced but never needed to hook it back up.
I was a very lucky engineer at the start of my career in 2006, who got to work on a pair of these. They had shared storage boxes, and SCSI attached in chain between them, and they were set up as a cluster. By that time HP-UX 11.00 was available for them. As you said, the software was specifically written to run on them, and not only that, written to support HP's clustering technology (HP Serviceguard). The systems were actually 2 complete cabinets, with the top part being something similar to this, and the bottom part being the storage enclosures. Using VFS you would create a RAID1 mirror of the disks from one box to another and then you would create the logical volumes. I wonder if they are still in use today :)
After I graduated in 2007, I did a extension program in one of the top Brazilian universities, and in the department I was they had some of these. I remember two are running openvms, one HP-UX and another NetBSD. Running custom software and with HP risc workstations as clients. One thing was these servers are almost always part of a DECNet cluster or a HP-UX cluster. So if a PSU failed HP would come and change while the other servers in the cluster could keep your operations running. Almost no x86 servers from the same era had two PSUs.
I think my workplace had one of these with a bank of modems for dial-up telnet sessions, when I was a child LOL! My mother worked here too, it's kinda how I got into IT... take your kid to work day(s), and I remember this beast in the server room and a Sun Solaris workstation for the Xerox electronic production printing press in the basement, and being absolutely fascinated. My fascination with all forms of Unix was dormant for a few years but then I discovered Linux around 2004 and haven't looked back since.
What a beautiful machine. The history of HP's RISC-based systems could be a documentary in and of itself- HP really did build their own little world for themselves with their server and workstation architecture- and some of those machines are still striking today. Great video sir, rock on!
I used computers like this in the early to mid 90s regularly in university. Based on my experience with them they didn't have redundant power supplies because they were actually good, not because we tolerated more downtime. There is more redundancy in servers today because they are garbage quality in comparison. Today we address downtime by adding more garbage machines instead of just building better ones.
I used to be System Administrator for one of these back in the mid 90’s when they first came out. The FDDI port wasn’t LAN access - FDDI was used to access storage pools. We had a few of these servers running in parallel being highly available clones of each other all hung off an FDDI ‘Ring’ which then also had the EMC Storage Array hung off the FDDI. You would probably now call it a SAN. Even then at this financial company I worked at it was a 3Tb disk array (it cost us around $3m from memory) which was effectively a couple of 42U racks stuffed to the brim with disks. The FDDI would interface with the storage controller in the rack and broker the discussions between the servers and the disks.
So before FC-AL (Fibrechannel Loop) people ran that over FDDI rings? Is there any any any link you can find me on that? I never heard of it but now my brain tells me i need to be able to support it, just in case i'd go back in time and have to work.
And - on the console port - this is still used today on all Cisco ISP scale routers (ASR/NCS/etc) - all have async console ports for low level management/booting/etc.
I think - but may be wrong - on the Cisco ACI systems the console ports have become more dysfunctional because of how they slide in the management layer.
@@udirt Hmm - not familiar with ACI systems - the only routers I've worked with are the ASR9K and NCS5K family. (Obviously Cisco has a ton of products, so it'd be no surprise that the management of devices would vary) - but you'd have to admit that a basic async connection is a low overhead way to manage something. :)
Nice find. I miss servers with proper serial consoles. I do work with AIX servers which have them, but even those hide most everything behind the HMC web UI.
Some pie-in-the-sky, never-gonna-happen ideas from Commodore in the months leading up to the death of the company had the future Amiga running on HP/PA RISC chips.
My dad worked with these up until he retired in 2016. The company he worked for still has them for back end inventory type stuff. I remember them upgrading the one they had in approx 2008 or so. You couldn't buy new parts for the most part so they bought an old one from say ebay or someplace and stripped out the memory and cpu's to max out the one they had. It still worked well and was as fast as a modern server for the job it did with a lot of that due to how well risc chips work for certain applications. The one Achilles heel it had was that it was so old there was no parts or support available from HP anymore.
The age problem of sysadmins are not as bad as with mainframe. I'm 42 and have installed and managed HPUX, Solaris and AIX plus other Unix dialects. And also had coworkers the same age as me do the same..
Exactly! I hate it when others start playing the ignorant ageism card, saying no one knows how these systems work because everyone that use to use them has all retired!
It's a real discussion point in the mainframe world - I'm a UNIX guy, but our mainframe team is all retirement age and they're looking at pulling me into a hybrid role since half of the Linux systems I maintain run on the mainframe. Going from Linux to proprietary UNIX is easy, especially if you spend a little time in OSX. Mainframes are a whole different beast.
My company is still running an HP9000 D250 in both the US and Canada. It still runs our old ERP software from back in 1998 and is happy to be chugging away in our server room. I loved those long gone days of being the HP-UX guy and was happy to see that maybe I can pick up some side work with whatever knowledge is left in my head.
Interesting! Reminds me of the MicroVAX II - not the narrow MicroVAX, but the fat box about the same size, and similar interfaces - SCSI, etc. Neat times!
I ran one of those when it was new. We used it to run billing software for private ambulance service. The software was originally ran on a MIPS vax that I nursed along briefly before it ran out of storage capacity. We purchased the 9000 and ran it with a 3com 48 port terminal server which was attached via ethernet and used IP. We tried to run the ambulance dispatch software on it too, but it proved to be too much. We went back to running the dispatch on a McDonnell Douglas PICK mainframe and then exported the data over to the 9000 for billing. Back in the day even more than now, if you pegged a UNIX machine out at 100% cpu you were in real trouble. That was the only issue we ever had with it, HP made fantastic hardware in those days.
It would be very unusual to use just a dumb serial terminal with an HPUX server. More common would be to use an XWindows terminal (XTerm) that could render the the desktop, or alternately your own workstation set as the terminal for the server. Source: worked as an engineer at Nortel in 90's where were based on HP workstations. I had an HP-9000 workstation (not a server) under my desk.
There's a familiar name, Hi there unixnerd! Back in 2000, then again in 2002 I bought a HP Visualize Workstation from you! Back then I was working for HP and as an employee was sent a CD pack with the latest software. Had the privilege on working on some of the most powerful Superdome systems back then... ... now then, who remembers HP Apollo Domain/OS? 😉 Now I really am getting old, nowadays trying to keep up with the cloud native kids!
@@matthewfranklin7541 We had Apollo Domains at uni (Robert Gordon's Aberdeen) in 1986. Used for SPICE, chip and PCB layout. Felt space age back then. Lovely kit. Awesome keyboards. Superdomes, wow. Drooled over them. HP still make good kit, my PC is a Z800. Also a BMW guy (the old ones) and found that BMW designed the Z series case for HP!
@@unixnerd23 Oh wow, yes that's how I used HP Apollo's at Uni! FPGA/ASIC design using HDL - amazing to think that that technology existed even back then. 😊
At my office, we used to use tape backups every night on those small DAT tapes, and we had the larger ones that were 36GB each, we used to have to do a backup every night on a rack of them. There are a bunch of old ones laying around here still in an upstairs closet.
11:30. I think the use of these ports is pretty obvious considering the label right beside saying "Workstation Multimedia". The "microphone" and "headphones" logos are pretty self-explanatory too. If this was used as a workstation (and not as a server) it must have been audio capabilities for multimedia content work.
When I was an undergraduate at MTSU, we had an HP-9000 v2200 server which served the whole campus. Everyone's email and a good chunk of homework was done on that system, which was named "frank". I was a computer science student, and so I was pretty well acquainted with the system. It was pretty cool to log in, run "who", and see like 1500 active users scroll by. I also played with a little assembly code for it. Not a bad instruction set, really. This one is like a scaled down version of the v series, but it still brings back some good memories. Thanks for sharing!
True, but I had a drawer full of scsi cables and terminators back in the day. I can't say where I would find one now except online. I do not know a single person or server (or old disused desk full of old crap) that has one in my current environment. That is most likely the context he was speaking from.
When I made the switch from OpenVMS to Unix back in 2000, the company I joined had a pool of these running HP-UX with us devs spread across 3 of them, 3 for test and a couple of others as Oracle database servers. I can't remember how much memory they each had but they all had 4 cpus. I remember them being super reliable and we used them for many years, then a few Itanium boxes started arriving and not long after that we switched to developing on Linux, maintaining just a couple of HP, Solaris and AIX boxes as build and test servers. It was good to see one again so thanks very much for this !!
We had one running Lawson Financial software, it was pretty much a turnkey system, I think we got ours around 1994 so it would have been a 32 bit model. You spent a lot of time talking about a "single power supply" but the power supplies for those older systems were seriously more reliable and even when redundancy is called for it may well have been built into the single power supply. Do not make the mistake of judging older systems by how things work today.
@@davidjames666 And I noted that the required redundancy may have been built into the power supply itself. The designs of yesterday are not the same as the designs of today and hot-swap was the exclusive realm of the Tandem non-stop computer (our Data Center had one of them too). That everyone now does hot-swap is probably because the patents by Tandem expired. That said, it is still a supposition on my part and not based on knowing the design. What I do know is that we never had any service calls made to fix or replace anything on that computer during the years we needed it. Like most HP equipment of the era it was extremely reliable. It was however just used for single application (Lawson Financial).
@@davidjames666 Our Data Center had a Uninterruptable Power System backed by a generator large enough to power the whole building not just the computers and that meant that I had Novell server at one point that had not been powered down for more than three years (updates didn't even require a reboot, MS should have payed attention) and even caught a secretary who had not logged out or powered down her IBM AT PC in over six months so she didn't have to change her password on a regular basis. The idea that redundant power would be required to keep a computer running because someone would flip it's breaker switch by accident was not on the radar of Data Centers at that time. A lot of the lessons we take as standard today had not been learned yet back then and yes we made those mistakes and corrected them in future projects.
Cool stuff, very nostalgic, I think. I worked for big fruity mobile phone company in the UK in the early 2000s and they pretty much ran their entire customer service system on rows of racks of K class servers in two data centres. I even have some photos of the racks with the servers in. They had big scsi disk expansions underneath that were almost half the height of these already huge servers. Getting them into the racks was fun!
Definitely the time when you go find an equipment lift because you'd need four people to lift one of these window AC unit impersonators. This was back when literally every server was twice as heavy as it looked like it should be. Lots of copper in those beasties.
Wrong. RISC is indeed an architecture, just as CISC is. cs.stanford.edu/people/eroberts/courses/soco/projects/risc/whatis/index.html#:~:text=RISC%2C%20or%20Reduced%20Instruction%20Set,in%20other%20types%20of%20architectures.
@@orbitalgolem91 From your own link: "RISC, or Reduced Instruction Set Computer. is a TYPE of microprocessor architecture" Key word emphasized in caps above. It's a TYPE of architecture, not an architecture in itself. Synonym to "type": Category, Classification, Family, Label. You could rephrase that sentence above to: "RISC is a category of microprocessor architecture" if it helps make more sense to you. An architecture would be things like x86, ARM, Risc-V, PA-RISC, MIPS, etc... They all are encompassed by a classification/type like RISC or CISC. Don't shoot the messenger. I didn't invent the ISA acronym (Instruction Set Architecture", where the term "Architecture" comes from in the context of computer science.
en.wikipedia.org/wiki/Instruction_set_architecture#Classification_of_ISAs RISC and CISC are mentioned there as classifications of a computer architecture. RISC and CISC are meant to classify an architecture's complexity.
4 years ago, I was a Junior Linux Admin at a fortune 50 company. We, at the time, still had a relatively significant HP-UX compute footprint running business critical apps. I never dabbled with the platform, but it was fascinating as a legacy OS. As a side note, my current company retired it's final Solaris server recently ;) we are largely an Oracle Linux shop with over 13,000 VMs.
I worked at a large print shop around that era that came with one of those. Essentially it came paired along a large Heidelberg offset printer. The thing did spooling and other things. It also came with its terminal display that ran a postscript server that allowed us to push, pause or cancel print jobs on that huge machine.
It was very rare to connect morr than one modem. These were often used to run manufacturing floors, and science research. The serial and parallel ports were typically connected directly to machines and/or sensors, and the server would sit and process that data.
7:15 if only I can say I still do that now, 😩 I always thinking go only digital is bad and this is one of best example of it, sure it's will take time but it's still got the job done than what we do now, well I still use half it manual so it's better now.
Clicked on this because it was very reminiscent of the type of servers the military was running back when in joined in 1998. Not sure we had this exact model, but not too dissimilar. As always, great video and good job breaking it down (without actually being able to break it down).
Many servers of the day were used as high-end workstations. That's what the graphical head was for, not "managing the server". That's also what those blanked out audio ports were for.
Proper rackmount kits were available for these, I’ve converted bunch of them back in the day. Internal modem was mostly used for software called Predictive, sort of early call-home solution. These servers were really popular - most of the banks, telekoms etc. had ton of them.
I supported 8 of the K's, email, database, file storage, backup.....I remember having to upgrade memory for the email server and it was 100$ a month added to the lease.... All were in HP racks with the disk arrays below them. I recall one of the leases the total was over 100k. $$$. And....the 64 bit s/w ran slower than the 32 bit. but could access more data. So it was still quite early for 64 bit systems.
This is correct. There was triple redundancy inside that power supply. And I'm later models you could order them with A/B power connections, so that all three sets of international (to the supply chassis) power circuits could switch between A/B.
That vent design around the case made me think it was something from mid-to-late-90s IBM at first. Edit: half way through video; I noticed the tower PC behind it and now I see just how big this thing is. WOW!
I am VERY interested on what servers from back in the day were actually running and how they were used on a day to day basis. Does anyone have any links to other videos that go more into detail on the software/OS side of things?
Thanks for this little visit to the past. i need to admit it makes me feel really old since we replaced many other servers back in the days with the then new K-Classes.
Wow! I had to deploy these servers K9000 RISC (10 of them) back in 2006 for a legacy platform because the customer couldn't upgrade to a newer platform.. Wow...
The way Colin gestures at everything with just his hand on camera while talking, never showing his face or anything else, makes me want to imagine that ALL there is to Colin is a talking hand
I’m almost certain that the FDDI board was used to access a Storage device. If such were the case, there is likely a twin of this machine that would have been in warm / hot standby mode if the app was mission critical. It would be interesting to get a recursive listing of the file system to determine what this machine would have been used for. If a /usr/local and/or /opt directory exists, that would be a great place to start spelunking. The timestamps of the files would also reveal how this machine was used. For what purpose? None whatsoever. Just because it would be fun.
This is the type of machine that would be used in a retail store setting for inventory DB and to run POS systems and probably dumb clients via direct serial connection to do lookups etc on menu based application to that DB.
@@adventureoflinkmk2 Point of Sale, but almost always also known as "Piece Of Shit" system. My own personal experience with POS systems is that the corporation always hires the cheapest bidder to develop the system and it's always garbage that looks and operates like a Geocities website designed by an 8 year old from 1997.
I appreciate the video but many things mentioned were wrong. It's still common to run servers headless today. The IBM hardware has a HMC (hardware management console) you use to partition the server, start/stop and other maintenance functions. For servers of this vintage, you'd either have a dumb terminal attached or more likely you'd have a terminal server attached so you could still connect to the console from your desk. You were correct in calling it a pedestal server but servers of that type were typically more overgrown workstations than servers. PA-RISC servers had their own flavor of media instructions that was a thing before the first Pentium MMX ever came out. Depending on the use case, HP/IBM/Sun systems of that era were also clustered as hardware was fairly reliable but nothing like today. We'd periodically get server reboots for no reason. The main memory had ECC but the cache on the CPU didn't have ECC, so it was a bit of a known issue. Since memory had low density, you could interleave banks of memory to get better performance too. I preferred Solaris but the HP-UX boxes were also OK.
I have a bunch of K-Class Core IO boards cluttering up my garage. I used to work on these machines and the big K580s were my favourites as the CPU cooler designs were amazing. And The way they sounded when powering on was like a jet engine!
WOW This brings back memory. We were using the called "Little Sister " HP 3000 until 2018. Long time no see this logon screen. Thumbs Up and more of these
My first Unix experience was on HP-UX. Our college workstation & server hardware was a few years older than this 64-bit server, but it was a wonderful time to be learning & working in the IT world. You needed all those modems for all the customer dial-ups at your small city ISP. This was the early public/commercial Internet too. AOL let all there users on the wider Internet in 1994 and things changed immediately for those of us already hanging out there.
That was considered a Modern server back in the day. I supported plenty of them. My history point of reference for servers (mini computers) was the Sun 3 from the mid 80's.
Back in the day when HP was a serious player in the enterprise market and had its own 1st rate version of Unix. Now HPE is just an x86 blade rebadger and HP itself just flogs cheap made in china consumer junk. All thanks to a bunch of CEOs that only cared about short term profits.
I wouldn’t say they are a blade rebadger - far from it. HPE still designs and makes custom hardware, whether Intel or otherwise. And don’t forget they bought both Silicon Graphics and Cray a few years back too.
Free Geek's eBay auction for this system ended at $455.
Hopefully local. I would hate to ship it.
Yea with $1,000 - $2,000 shipping I'd guess. Quite possibly even more $.
@@AnthonyGoodley remove a zero from that. If you strap it to a pallet and ship it via ground it's not ridiculous.
@@joeofloath Sure, but would you really want to ship a 25 year old card laden box clad in brittle plastics on a pallet via ground? Survey says pay for the next tier up if you want it to live.
It was local pickup only, which is probably why it didn’t sell for more
Oh wow, we used to use HP 9000’s in Mission Control in Houston to support mission operations of the Shuttle back in the day. Cool.
(By the way, fingers crossed for Artemis I 😄🤞)
Seeing how they've "restored" the Apollo Mission Control (it's mostly show, real consoles with modern hardware inside, like LCDs instead of CRTs) makes you wonder why they won't stick to this or at least part of it for historic reasons. Especially when they spent ages figuring out the correct type of carpet the room originally had...
I mean, sure you can emulate the entire Apollo era complex at Houston on a desktop these days but that's not as cool as the real deal imho.
Did they dial in with the spaceshuttle?
@@thorsteinj I saw some video of this, the LCD's stand out like a sore thumb, they've ruined 'em.
@@EdwinNoorlander No, that was left to ET
I remember like 10 years ago while working as a backup consultant, I went to work on backups for a government office they where in charge of public water, they had exactly this model and it was absolutely forbidden to get close to it, it was so old and nobody there really wanted to touch it, they had cones and “police” tape it was funny but it was pretty serious thing. They warned me that there was no spare parts and support for it and the app was a legacy one so they couldn’t just update it or migrate since the guys who developed it just disappeared many years ago. Funny days as a consultant.
heck... hp-ux was so damn backward compatible, as long as you'd have stayed on PA-RISC you could have run HP-UX 8 code on even the last series of HW and OS... always sad when fear and ignorance cross paths. They had people who knew, so they didn't bother to learn and then they're left ... putting cones around it. lol.
hey, whats the work of a consultant?
@@udirt To be fair though, I heard unix consultants charged out the ass so if they didn't have an inhouse guy know knew the thing. I have seen this fear from old PDP systems to even modern web servers.
It's only an issue if they're too cheap to find a decent developer or 2 and pay them a reasonable salary to migrate whatever it is to something modern.
The old "Let's hire an external consultant to write a bespoke business critical application and not demand the source code or comprehensive documentation because it costs extra." - classic! What could possibly go wrong?
I disassembled and scrapped about 40 of these suckers in 2005. GE donated them to the local refurbusher/e-waste place I volunteered at at the time. The CPUs are freaking huuuge
horrific... i guess it was was your job and all but still :(
GE Appliance still has three of them in production. I just fixed one of them
@@kedamono2562 obsolete tech becomes recycled or gets dumped in a landfill. This stuff has very limited uses in modern day. People out there who collect this junk and do anything meaningful with it are scarce and when they die the same thing will happen to their collections. Yeah you paid $400 for a outdated e machine but your next of kin is going to dump it to the highest bidder or throw it in the trash
perspective, a 1996 system in operation until 2008, that is 12 years. My home server is an i7-970. made in late 2010/early 2011 and is still running today and likely for some more time, also about 12 years old. No one would flinch about a i7-970. I recall working in the 2007~2012 range and losing my mind whenever I saw a min/late 90s server running something like NT and making it my mission to get rid of these relics. Today we have server 2012 boxes that are starting to be replaced and I feel like 2012 was just yesterday and I'm getting old.
The difference is the OS. A server from 96 isn't going to run an OS from 2008 very well, or if at all. Never mind the many hardware changes in those 12 years. Compared to now where an i7-970 will run pretty much whatever you want and if you need something hardware wise its board doesn't support, you can hop on over to ebay or where ever and most likely find exactly what you need.
@@osrr6422 at my Workplace there are Servers still running Xeon X5650 which are from the same generation.
Have to remember that hardware was improving at a much faster rate back then compared to today. A 10 year old 486DX in 2000 was no where near as capable as a second generation core i series workstation from 2010 is today. My main desktop is still a dual dual core Xeon 5150 with 32gb ram from 2006 with hardly any upgrades. I could certainly use an upgrade but it is handling Windows 10 and kubuntu with no issues so no big incentive to upgrade yet.
@@kathrynradonich3982 I agree and I wasnt trying to imply that these 90s systems were actually still good in late 2ks but more to reflect on how much has changed and accelerated since the 90s.
The main thing is single user Vs. 1000s. Just like my 1993 386 was still tolerable for games in 96, the i7 is still ok. But for a lot of users a desktop grade system lasts one generation, while servers need to be good for 2-3 even IF you start replacing after 3yrs. And what always pissed me off in the past: people never upgraded Ram at all, ignored broken raid bbus / never properly enabled caching and then they throw out systems that never ran well... Like Xeon 5500 series that could easily fit 96GB and they run one as their SBS server with like 8GB Ram.
We used the audio ports (missing in your model) for condition alerts over loudspeaker in the server/PBX room(s).
Could they have also been used with those weird modems where you put the handset up to a speaker and it listens for the data as a series of tones? Like as a backup of a backup of a backup or for backwards compatibility with really old stuff that used that?
@@forgottencameras those modems with "acoustic couplers" (where you put the handset up to a speaker) also used the standard RS-232 ports, on this server it would have been plugged into the "external modem" connector.
In older HP-UX releases, you could even link audio to /dev/null which created some funny noise in the datacenter ;)
I used to have one of these in my house. I got it from a school on the cheap, and used it to manage my file archive, DVR, and ran a BBS for Text Games and classic computer game files. Man, what a time.
@Peter Kurten Not bad honestly, at the time only $230.
Nice
I could not agree more I had one for 5 years I ran a BBS as well great times!
@@jonathankovacs1809 I miss my BBS. You may have heard of it, if you were in the BBS scene. "Divided by Zero" was the BBS I ran.
@@ninja011 I do recall a BBS "Divided by Zero" being talked about in my company's break room at that time. those were the days I miss them. The young folks have high speed internet today and smart phones its great but they missed out on the BBS glory days!
I worked for Sun Microsystems in the early 90’s. This was a “competitor” to Sun. Your presentation was fun to listen to but contained several factual, how would you say, deviations from reality. But that’s understandable and I commend you on keeping this part of computer history alive.
HP made incredibly reliable hardware. The fact you can turn it on in 2021 and get a prompt supports that statement. Hence the requirement for a single power supply. Electrical sources were less reliable than the machines themselves.
Turning off a Unix based machine and then turning it on again at a later time was always a very stressful event. Disk corruption was not uncommon. Then you’d have to perform an fsck and hope that repair was possible. For this reason, you’d always want to “sync”, twice for good measure, to flush all the in-memory writes not committed to disk, prior to a “shutdown” or “halt”. On a server if this type you’d run a shutdown command that would send a message to each logged on user giving them ample warning to save their vi or ed sessions and log off.
1996 seems a long time ago but communications protocols were not very much different from what they are today. Client-Server computing was a thing. So you could be running something like Ingres (precursor to Postgres) serving multiple client programs.
The multitude of serial ports could be used for modems or more likely “terminal servers” such as a Xyplex that gave you access to the machine via a telnet session. Running a bank of modems into a terminal server and then to the actual server was also common.
To put things into perspective, I was running Perq computers with an operating system called Penix (no kidding) having a portrait graphical display (1024x768) and a bit pad (a deluxe optical mouse like device) with a whopping 10 mb Harddisk in 1983 or so while at McGill University.
HP was rarely ahead of the technology curve when it came to workstations. But the hardware was solid. If ever you get another one of these, go ahead and open it up. There is nothing to break.
By the way, the CLI on that machine was definitely “sh”. “sh” exists on all Linux machines to this day. RaspberryPi users would feel quite at home on that thing.
The 9000s were workstations as well as servers. The multimedia was used your application had a multimedia element. My fathers company used then for acoustic testing in the 90s.
@@jamesjross yes indeed. SGI had similar footprints. Imagine a workstation like this under your desk. Imagine 4 in close proximity. The noise must have been deafening.
@@AllanKobelansky it was in a room attached to a chamber of sorts.. It was 20 years ago so I can hardly remember. They had 4 Sun workstations for PCB designing - that was more like 92. I can't remember it being loud - I'm guessing the terminals where in a different room
I currently work at an HP campus and believe it or not these things are still in use. I find it entertaining to see all of the old 90s tech around that site
@@silversurfer420 and why would anyone need to change them? If the box is doing what it’s supposed to do, there is no need to replace it. I’ll wager that machine will continue to work till 2038.
I worked for HP back in the '90s doing software dev on the 9000 series with HP-UX. PA-RISC was a very unique architecture at the time.
I really hated having to deal with HP-UX. I especially hated dealing with HP's pre-boot environments, because it was always a fight to get them to pick up and use the boot media I wanted them to use.
Sold and supported the same. I came from the MPE side with HP3000’s. These were the smallest server system we made at the time. I did have a HP 712 desktop “workstation” on my desk and I really learned HP-UX that way…
I remember there was some compatibility with Power/PPC architectures as they started as a joint effort, am I correct?
@@tomsaltner3011 Completely different.
I worked at Gateway 2000 in 1996. Memory at that time was $50 a megabyte, so to upgrade this to 4gb would have cost $193,600 1996 dollars.
And I'm pretty sure that this is a low estimate, because like today, servers used to have different, more complicated and higher quality memory than consumer or even workstation machines. I wouldn't be surprised if the actual number was much closer to half a million.
Edit: We see the price in the video later on: 256MB = $12000. A full 4GB adjusted for inflation would then be at ~$335000. So not quite as bad as I thought, but still wow.
@@craigjensen6853 you could, yeah, but I'm Not Sure you'd want to buy a Chromebook. And for me that's Not even because they're that Bad. I wouldn't want to buy a Chromebook because their ads are so useless. "Tired of having to Charge your Laptop? Switch to Chromebook". That doesn't make any sense. As If the Chromebook ran on a Solar Panel or some other Kind of bullshit. No it doesn't, it's a regular dumb Laptop. You'd have to Charge it less often than say a ThinkPad with an AMD Ryzen, but the ad suggests you'd never have to Charge it. Same with the Updates. They try to get to the specific Kind of User Base that doesn't know how to disable automatic Windon't updates. "Are you tired of your Laptop rebooting while playing a Game to Install an Update? Switch to Chromebook. Because then you can choose when to Update". Like duh, I can do that with Windon't too, Not that I'd want to because I don't really use it, but it's Not that hard. Also this ad suggests you can Play Games on a Chromebook. You could If you were to Install Chrome OS on a powerful Laptop but a Chromebook itself is really only suitable for Things Like Solitaire or maybe among us.
A K450 with like 4gb ram was something like $270k, not counting the yearly hardware and OS support - in about 1998/1999.
To be clear, mainframes are used for a specific reason, such as reliability and scalability. The banking sector cannot afford to lose a single transaction, much less "move fast and break things."
100%, and back in the early 80's when I was still a toddler my mother being a single parent(my parents split when was 2 1/2) was an oncall 24/7 ATM/computer technician to make extra money for the bank she worked for(First Citizens Bank & Trust Co.), and even at that age I can remember her grabbing me out of bed in the middle of the night to make a call, and some of those systems she worked on at the time with 10-40MB HDDs the size of a small lawn mower engine where still in use well into the late 90's simply because of cost, and the fact they could not loose customer account data, where some customers had $100,000's, or even millions of dollars in them that had been with that bank for well over 1/2 a century at that point.
Also mainframes are really good at transactions, often with reliability down to the level of hotswapping CPUs and RAM
@@CommodoreFan64 EEVblog has a teardown of one of those massive HDDs. Crazy stuff. Was 10MB off the top of my head.
@@TomStorey96 I've been a long time sub of Dave, and saw that video, it's indeed crazy how big that HDD was, and to think these days we carry around 128GB MicroSD cards in our phones like it's nothing.
Also System-Z is a HUGELY profitable enterprise for IBM so they're going to continue making it. And they're actually doing a lot to support training "the next generation" of Mainframe operators (they even have a lab you can access for free if you're interested in it)
It is pretty much guaranteed that you have interacted with a Mainframe today, they aren't going anywhere any time soon.
I ran these machines. The reason it does not have redundant power supplies is this is a lower end model, it likely didn't come out of a server room. That I/O panel gives that away, It does not have the color workstation options which was another market HP was after. A lot of these were used in industry and control rooms to run experiments and production equipment, if it were it would have the additional I/O boards and a large color monitor. The console connector (often this was duplicated on the front panel on rack mounted machines) was mostly used to monitor the boot up. These machines don't have BIOS so this connector was pretty important to keep them running. I used to access the console with my palm pilot to equipment like this. At the time this part of the market was pretty heavily competed by not just IBM and SUN but SGI, DEC, Data General, and a number of German and Japanese companies all had mobile ped models. A number of these machines were market specific but if you knew unix you could sit down at anyone of them and get some work done. I used to say that there were more similarities between any two varieties of unix than any two major versions of Windows.
I thought I remembered our HP 9000's being about 5 feet tall. Is that what I was thinking of, server room model?
Back when HP was a decent company and made bloody good hardware. This takes me back. Thank you.
for PCs the competition is tough for everyone, but I think that HPE still designs and manufactures excellent servers
They are absolutely ROCK SOLID servers still. I can't attest to their consumer market products but the two HPE personal servers I've dealt with the past 7 years have been on 24/7 with a minor hiccup of the integrated SD card reader.
Started my IT career as admin in '97 in Germany, and one of the first servers I was responsible for, was this HP9000 K class. It was the main server for the whole SW development team, later we got an add-on tower with 10 (iirc) 18GB SCSI harddrives, alone this did cost a fortune at that time :-) We decommissioned it in 2010 or so, maybe it is still somwhere in the basement, besides the MicroVAX :-)
"It's a Unix System! I know this!"
Related to this quote from the film, the GUI she was using is a real file manager called 'fsn' (Fusion). It runs on SGI machines.
That was for an SGI system of the time.
Back in the day (1985 thru 2000), these systems were all over the place. They required real skills to implement and manage. Now everything runs on Linux servers and things are relatively simpler. The only thing complicated now is the Web UI layer. (Which hopefully at some point will go through it's simplification phase.)
SaaS, basically.
HPUX, AIX, or even dual universe unix flavours - once you know the core, it's all good. Ported a radio dispatch system from NCR Tower onto HPUX back in this time - still not no DEC VAX environment but still, fun.
Oh and early HPUX had an error code of 8008 - that ended up getting changed due to early PC mentalities in marketing overruling that showing on a display due to calculator word legacy.
And today, you fire up an EC2 instance and you have your own Server in the cloud as an Infrastructure as a Service.
Really? Thought sam made config pretty easy
I was REALLY hoping to see this running HP/UX Common Desktop Environment, it's my absolute favorite 90s desktop environment
I've worked on a number of them in a dev datacenter and I've never seen a single one running their Common Desktop Environment.
@@mndlessdrwer It was always installed. If you ever saw it depended on the usage scenario - most devices probably ran just databases and server SW for rich clients. XWindows desktops for terminals or X emulators were seldom seen in the nineties.
I know for an absolute fact that as recently as 2018, SOME hospitals with pneumatic tube systems to transport blood samples and other sensitive items, were STILL using servers similar to this, sourcing spare parts from friggin nasa, just because the entire system was relying on some proprietary OS from 30 years ago that has no possible hope of ever being reverse engineered or ported into a modern OS. It was only after literally investing tens of millions of dollars to basically gut out the entire tube system and replace it that the old beast of a dinosaur finally came to a rest.
Are developers that can reverse engineer something really that hard to come by? I'm sure the tube thing isn't that complicated, and could've been reverse engineered for much less than "tens of millions" of dollars. I would've done it for $1M easily!
@@gorak9000 $150K to reverse engineer and port to Linux, several million more to shepherd it through FDA device approval. Been there. :-)
@@gorak9000 do you really want to be responsible for anything a life depends on? I decided when I was like 10 (started coding when I was 5) that I wouldn't ever take that responsibility personally.
@@forbiddenera The pneumatic tubes they're talking about transport samples from the hospital to the lab. If the tube breaks or stops working, the worst that happens is they take a new sample and someone walks it over to the lab instead of sending it down the tube system. It's not directly connected to the patient. It's the same kind of thing some stores use for moving money to the cash registers and back, and some banks use at their drive through tellers for deposits. Not sure if they're used anymore, but most Home Depots have them.
@@gorak9000 I suppose but extra time could still cause problems..also I assumed it might be used for other purposes in some hospitals like sending med ampules or something, I have no idea. Though, even if you're 100% right, hospital admin/insurance might disagree or require a certain level of service or guarantee or something. I dunno, I'm paranoid enough doing financial stuff lol
The lack of redundant power supply is most probably due to that this is a small workgroup server, the larger HP servers for datacenters had redundant power supply's...
I was system administrator for the a couple of K460 versions from 2000-2005. The chasis was the same size of the one in the video. That was 20 years ago, but I don't remember having a redundant power supply in that chasis. For sure the bigger ones did. We did have the HP AutoRAID 12h which had triple power supplies, 12 drives, and dual controllers.
@@timkr66 and redundant batteries (but i hated the autoraid). agreed that k-klass never had multiple PSU, but there was the fat brother in the 90s, i think T-class as SuperDome's business use case successor. never saw a T-class outside of history websites.
Nope, the K class just didn’t have redundant power full stop. It wasn’t a “small workgroup server” either. At the time of its introduction the only HP9000 systems that would expand more were the T Class, and I don’t think they had redundant power supplies either, although I think they did have more power resiliency than the K class.
Their rack-mount server power supplies were extremely loud and, well, less reliable than the massive single PSU. They were prone to letting out the magic smoke and causing problems. I suspect it's because they had to shrink everything down to make it fit and it couldn't keep up with the cooling demands of the components in the PSU. Still more reliable than some of the modern PSUs, but finding replacements without constantly renewing a service support contract with HP was always a bit of a nightmare.
Man, this brings back memories... When I started working, I had access to this beast, along Sun Solaris servers, Compaq TruUnix servers, Digital Unix, Red Hat 6 clients... Being at work looked like being on a spaceship 20 years ago :D
Lol redhat - I had Suns and SGIs
Same! My first job out of high-school back in 1999 was working for a company called DCG Inc and they specialized in Alpha Linux clusters. I built a lot of Alpha systems, the biggest one being like a 20 server cluster for Bloomberg! I even had my own Alpha workstation that I got to use at work for things like AIM and Email. Those were the days!
This was the golden age of computing - those days were fun!
i miss the 90's@@markymark2648
My first Sys Admin Job had one of these and a bunch of HP LC2000 Net Servers. Brings back memories / nightmares, lol.
Thank god you never had to work on VaxVMS machines lol
Omg..I still have a lc2000 netserver in my garage. I need to get rid of it. Funny to hear it mentioned
This system was designed by the General Systems Lab in Roseville, CA. I was the system design lead for the corresponding entry level platform, the D-Class systems and I adapted the four to six processor designs of the K-Class to the dual processor D-Class systems. At the time, PA-RISC/HP-UX was used both by the Systems Technology Division for enterprise computing and by the Workstation Division for engineering workstations. After the acquisition of Apollo, there was a more concerted effort to leverage ASICs and engineering between the divisions, and this is reflected some of the core I/O hardware, like mouse and keyboard, but it was limited in the K-Class. The D-Class actually shared much more with the corresponding generation of workstations, as we defined a common two socket CPU/Memory design that allowed us to share much more.
Kirk Bresniker is the real deal when it comes to these platforms and he is of course spot-on with all his comments. HP sold a ton of the D and K-class hardware at that time, targeted for the OLTP customer using big-metal databases, and Kirk was right there, in the center of the action. I'm glad to read his commentary and happier that he has time to look at TH-cam videos.
So how did this crime happen that the L2 Cache was an optional component most people would not afford? You were selling castrated servers!
I got my first D-Class in late 2000s as payment for "i was never here" storage support at an insurance co, proving their EMC array was underperforming. I think I later sold it for a crate of lemonade, minus the cache module that went to my A180.
@@udirt Why is "you get what you pay for" a crime? And how is it different than paying more for a 5% resistor than a 20% resistor than came from the same line or an x86 CPU SKU with more cache than the cheaper SKU with less that came from the same wafer? The D-Class offered a 10X performance range over a series of in-package upgrades and that was really important to our customers since switching to a different vendor was a big investment. Whenever we introduced a new upgrade the biggest bump in sales was almost always for the prior top end because our customers always wanted one more easy performance upgrade, they were planning on their future success and we were helping them realize that future.
Thanks, @@curtthompson7279, I'm on a 665 day streak with the NY Times Crossword so every night I have between 5 and 60 minutes (depending on the puzzle) to also watch TH-cam while I solve and vids of old HP equipment are my hands down favorites!
100 years from now someone is gonna read these comments and think to themselves "All these people are dead now..."
this took ... sad turn
If TH-cam is around by then.
You have a better outlook on the human race then I would expect...
The worst part is than maybe are our newer generations XD
Damn
I'm a field server engineer and I've worked on these twice. They are so foreign compared modern servers. I still have a couple customers that still use them for legacy data.
The CDE login screen on the "graphical interface" is so nostalgic for me. CDE was also the default window manager for Suns Solaris OS and I used that in college. Well, I didn't USE it because I hated it. I always switched to KDE. Yes people KDE existed in the the late 90s and it was so much more familiar and configurable then CDE.
Also fun fact, XFCE one of the least resource intensive desktop environments currently for Linux. Started as a CDE clone. Then they got their act together and made it much better. ;)
For those of you interested in what CDE was like:
en.m.wikipedia.org/wiki/Common_Desktop_Environment
CDE is Open Source these days and you can run it on Linux.
Of course, what we really need is een Open Source version of 4dwm ;-).
KDE was bloated and buggy back then and is still bloated and buggy today. Worst DE ever.
@@ps5hasnogames55 maybe up to 5 or 6 years ago but not so much these days, GNOME in my experience is waay more resource intensive and far less usable, a nice minimal KDE-Plasma install is closer to xfce and mate in resource use, and is by far the most consistent in look, feel and configurabiliy.
Kmail does still suck though
@@mrmysteryguest Warning... long rant. "Minimal" KDE just doesn't exist. All of it ties into these useless bloat framworks and Qt is semi-proprietary trash. At least it looks nice and works somewhat like how a normal person expects a computer to work. Too bad KDE crashes on you if you blink at it wrong, and KWin is still a horrible window manager that doesn't work with mutliple monitors. Steam Deck customers are going to be in for a world of pain when they connect their shiny new hardware to an external monitor only for KWin to outright crash on them, or for their panels to go missing when they disconnect the external monitor.
GNOME's idea of a "desktop" is some abstract bullshit that's unusable for an average user. You want desktop icons, like every other desktop has? Too hard to maintain in the file manager so we'll just remove it and force you to install an extension, and if you complain, well you just shouldn't use your computer that way because "minimalism is the future". Tell that to every normal Joe and Mary who just want to get their work done and puts their most-needed documents right on their desktop so it's just one click away. Their developers are arrogant prats (like every other project that Red Hat usurps while parading around that they support free software, lol whatever). While GNOME doesn't have a tonne of framworks, instead it relies on (and is the reason every other distribution is forced to include) subsystems like Systemd, Wayland and PipeWire that nobody asked for. Classic Red Hat forcing everyone to use Linux how they want you to, right down to the init system. Funny how Canonical gets hate for making Linux easy with Ubuntu, but Red Hat gets zero hate when they're the reason your desktop has graphical glitches and random crashes (GNOME on Wayland randomly booting you back to the login screen causing you to lose all your work with no explanation sound familiar?) and your audio doesn't work (PipeWire, just like Wayland, is incomplete, beta-quality software yet they're still trying to force it on everyone), and has loads of security bugs (systemd and all of its tentacles and everything it touches and tries to integrate... and runs all on PID 1 :wink:).
Cinnamon is alright, they have the right idea of how to make a usable desktop and it was forked from GNOME 3 right at the start before Red Hat forced everyone to use insecure and buggy bloat software no one asked for like systemd, Wayland and PipeWire in the name of "innovation". But at the end of the day, it's still GNOME 3. Bloated. MATE and Xfce are lighter, but that's only because they're outdated. You show a desktop runinng MATE or Xfce to an average user, they'll laugh at it thinking it's from 2005. Of course there's still more like that one that Elementary OS has (wow it's just GNOME re-skinned to be even more iPad like... who cares) or that Pop OS Cosmic thing (it's literally just GNOME with more bloat extensions for window tiling, as if any normal computer user works like that).
@@ps5hasnogames55 for the last 17, almost 18 years I have been a Linux user, a distro Admin/Dev, minimalist, and for the last few years recently a more pragmatic user of whatever tool gets the job done.
KDE-Plasma has for the past few years been more stable than any other environment I have tried/used, including the self styled wm only setups I used in the past and does not require a massive investment of time and effort to be usable.
There is no such thing as a perfect distro or desktop environment, if there were, then there would not be so many available.
Plasma Desktop has evolved and is evolving to be among the most usable and configurable environments, both on Linux and FreeBSD.
The issue surrounding the licencing of qt may not become an issue as KDE has documented their plans should it become problematic.
Although I currently use Kubuntu, I would whole heartedly recommend ArchLinux+PlasmaDesktop to anyone, even to adventurous folks not afraid to experiment that are new to Linux
Pssst, it doesn't have a motherboard, it has a backplane. 😉
Yeah, there's a ton of misinformation and misunderstanding in this video that it hurts. Particularly as he rants about the power supply...
Not knowing the difference between DCE and DTE is frankly a big red flag...
This is beautiful. I worked on these and installed a few back in the day running HP/UX. All with FDDI. That card is single attach, by the way. Dual attach was also available. We also ran these in a cluster (actually active/passive pair), so the lack of a redundant power supply wasn't that big of a deal, since we would purchase two and have them failover. Such good times. I remember the failover code was version 1.0 and I worked with HP support to get it running. Very basic but worked very well. It included shared SCSI storage across two nodes, and the active system would seize r/w access with the backup system having r/o access. Also the days of yellow pages (precursor to NIS)... So much fun.
IBM are still selling AIX and IBMi systems in 2021, although they're rack mount only now. And you can still buy Solaris systems from Oracle. They're not dead yet!
Unfortunately
@@dupajasio4801 Why do you say this? They still do things that other OS's can't. AIX is renowned for it's very long uptime, and the RAS features of Power systems are second to none!
@@petergathercole4565 Oh just bad experience with those systems. We used to run D class version of this server for our ERP. I still have to keep it's Oracle DB running on 2003 and XP client. Also we still run Teamcenter on Sun and IBMi for ERP. We are too small to properly manage these. So it's all voodoo kind of DR etc. Perhaps these systems are good for some apps but knowing Oracle and IBM they always are overpriced or are missing some features that are better supported or implemented in Linux or M$. It could be that our apps suck on those systems. Just my opinion. Regards
Being a massive HPPA fan, I still have PA-RISC systems here at home and work still running. My oldest is a G40 :) Build date is 1989. I've worked on and setup K-Class's back in the day. (And T's and D's and C's and snakes, and the instruments)... don't get me started.
I also had a few, have one demonstrating at my channel, the line I never quite understood is the E series, single cpu server not that much powerful than a workstation and I think at the time undercompetitive in performance vs. a Pentium or Alpha cpu, I had an E25 fully loaded and it was silly to run 11i on it, it was so slow for a mid 90's server.
When I worked at Sprint, we would indeed put them in a rack on a shelf. Floor space was at a really high premium in our switch houses.
The second modem port is for an extra modem that was used by HP for support of the server.
yup
My still working D230 is even bigger and very heavy. It was used for about 20 years, for administration of several network elements of a submarine cable, between Italy and US
Given the "Workstation" label next to the audio ports, they are probably literally for using it as a multi-media one person workstation.
Yes, my dad worked for HP back in the day and they still had a few of these as workstations in the office.
There were workstation form factor versions of this line of computer. Those versions had the multimedia ports.
@@benjames866 Yeah, I've a couple of the C-Classes in my collection, same CPU as this badboy but IIRC they're Uniprocessor machines. And I can confirm the multimedia ports are for exactly what they're for today.
I remember the electronic engineering students at uni in the late '80s early '90s used HP workstations with all the multimedia bells and whistles for designing ICs.
My first big IT job was to source a replacement for an HP 9000 G50. I bought a D270 with an external "Jamaica Box" SCSI array. It cost more than my house at the time.
A company I used to work for used this server to support their sales and inventory systems. They had around 100 users on WYSE terminals in the office and several at the warehouse.
This video resonates so hard with me. I worked as a digital analyst for a large financial services company for about 4 years. We had applications running code written in the 70s running on old IBM mainframes, and we used to joke that we probably had at least one random computer stuffed in a closet somewhere running a critical process.
Great video. I worked in an IT shop in the 90s that used pretty much the same K class machine. Will definitely take issue about businesses not needing uptime as much in the mid 90s. Even in 1996 we heavily relied on our Oracle ERP business system running on HP9000s to ship product and to get paid. I do remember CFO angst about the money to upgrade the memory!
I was a field engineer at HP and back in the late 90s and early 2000s this server was all over the data center. Sure brings back a lot of memories.
Even the plebs (me) on the office floor got to be familiar with the character of these things, as all the problems would filter down to us. And management was always keen to brag their systems. Brings back happy memories.
I used to work with a HP9000. It was only turned off 1 time in the years I worked there and all I ever had to do with it was run the daily backups, It worked perfectly! We also still had the HP 3000 it replaced but never needed to hook it back up.
I was a very lucky engineer at the start of my career in 2006, who got to work on a pair of these. They had shared storage boxes, and SCSI attached in chain between them, and they were set up as a cluster. By that time HP-UX 11.00 was available for them. As you said, the software was specifically written to run on them, and not only that, written to support HP's clustering technology (HP Serviceguard). The systems were actually 2 complete cabinets, with the top part being something similar to this, and the bottom part being the storage enclosures. Using VFS you would create a RAID1 mirror of the disks from one box to another and then you would create the logical volumes.
I wonder if they are still in use today :)
I love Free Geek Twin Cities! I saw the posting and that it said ebay. Was a little sad but it makes sense since they would get way more $ for it.
After I graduated in 2007, I did a extension program in one of the top Brazilian universities, and in the department I was they had some of these.
I remember two are running openvms, one HP-UX and another NetBSD.
Running custom software and with HP risc workstations as clients.
One thing was these servers are almost always part of a DECNet cluster or a HP-UX cluster. So if a PSU failed HP would come and change while the other servers in the cluster could keep your operations running. Almost no x86 servers from the same era had two PSUs.
I think my workplace had one of these with a bank of modems for dial-up telnet sessions, when I was a child LOL! My mother worked here too, it's kinda how I got into IT... take your kid to work day(s), and I remember this beast in the server room and a Sun Solaris workstation for the Xerox electronic production printing press in the basement, and being absolutely fascinated. My fascination with all forms of Unix was dormant for a few years but then I discovered Linux around 2004 and haven't looked back since.
What a beautiful machine. The history of HP's RISC-based systems could be a documentary in and of itself- HP really did build their own little world for themselves with their server and workstation architecture- and some of those machines are still striking today.
Great video sir, rock on!
I used computers like this in the early to mid 90s regularly in university. Based on my experience with them they didn't have redundant power supplies because they were actually good, not because we tolerated more downtime.
There is more redundancy in servers today because they are garbage quality in comparison. Today we address downtime by adding more garbage machines instead of just building better ones.
I used to be System Administrator for one of these back in the mid 90’s when they first came out. The FDDI port wasn’t LAN access - FDDI was used to access storage pools. We had a few of these servers running in parallel being highly available clones of each other all hung off an FDDI ‘Ring’ which then also had the EMC Storage Array hung off the FDDI. You would probably now call it a SAN. Even then at this financial company I worked at it was a 3Tb disk array (it cost us around $3m from memory) which was effectively a couple of 42U racks stuffed to the brim with disks. The FDDI would interface with the storage controller in the rack and broker the discussions between the servers and the disks.
It would probably cost $3m just to cool those 42U disk racks...
So before FC-AL (Fibrechannel Loop) people ran that over FDDI rings? Is there any any any link you can find me on that? I never heard of it but now my brain tells me i need to be able to support it, just in case i'd go back in time and have to work.
And - on the console port - this is still used today on all Cisco ISP scale routers (ASR/NCS/etc) - all have async console ports for low level management/booting/etc.
I think - but may be wrong - on the Cisco ACI systems the console ports have become more dysfunctional because of how they slide in the management layer.
@@udirt Hmm - not familiar with ACI systems - the only routers I've worked with are the ASR9K and NCS5K family. (Obviously Cisco has a ton of products, so it'd be no surprise that the management of devices would vary) - but you'd have to admit that a basic async connection is a low overhead way to manage something. :)
Well, you need some way to type "transport input ssh" to a new device ;')
@@shana_dmr Haha - very true - or even more basically, transport input telnet (enabing ssh requires generating additional steps on the router) :)
Nice find. I miss servers with proper serial consoles. I do work with AIX servers which have them, but even those hide most everything behind the HMC web UI.
Some pie-in-the-sky, never-gonna-happen ideas from Commodore in the months leading up to the death of the company had the future Amiga running on HP/PA RISC chips.
The amiga 2000 ran a custom version of UNIX and Xwindows out of the box, from day one!
@@danielktdoran cheers! thanks for the correction. My bad!
My dad worked with these up until he retired in 2016. The company he worked for still has them for back end inventory type stuff. I remember them upgrading the one they had in approx 2008 or so. You couldn't buy new parts for the most part so they bought an old one from say ebay or someplace and stripped out the memory and cpu's to max out the one they had. It still worked well and was as fast as a modern server for the job it did with a lot of that due to how well risc chips work for certain applications. The one Achilles heel it had was that it was so old there was no parts or support available from HP anymore.
The age problem of sysadmins are not as bad as with mainframe. I'm 42 and have installed and managed HPUX, Solaris and AIX plus other Unix dialects. And also had coworkers the same age as me do the same..
Exactly! I hate it when others start playing the ignorant ageism card, saying no one knows how these systems work because everyone that use to use them has all retired!
It's a real discussion point in the mainframe world - I'm a UNIX guy, but our mainframe team is all retirement age and they're looking at pulling me into a hybrid role since half of the Linux systems I maintain run on the mainframe.
Going from Linux to proprietary UNIX is easy, especially if you spend a little time in OSX. Mainframes are a whole different beast.
My company is still running an HP9000 D250 in both the US and Canada. It still runs our old ERP software from back in 1998 and is happy to be chugging away in our server room. I loved those long gone days of being the HP-UX guy and was happy to see that maybe I can pick up some side work with whatever knowledge is left in my head.
Interesting! Reminds me of the MicroVAX II - not the narrow MicroVAX, but the fat box about the same size, and similar interfaces - SCSI, etc. Neat times!
I ran one of those when it was new. We used it to run billing software for private ambulance service. The software was originally ran on a MIPS vax that I nursed along briefly before it ran out of storage capacity. We purchased the 9000 and ran it with a 3com 48 port terminal server which was attached via ethernet and used IP. We tried to run the ambulance dispatch software on it too, but it proved to be too much. We went back to running the dispatch on a McDonnell Douglas PICK mainframe and then exported the data over to the 9000 for billing. Back in the day even more than now, if you pegged a UNIX machine out at 100% cpu you were in real trouble. That was the only issue we ever had with it, HP made fantastic hardware in those days.
It would be very unusual to use just a dumb serial terminal with an HPUX server. More common would be to use an XWindows terminal (XTerm) that could render the the desktop, or alternately your own workstation set as the terminal for the server. Source: worked as an engineer at Nortel in 90's where were based on HP workstations. I had an HP-9000 workstation (not a server) under my desk.
I worked for HPE and the customer we supported had K class machines still online and running in 2020.
HP--UX was a very nice unix, I still use it. Sold quite a few of those servers over the years.
There's a familiar name, Hi there unixnerd! Back in 2000, then again in 2002 I bought a HP Visualize Workstation from you! Back then I was working for HP and as an employee was sent a CD pack with the latest software. Had the privilege on working on some of the most powerful Superdome systems back then...
... now then, who remembers HP Apollo Domain/OS? 😉 Now I really am getting old, nowadays trying to keep up with the cloud native kids!
@@matthewfranklin7541 We had Apollo Domains at uni (Robert Gordon's Aberdeen) in 1986. Used for SPICE, chip and PCB layout. Felt space age back then. Lovely kit. Awesome keyboards. Superdomes, wow. Drooled over them. HP still make good kit, my PC is a Z800. Also a BMW guy (the old ones) and found that BMW designed the Z series case for HP!
@@unixnerd23 Oh wow, yes that's how I used HP Apollo's at Uni! FPGA/ASIC design using HDL - amazing to think that that technology existed even back then. 😊
@@matthewfranklin7541 Yeah, ASIC design. Small world :-)
At my office, we used to use tape backups every night on those small DAT tapes, and we had the larger ones that were 36GB each, we used to have to do a backup every night on a rack of them. There are a bunch of old ones laying around here still in an upstairs closet.
11:30. I think the use of these ports is pretty obvious considering the label right beside saying "Workstation Multimedia". The "microphone" and "headphones" logos are pretty self-explanatory too.
If this was used as a workstation (and not as a server) it must have been audio capabilities for multimedia content work.
When I was an undergraduate at MTSU, we had an HP-9000 v2200 server which served the whole campus. Everyone's email and a good chunk of homework was done on that system, which was named "frank". I was a computer science student, and so I was pretty well acquainted with the system. It was pretty cool to log in, run "who", and see like 1500 active users scroll by. I also played with a little assembly code for it. Not a bad instruction set, really.
This one is like a scaled down version of the v series, but it still brings back some good memories. Thanks for sharing!
You say finding a UW SCSI2 terminator would be hard, but they're literally $16 on Amazon...
True, but I had a drawer full of scsi cables and terminators back in the day. I can't say where I would find one now except online. I do not know a single person or server (or old disused desk full of old crap) that has one in my current environment. That is most likely the context he was speaking from.
When I made the switch from OpenVMS to Unix back in 2000, the company I joined had a pool of these running HP-UX with us devs spread across 3 of them, 3 for test and a couple of others as Oracle database servers. I can't remember how much memory they each had but they all had 4 cpus. I remember them being super reliable and we used them for many years, then a few Itanium boxes started arriving and not long after that we switched to developing on Linux, maintaining just a couple of HP, Solaris and AIX boxes as build and test servers. It was good to see one again so thanks very much for this !!
We had one running Lawson Financial software, it was pretty much a turnkey system, I think we got ours around 1994 so it would have been a 32 bit model.
You spent a lot of time talking about a "single power supply" but the power supplies for those older systems were seriously more reliable and even when redundancy is called for it may well have been built into the single power supply.
Do not make the mistake of judging older systems by how things work today.
remember, the redundancy is not only for the power supply itself, but for the reliability of a single power source.
@@davidjames666 And I noted that the required redundancy may have been built into the power supply itself. The designs of yesterday are not the same as the designs of today and hot-swap was the exclusive realm of the Tandem non-stop computer (our Data Center had one of them too). That everyone now does hot-swap is probably because the patents by Tandem expired.
That said, it is still a supposition on my part and not based on knowing the design. What I do know is that we never had any service calls made to fix or replace anything on that computer during the years we needed it. Like most HP equipment of the era it was extremely reliable. It was however just used for single application (Lawson Financial).
@@davidjames666 Our Data Center had a Uninterruptable Power System backed by a generator large enough to power the whole building not just the computers and that meant that I had Novell server at one point that had not been powered down for more than three years (updates didn't even require a reboot, MS should have payed attention) and even caught a secretary who had not logged out or powered down her IBM AT PC in over six months so she didn't have to change her password on a regular basis.
The idea that redundant power would be required to keep a computer running because someone would flip it's breaker switch by accident was not on the radar of Data Centers at that time.
A lot of the lessons we take as standard today had not been learned yet back then and yes we made those mistakes and corrected them in future projects.
Cool stuff, very nostalgic, I think. I worked for big fruity mobile phone company in the UK in the early 2000s and they pretty much ran their entire customer service system on rows of racks of K class servers in two data centres. I even have some photos of the racks with the servers in. They had big scsi disk expansions underneath that were almost half the height of these already huge servers. Getting them into the racks was fun!
Definitely the time when you go find an equipment lift because you'd need four people to lift one of these window AC unit impersonators. This was back when literally every server was twice as heavy as it looked like it should be. Lots of copper in those beasties.
RISC isn't an architecture. It's a concept/design principle.... It simply means a reduced set of instructions, as opposed to CISC like x86.
Wrong. RISC is indeed an architecture, just as CISC is. cs.stanford.edu/people/eroberts/courses/soco/projects/risc/whatis/index.html#:~:text=RISC%2C%20or%20Reduced%20Instruction%20Set,in%20other%20types%20of%20architectures.
@@orbitalgolem91 From your own link: "RISC, or Reduced Instruction Set Computer. is a TYPE of microprocessor architecture"
Key word emphasized in caps above. It's a TYPE of architecture, not an architecture in itself.
Synonym to "type": Category, Classification, Family, Label. You could rephrase that sentence above to: "RISC is a category of microprocessor architecture" if it helps make more sense to you.
An architecture would be things like x86, ARM, Risc-V, PA-RISC, MIPS, etc... They all are encompassed by a classification/type like RISC or CISC.
Don't shoot the messenger. I didn't invent the ISA acronym (Instruction Set Architecture", where the term "Architecture" comes from in the context of computer science.
en.wikipedia.org/wiki/Instruction_set_architecture#Classification_of_ISAs
RISC and CISC are mentioned there as classifications of a computer architecture. RISC and CISC are meant to classify an architecture's complexity.
Von Neumann, Harvard, and ISA are examples of computer 'architecture'. RISC is a 'type' or 'implementation' of the Von Neumann architecture.
4 years ago, I was a Junior Linux Admin at a fortune 50 company. We, at the time, still had a relatively significant HP-UX compute footprint running business critical apps. I never dabbled with the platform, but it was fascinating as a legacy OS.
As a side note, my current company retired it's final Solaris server recently ;) we are largely an Oracle Linux shop with over 13,000 VMs.
I used to learn to code Java on Solaris in 2015 in a German university! When I finished university, they adopted Suse linux for their unix labs.
Also, don't forget DEC with their various Alpha-based unix machines, or SGI with their MIPS stuff.
and VAX/VMS
@@tekvax01 VAX/VMS lasted well into the early 2000's running in steel mills, power plants, and other places where operation had to be rock-solid.
I worked at a large print shop around that era that came with one of those. Essentially it came paired along a large Heidelberg offset printer. The thing did spooling and other things. It also came with its terminal display that ran a postscript server that allowed us to push, pause or cancel print jobs on that huge machine.
I love old technology
It was very rare to connect morr than one modem. These were often used to run manufacturing floors, and science research. The serial and parallel ports were typically connected directly to machines and/or sensors, and the server would sit and process that data.
7:15 if only I can say I still do that now, 😩 I always thinking go only digital is bad and this is one of best example of it, sure it's will take time but it's still got the job done than what we do now, well I still use half it manual so it's better now.
Clicked on this because it was very reminiscent of the type of servers the military was running back when in joined in 1998. Not sure we had this exact model, but not too dissimilar. As always, great video and good job breaking it down (without actually being able to break it down).
I love how HP (and dell) make amazing enterprise grade hardware yet some of their consumer hardware is such junk. This thing is awesome.
Consumer grade hardware at consumer prices vs enterprise hardware at enterprise prices. Consumers want cheap. Enterprises want 24/7/365 uptime.
Many servers of the day were used as high-end workstations. That's what the graphical head was for, not "managing the server". That's also what those blanked out audio ports were for.
Most DDS2 tape drives would play out standard DAT audio tapes too!
Cool! shutdown -h now still does the job in Linux Mint.
and btw a top spec Mac Pro (28 Cores, 8TB SSD, 1TB RAM, 32GB Video) is USD 53,799(!!) today.
I think shutdown -h now has worked for quite a long time dude... I remember using it quite a bit
@@adventureoflinkmk2 You are correct. I was probably on win95 back then. We have all been noobs 😘
Proper rackmount kits were available for these, I’ve converted bunch of them back in the day. Internal modem was mostly used for software called Predictive, sort of early call-home solution. These servers were really popular - most of the banks, telekoms etc. had ton of them.
That explains why its 64 bit, as 4 gb is at that ram limit on 32bit
That's only on widows 😂 Even Linux can run more than 4gb of ram on 32bit cpus (with PAE).
@@lenoohpuls It can't directly address over 4GB of ram.
Yes, you're right. Windows also has support for PAE, but it's not usually enabled. And yes, it needs to be supported by the hardware as well.
I supported 8 of the K's, email, database, file storage, backup.....I remember having to upgrade memory for the email server and it was 100$ a month added to the lease.... All were in HP racks with the disk arrays below them. I recall one of the leases the total was over 100k. $$$.
And....the 64 bit s/w ran slower than the 32 bit. but could access more data. So it was still quite early for 64 bit systems.
That PSU was probably a lot more reliable too. Built for server use. Redundancy is especially needed when your PSUs are prone to failure.
This is correct. There was triple redundancy inside that power supply. And I'm later models you could order them with A/B power connections, so that all three sets of international (to the supply chassis) power circuits could switch between A/B.
That vent design around the case made me think it was something from mid-to-late-90s IBM at first.
Edit: half way through video; I noticed the tower PC behind it and now I see just how big this thing is. WOW!
I am VERY interested on what servers from back in the day were actually running and how they were used on a day to day basis.
Does anyone have any links to other videos that go more into detail on the software/OS side of things?
Thanks for this little visit to the past. i need to admit it makes me feel really old since we replaced many other servers back in the days with the then new K-Classes.
Wow, my brother, that computer he’s my brother
Wow! I had to deploy these servers K9000 RISC (10 of them) back in 2006 for a legacy platform because the customer couldn't upgrade to a newer platform.. Wow...
Hp still doesnt have dual psus i have a pretty new hp server and it still only has one psu
Wow, dell has 2psu. I seen an HP server to have 2psu option.
Depends on which server you get. The HP DL series will have dual PSU. The DL580s can take 3-4 PSU depending on config
The way Colin gestures at everything with just his hand on camera while talking, never showing his face or anything else, makes me want to imagine that ALL there is to Colin is a talking hand
hmm a talking hand!
I’m almost certain that the FDDI board was used to access a Storage device. If such were the case, there is likely a twin of this machine that would have been in warm / hot standby mode if the app was mission critical. It would be interesting to get a recursive listing of the file system to determine what this machine would have been used for. If a /usr/local and/or /opt directory exists, that would be a great place to start spelunking. The timestamps of the files would also reveal how this machine was used.
For what purpose? None whatsoever. Just because it would be fun.
This is the type of machine that would be used in a retail store setting for inventory DB and to run POS systems and probably dumb clients via direct serial connection to do lookups etc on menu based application to that DB.
Running Piece Of Shit systems? 😂😂😂
@@adventureoflinkmk2 😅 "Point of Sale system"
@@adventureoflinkmk2 Point of Sale, but almost always also known as "Piece Of Shit" system. My own personal experience with POS systems is that the corporation always hires the cheapest bidder to develop the system and it's always garbage that looks and operates like a Geocities website designed by an 8 year old from 1997.
I appreciate the video but many things mentioned were wrong. It's still common to run servers headless today. The IBM hardware has a HMC (hardware management console) you use to partition the server, start/stop and other maintenance functions. For servers of this vintage, you'd either have a dumb terminal attached or more likely you'd have a terminal server attached so you could still connect to the console from your desk. You were correct in calling it a pedestal server but servers of that type were typically more overgrown workstations than servers. PA-RISC servers had their own flavor of media instructions that was a thing before the first Pentium MMX ever came out. Depending on the use case, HP/IBM/Sun systems of that era were also clustered as hardware was fairly reliable but nothing like today. We'd periodically get server reboots for no reason. The main memory had ECC but the cache on the CPU didn't have ECC, so it was a bit of a known issue. Since memory had low density, you could interleave banks of memory to get better performance too. I preferred Solaris but the HP-UX boxes were also OK.
That's an awesome machine. Thank you for the video. Did the CPUs share the whole 4 GB of RAM or was it split up?
I have a bunch of K-Class Core IO boards cluttering up my garage. I used to work on these machines and the big K580s were my favourites as the CPU cooler designs were amazing. And The way they sounded when powering on was like a jet engine!
I would only be impressed if the model was OVER 9000
WOW This brings back memory. We were using the called "Little Sister " HP 3000 until 2018. Long time no see this logon screen. Thumbs Up and more of these
13:58 nice voice crack lol
My first Unix experience was on HP-UX. Our college workstation & server hardware was a few years older than this 64-bit server, but it was a wonderful time to be learning & working in the IT world.
You needed all those modems for all the customer dial-ups at your small city ISP. This was the early public/commercial Internet too. AOL let all there users on the wider Internet in 1994 and things changed immediately for those of us already hanging out there.
Are you wearing a mask? Because there's something kind of odd about your voice in this video.
Likely bc he mentioned he's in free geek twin cities
Yep.
Ahhh, I thought he was eating a sandwich through the video...
That was considered a Modern server back in the day. I supported plenty of them. My history point of reference for servers (mini computers) was the Sun 3 from the mid 80's.
Back in the day when HP was a serious player in the enterprise market and had its own 1st rate version of Unix. Now HPE is just an x86 blade rebadger and HP itself just flogs cheap made in china consumer junk. All thanks to a bunch of CEOs that only cared about short term profits.
I wouldn’t say they are a blade rebadger - far from it. HPE still designs and makes custom hardware, whether Intel or otherwise. And don’t forget they bought both Silicon Graphics and Cray a few years back too.
@@TheMrMarkW You'd currently struggle to find any influence on the HPE lineup from either of those companies.
@@boltar2003 what, like HPE Exascale with Slingshot? Like they’re using in frontier?