Bye Bye Solaris, It Seems.

For readers of A Certain Age, this may bring a tear to the eye. Reports have been circulating of the decision by Oracle to lay off a significant portion of the staff behind its Solaris operating system and SPARC processors, and that move spells the inevitable impending demise of those products. They bore the signature of Sun Microsystems, the late lamented workstation and software company swallowed up by the database giant in 2009.

So why might we here at Hackaday be reaching for our hankies over a proprietary UNIX flavour and a high-end microprocessor, neither of which are likely to be found on many of the benches of our readers in 2017? To answer that it’s more appropriate to journey back to the late 1980s or early 1990s, when the most powerful and expensive home computers money could buy were still connected to a domestic TV set as a monitor.

If you received a technical education at a university level during that period the chances are that you would have fairly soon found yourself sitting in a lab full of workstations, desktop computers unbelievably powerful by the standards of the day. With very high resolution graphics, X-windows GUIs over UNIX, and mice that weren’t just used for a novelty paint package, these machines bore some resemblance to what we take for granted today, but at a time when an expensive PC still came with DOS. There were several major players in the workstation market, but Sun were the ones that seemed to have the university market cracked.

You never forget your first love, and therefore there will be a lot of people who will never quite shake that association with a Sun workstation being a very fast desktop computer indeed. Their mantra at the time was “The network is the computer”, and it is the memory of a significant part of a year’s EE students trolling each other by playing sound samples remotely on each other’s SPARCStations on that network that is replaying in the mind of your scribe as this is being written.

A Raspberry Pi with a Raspbian desktop probably outperforms one of those 1980s SPARCStations in every possible way, but that is hardly the point and serves only to demonstrate technological progress. It feels as though something important died today, even if it may be a little difficult to remember what it was when sat in front of a multi-core x86 powerhouse with a fully open-source 64-bit POSIX-compliant operating system running upon it.

Unsurprisingly we’ve featured no hardware hacks with such high-end computing. If you’d like to investigate some Sun Microsystems hardware though, take a look at the Centre for Computing History’s collection.

96 thoughts on “Bye Bye Solaris, It Seems.

    1. Yeah, the network bandwidth on a Pi is absolutely atrocious. The Pi2 could only do about 60-70Mbit in real-world use (with iperf just pushing packets, you could almost get 100Mbit).

      1. Yeah should compare it to a pizzabox or a netra or something.
        But in about 20 more years, when the netra and pizza box will still be humming away quietly without requiring reinstalls and new sd cards on a periodic basis…
        Still got two netras, a disk array and a sparcstation kicking round in the rack, they just keep on working, unlike the pi’s which are mostly consigned to a drawer in disgust. I spin them up when I need to test something on solaris on native sparc arch…
        Now fingers crossed I can snag some bargain bucket nicely spec’d v480’s to carry on the good work.

    2. My 1st Sun was a sparcstation 1+ with SunOS. At home, I was running Linux (0.95) on a 486 w/ a VLbus ATI graphics card. The PC graphics were faster, but the sparc could do a compile in the background and still be responsive.

      A PI has more RAM (10x!), faster network, more cores and probably faster graphics.

      1. My interview with Sun in the early 2000s consisted of “Name THE 5 ways to drop a machine to an OK prompt”. They were quite upset when I came up with 6 ways (pulling the keyboard connector is probably not a good way to do it, but it still will get the job done!).

          1. Accidentally hitting the watchdog reset button that was conveniently located next to the AUI connector — which you were groping for under the desk because the mounting clip had too many washers on it as shipped from the factory, so the cable would keep falling out — for the 1000th time.

            (Yeah, technically, that would get you to the “>” prompt, not the “OK” prompt, since those particular machines were pre-OpenBoot.)

  1. As a programmer I always thought of Solsris as something between a duck-billed platypus and Frankenstein’s monster; Its heritage was in BSD but they’d gone and mutated it back to have a _mostly_ System V userland, with a few BSDish features and a few uniquely Solaris quirks (iirc either select() or poll() had some weird corner case behavior that didn’t resemble either parent) and some userland programs had wacky non-standard command line flags. At one job I had to maintain and extend an in-house VoIP switchboard daemon (this was in ’98 so bitrates we’re low and the fact that it worked at still blew non-techie’s minds) that needed to compile and run on FreeBSD, Linux, Solaris, and IRIX. I, for one, will not miss Solaris one little bit.

  2. RIP. I have fond memories of my 2/120, 3/50(milo!) and later sparc pizza box machines. They were a joy to work on compared to the DOS machines of the time, and were really well engineered.

  3. Solaris was also responsible for running a lot of early e-commerce sites and web services. If you use iCloud services then your emails probably ran on Oracle Communications Messaging Server running on Oracle Solaris Cluster software. The new macOS due this autumn has a new file system that borrows heavily from ZFS – another Solaris innovation (yes I know there was a legal bun fight over ZFS)

      1. I don’t miss the low-rez mouse usually attached to a Type 4 keyboard (of which you speak). The Type 5 (but not Type 5c <– "cheap") was the best of any keyboard Sun ever produced.

        We had endless fun in the computing labs rotating the gridded glass mousepad by 90 degrees and watching people try to figure out why they didn't work anymore. :^)

        1. When I left my job working for the phone company in 1999, I put a yellow sticky under everyone’s mouse on the Spark 3s… I remember starting 4 years earlier being amazed that all the machines came with 64 Megabytes of ram! I did once manage to open 3 windows, one running Solaris, one running Windows, and one running the Mac operating system, all at once. Took a long time to shut those windows it was running so slow.

  4. To qote one of the old commercial battles between I think SUN and Next:
    “The sun ain’t gonna shine anymore” at least it outshone NeXt for quite a while.
    And yes I fonldy remember Mosaic as my first Web Browser on an X Window served by a Solaris machine!

      1. Sorry MacOS is now more like Windows Vista than Un*x…Apple is more interested in turning “computers” into crippleware iOS devices where a user’s only function is to interact with the app store and upload personal contact lists and browsing habits for third party companies to monetize their info

  5. Sun had so many things before anyone else really did it well. Yellow Pages, well, too bad that was a trademarked name, but you know one login anywhere on the network. NFS that made any disk anywhere a reality, not goofy MS-DOS disks, but real partitions or whole disks. Sun Windows was kind of a mess, but X windows eventually worked on the work stations, then a foray into Visual Postscript (NeWS), when the hardware wasn’t quite ready.

    Multiple attempts to make Unix affordable, i386, SLC, IPC, etc.

    Mainstream of RISC processors.

  6. Back in the mid-90’s the company I worked for was still very much a Mac shop with a growing influence of Windows 3.1. I had an Ultra10 on my desk with a pair of 19″ monitors instead and ran Sun’s WABI and Apple’s Mac emulator on Solaris. Demoing that our IT management was why we didn’t try to go with NeXTStep/OpenStep on the desktop, despite our VP’s insistence that it was the right thing to do.

  7. I have a fondness for Sun Microsystems. However, IMO their tragic flaw is they were a Software company that thought they were a Hardware company. They could have owned the Internet space if they stopped dissing Intel hardware as “toys” and Linux as a “hobbyist” OS. In short, they should have gotten behind Solaris x86. As a result, the market moved to Linux on Intel and left Sun and their “superior” hardware to the dwindling, late adopter, market.

    1. Their hardware really *was* superior, and they could make the case that a single Enterprise could do the work of a big rack full of PCs with higher per system uptime. This was a valid sales pitch even after x86 Linux took over the expendable web server frontends. But eventually PCs got powerful enough that VMs and containers could work around the reliability problems of cheap hardware, and were easier to scale without hardware or licensing costs.

      Sun did sell some more modern x86 systems, but by then HP and Dell already owned the x86 datacenter markets.

      1. No doubt, Sun made better server hardware “back in the day”. However, it was 10x the price of Intel gear. I’d rather have my app distributed across 5 systems, for 1/2 the price rather than put all my eggs in one, superior basket.

        Re Suns x86 system
        Precisely my point. They thought they were a hardware company. Why would I buy Sun’s Intel hardware when I could get much less expensive, better gear elsewhere? Had they fostered Solaris x86 instead of threating to kill it on every release, the Internet may have been running on Solaris today instead of Linux.

        1. “This Sun basket holds 100 eggs and it’s made of the highest grade wicker.”

          “Can the handle still break?”

          “Uh, well, yes, it might. But it will absolutely last much longer than the handles on those cheaper PC baskets made of lower grade wicker. And did I mention it holds 100 eggs VS the 25 those PC baskets hold?”

          “Yes, you have, many times. But if a Sun basket’s handle breaks, I lose all 100 of my eggs. If a PC basket handle breaks I only lose 25 eggs, and for the cost difference between four PC baskets and one Sun basket, I can afford that amount of egg loss, if it ever happens. I can even afford a fifth PC basket and hold 125 eggs. So why should I spend more money to put all my eggs in one basket?”

      2. Yes, but then we have seen this model before. Technical superiority does not mean you will come out on top. Think (VHS and Microsoft) vs (Beta-max,Amiga,Unix, Linux). The general public desires a certain inferiority to hold back technology to it’s proper place. We don’t want technology to advance too far too quickly. Even the big names like Musk are afraid of advancing to SkyNet capabilities.

  8. Nope. It isn’t going anywhere anytime soon, if AIX3.2 is any indicator. The FAA has a specific contract with IBM to maintain a particular version of AIX for the Display Systems Replacement project. TFM uses Solaris, and I’m sure many other government projects.

  9. Still have a few Solaris servers around my workplace — slow to move a couple of legacy enterprise applications to Linux, I’m afraid. Will probably have them running through the end of the year…. Most of them have multiple years uptime.

    1. you know you have a well maintained machine (or ungodly luck if it isnt) if you start measuring uptime in ..not days,or months…..but YEARS. reminds me of a few servers i have around my place with linux that are breaching the 2 year uptime mark without any hiccups

      1. It is the ungodly luck one… We’re actually too scared to reboot in fear that we might crack a motherboard or blow a cap on something irreplaceable if we do have to reboot. With the one V440 we did reboot (with 9 years uptime) I sweated through several layers of clothes that day.

      2. The years that it was cool to have a lot of uptime are gone. Nowadays, a system administrator shows his competence by having a recently patched system, including any kernel patches. That means having to reboot occasionally.

        I’ve heard an interesting talk on how this uptime attitude has actually hurt a few companies badly.

  10. Oracle priced Solaris and SPARC out of the desktop market. If they were cheaper I’d get a used high end SPARC workstation and run the last free release of Solaris from Sun. But S10_u8 on x86 hardware is more cost effective even if it is the Little Endian spawn of the devil.

    Why you ask? Because the Sun/Forte development tool set is the best ever implemented. And I’ve used the *nix dev tool chains from IBM, Intergraph, HP, SGI, DEC, Evans & Sutherland, Alliant, Intel, Gnu and probably some I’ve forgotten. The Sun dbx is the only debugger which will evaluate intrinsic function calls on the debugger commandline. That is *really* valuable dealing with large, dusty deck scientific codes written in obscure subdialects of FORTRAN by scientists who never heard of “software engineering”.

    Fortunately, there is the Illumos code base which Sun made open source. It is alive and well, though somewhat of a niche. Sadly it seems unlikely that Ellison would release the later code, but there is a small chance it might happen. The system I’m using for Internet access is running the Illumos based OpenIndiana on x86. I keep my S10_u8 system off the Internet on an isolated LAN to deal with the lack of security patches. It runs like a champ. After HP disbanded the Alpha group Intel and AMD finally learned how to do really fast floating point from the refugees. In 1999 a 533 MHz Alpha 1164LX was 3-4x the FLOPS of a similarly clocked Intel or AMD processor.

    1. About 10-15 years ago, my company was selling software on multiple Unix/Linux platforms. Whenever we had a bug pop up on one, the first step for fixing it was to reproduce it on Solaris so you could debug it with dbx.

  11. OpenIndiana is a continuation of OpenSolaris, albeit one that is not updated very often: https://en.wikipedia.org/wiki/OpenIndiana
    Apparently there’s been enough changes on each side that it’s not really compatible with the closed source form of Solaris anymore. I wonder if Oracle will decide to release the source code to the final version since they are ending their development on it.

    1. Doubt it, Oracle gutted Sun, the only reason they bought it was to bury MySQL that was destined to eat into it’s marketshare of the enterprise RDBMS market share. unlucky for the rest of us, they also took posession of solaris, java, openoffice, sparc, virtualbox and countless other good things Sun helped bring into the world.

  12. Cue the “Pi mounted in a SPARCstation 5 box” project!

    I still remember being floored walking into a lab in college and seeing SPARC, HP, DEC, IBM, and SGI workstations humming away with their giant 20″ CRT’s.

      1. None, I’ll bet. IIRC, the U-class motherboard was ATX form factor. Or at least very similar. I used to work with a reseller that put U10 and U60 motherboards in heavy-duty ATX rackmount custom cases all the time — I assume it could work in the other direction as well.

  13. Meh.. Oracle is where good things go to die. I’m sure they would even kill off MySQL if it weren’t for the fact that they know all the forks would just carry on happily without them. with the exception of that I just consider anything Oracle manages to purchase as already dead. This is just an announcement that the corpse has no heartbeat.

    I’m not sure what exactly Oracle exists for these days. Do they actually have a commercial project with a significant user base? I’m guessing they are either on borrowed time themselves, living off money accrued in decades past or they are kept alive by government contracts, mostly for keeping large databases of innocent civilian’s personal lives.

      1. I don’t know anything about Hudson.

        MySQL is kind of a weird one. Sure.. the open source developers mostly moved on to Maria but Oracle keeps pushing out new MySQL versions. Meanwhile there are at least 3 or 4 other forks as well. We use Percona where I work.

        Anyway, because they all strive to remain compatible I think they still kind of follow what Oracle does. So.. Oracle still keeps a big seat at the MySQL (and derivatives) community.

  14. I did some work at quite a few telco switch sites back in the mid 90’s. Lots of Sun hardware, Solaris and HP OpenView in those places back then. I spent a lot of time on them in the late 80’s – early 90’s. A little at McDonalds Corp, and a lot at Novell in San Jose. Good times indeed.

  15. Geezing, I remember the 40Mbyte HDs that if you turned them off you had to smack them like an old tv set to unstick the heads. It’s been a long time but weren’t those in the Sun pizza boxes?

    1. I remember there was one flavor of 120MB or so SCSI from Quantum that popped up in some of the SUN4C workstations as well as mid ’90s Macs. If they were powered off for more than a few weeks at a time, you would have to tap them on the front after powering up since the grease in the spindle would congeal.

  16. There was one well-known hardware hack on the Sun stations. The mouse pads were metal with grids printed on them. You could hold them up to the giant CRTs for a few seconds and then quickly pull the pad away and draw a GIANT spark! (I don’t think that’s why they were called Sparcstations…)

  17. I’m now using and typing on a Type5 keyboard on this multicore i7 PC. Still “The keyboard” for a real unix geek with some (fair) rivalry with the IBM model M. Hacked and adapted from Sun bus protocol to USB with a classic atmel hack. It’s getting old but it’s still nice, Sun deserved more from Oracle.

  18. My first ethernet installation here in Sweden in 1982 as to connect our VAX 750 running BSD to the SUN 1 running Sunos for our data research department.
    The CPU was a 68010 with a custom MMU on a VME card in a black box, a Digital vt100 keyboard and a square 19 inch monitor.
    Anyone have a burning need for a Sunos 1.1 tape cartridge :-).

  19. My memory is appearently not correct, multibus not come and 68000 appearently, my last sun I used was a ultra 10.
    I was a bit proud when I had sun2/120 for my use, it was in the same price range as a 1 family home.

  20. My first experience with them was using packing tape to keep the thick ethernet connectors from breaking off the back, though that was wos’s name’s fault. Also that Computervision didn’t port CADDS to them actually, just wrote a CV-FORTRAN to C convertor and patched a really poor GUI on top. What really cheesed me was they were licensed boxes via Computervision and Computervision didn’t include any development tools so engineers were unable to create C programs for themselves.

    As I recall Sun was created to undercut Apollo. They did a lot of smart things, but Sun didn’t come down in price as fast as Microsoft came up in capability, though the same could be said for all the Unix variations. It’s very hard to grow from the top down.

  21. Ah those were the days.. I managed a whole park at my employer with a big 4-280 as central server for a bunch of discless clients., Mostly 3/50 and 3/60 systems. later on sparcstations. What a panic if some of the more creative types were mangling the ethernet thin coax cables causing excessive drops :-). a few years agoo I had to ressurect one of them by repairing the QIC-120 tape unit with a tap rubber tape pressure roll replacement, All gone now.

  22. I worked for DEC back in Sun’s time. There were some Unix workstation haters in high places at DEC and we just watched Bill Joy and Sun walk right by. Eventually MIPS-based and Alpha-based DECstations did very well but others had already earned the central market share. IIRC, early Sun workstations had a pair of 68000s or 68010s or whatever that didn’t have instruction retry to effect virtual memory. One processor followed the other until a page fault at which time the follower was stopped, pages were swapped, TLB updated, and the previously follower CPU became the leader and the previous leader the new follower (of course I could be ascribing this technique to the wrong company). Pretty cool for a HW EE to hear about back then. I used SparcStations for HW design at a later company and they behaved well and seemed to be very manageable. If you were running HW design tools back then, SparcStation/Solaris was often the most supported option for running a given tool on Unix. I agree, Sun, Sparc, and Solaris more than made their mark in workstation history. Hoist one to very capable group.

    1. The 68020 was the first 68K CPU to be able to do virtual memory, but it required an external Memory Management Unit chip to do it. The 68030 integrated the MMU – in some versions. Motorola had them available without MMU, same for the 68040. One could get an 040 minus both FPU and MMU, possibly also minus some other features if your system didn’t need them. IIRC they did similar things with the 60x series.

  23. I was far sadder at the demise of SGI Irix than Sun Solaris.

    Solaris was well marketed, and had some nice features, but Irix was an operating system which was truly ahead of its time (scaled to thousands of single-system image CC:NUMA processors). It was the OS which gave us OpenGL. It’s user interface was jaw dropping, and in 1990 did things which even modern OS’s don’t do as well today. Solaris matured well, but it wasn’t especially innovative (ok, maybe ZFS).

    And before you accuse me of bias, I ran SunOS, Solaris, Ultrix (DEC), Dynix/PTX (Sequent), UTS/M (Fujitsu) and UNICOS (Cray) systems for a major university (1990-1993). I still have the memories of turning off two of the four processors on a Sun Galaxy 4/690 MP to massively improve performance, because their crap SunOS kernel wasn’t multithreaded, and Solaris at the time was unusably unstable. I worked for Silicon Graphics (1994-1999), then Sun Microsystems (1999-2004). Upon arriving at Sun and sitting down at Solaris, I couldn’t believe what a piece of crap it still was.

    Sorry, I won’t miss the SPARC ISA, and I won’t miss Solaris.

  24. I still think Silicon Graphics takes the cake. What a gorgeous desktop machine. I got to unbox several Indy desktops for some medical product engineers and taken aback by their aesthetic engineering and beautiful graphics.

  25. Reading this article just made me remember that our laser plotter used to print pcb films is going to malfunction some time in the near future. It was such a pain to get another Sun SPARC workstation working in the past, sourcing LVDS scsi drivers, installing HP-UX etc. I’ve pretty much made good notes on what I did 3 years ago, but the issue is not in our end this time. There is a increasingly lower chance of a person being in the other end of the software vendor that actually remembers how to generate new licenses from the hardware ids and/or whether they still have a working HP-UX machine to do the licenses on.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.