The Sun Also Sets
by
Jim Frost
July 3, 1997
Sun is in trouble.
This one simple statement will inflame a lot of people. They’ll point to Sun’s growth pattern and show that it’s healthy. They’ll point to Sun’s Java and Network Computer initiatives and show that they’re trying to expand their market. These things are true; for the moment, Sun is doing well. But the entire underpinnings of their business have been being ripped out from beneath them for the last several years, and they’re left standing only by the most rickety of supports right now. Sun will be a pale shadow of today by the turn of the century.
I got the first wind of this back in 1994 when I was at the Solaris Developer’s conference. Bill Joy was giving a great speech – he had the crowd laughing along with him, and he was having a great time. The thing that struck me, though, was that he kept ribbing Microsoft. He had to have dropped a dozen jabs at the company, particularly at the company’s new Windows NT product, in the span of fifteen or twenty minutes.
At the time I’d just started using Windows NT, and I had been astounded by how good it was. Not great, mind you, but it had lots of UNIX-like features, they actually worked, and for the most part no matter what I did it just kept chugging along. After having fought with MS-DOS, OS/2, and Windows over the years this came as something of a shock to me. The contrast between this experience and Joy’s glib comments was startling. The very first thought in my head was, “He doesn’t know what he’s up against.” I thought there was just a little bit of nervousness in those jokes.
Looking back, I think I was wrong. NT is really only a side issue; Sun would be in trouble even if it didn’t exist. It’s the PC that Sun was worried about. The PC is going to destroy Sun, and it’s going to do it very quickly.
Historical Perspective
In order to understand what’s going on, it’s good to use historical perspective. This whole situation harkens back to the days of VMS. I remember the day when Ken Olsen joked about UNIX, dismissing it as it “snake oil.” At the time DEC looked untouchable; they could ride this UNIX thing out, no problem. DEC was competing effectively with Sun and Apollo in the workstation arena (not dominating, but competing), and nothing these competitors made could touch DEC’s big iron. If you needed something better than what they offered, who were you going to go to?
Just three years later DEC sales had dropped off precipitously and Olsen found himself out on the street. Despite tremendous advances in technology the company has been totally ineffective in the market since then. What happened? Many industry observers say that UNIX won on account of its “openness.” They say that UNIX had a natural advantage because it was available from more than one vendor.
Those explanations are wrong. UNIX wasn’t really open; its much vaunted inter-vendor compatibility came naturally as a result of its common original source base. Everywhere that vendors extended it they did so in incompatible ways, leading to myriad incompatibilities that caused a big headache when it came time to move.
What really did it, and if you look at the huge growth spike of UNIX in the late 1980’s this is pretty obvious, was a hardware shift. About a decade before we’d seen a huge jump from mainframe solutions to minicomputer solutions; the price/performance was much better than mainframes, so users got more computer for less money by buying minicomputers. Minicomputer companies like DEC and DG and Prime exploded. This transition took awhile; while minicomputer systems were available by the late 1960’s, they had little in the way of useful software. It wasn’t until the second generation of minicomputer systems that the software and the hardware were in synch, and from that point on the mainframe industry took a pounding.
UNIX grew up in that same time frame, and was unique in two ways: you could buy the complete source and distribution rights, and it wasn’t very difficult to make it run on a new hardware base. UNIX rapidly expanded to support much of the hardware that was out there, from the new “personal computers” to large mainframes. But it didn’t have much effect on the market; when in direct competition with proprietary systems such as VMS, UNIX didn’t do well at all.
When RISC hit the scene starting in about 1987, UNIX was tailor-made for it. The source was there, it was a lot less expensive than writing it yourself, you could get it up and running in a matter of a few months, and it was a no-brainer to move applications software. There really wasn’t any other choice that made any sense whatsoever. By 1988 RISC systems, all running UNIX, were blowing the doors off of minicomputer and CISC-based workstation systems. When Sun released the SparcStation they improved baseline price/performance by 300% in one fell swoop. The market jumped – just as they did to mainframes a decade earlier, and for the same reason: you got more computer for your money.
We are currently in the middle of another such hardware transition, and the similarities between what happened in the minicomputer to workstation transition, and how the respective companies reacted, are remarkable.
The Personal Computer Onslaught
Most of you reading this have probably noticed that PCs have become increasingly common in engineering environments which, only a few years ago, were entirely equipped with workstation systems. This change occurred because of several simultaneous improvements in the PC hardware base. First off, processor performance improved by tens of times in just a couple of years; this gave PCs the raw computational ability to run many applications which would have been too slow otherwise. Secondly, display density increased from a baseline of 640x480 – about a quarter of the density of workstations – to densities equivalent to, and in some cases superior to, most workstation displays. Thirdly, the standard PC bus went from the rickety old bus introduced with the IBM AT in 1985 to the much more modern PCI bus – essentially eliminating the I/O bottleneck that had so plagued PCs.
Despite these advances PCs could not compete effectively with workstations, primarily because of their operating system software. The Microsoft Windows 3.1 platform that was standard at the time was, as we all know, horrible to write software for. Segmentation, lack of address space protection, and horrific I/O support made it all but impossible to build high-end applications. UNIX solutions, while long available for PCs and far more capable than Windows, never had the right combination of performance, low price, and software necessary to sell in respectable numbers.
This began to change in 1993 with the release of Windows NT. NT broke the resource limitation barriers that crippled Windows 3.1; it was perfectly capable of running all the same software that ran on UNIX workstations, it was dirt cheap compared to popular UNIX variants, and it promised the same kind of mass applications market as Windows 3.1. Unfortunately much of the software that made Windows popular wouldn’t run on a more restrictive environment like NT, and the operating system required hardware that was relatively high-end even for workstations at the time. It languished in the market; there really wasn’t a good reason for anyone to switch. What it needed was a drop in hardware prices and lots of native software. Plummeting storage and memory prices supplied the former, and the latter was the job of Windows 95 – an operating system written for no other reason than to act as a bridge between old and new Microsoft technologies.
Today we stand on the brink of another mass hardware jump. PC performance is equivalent to that of far more expensive workstations. A host of different operating systems capable of utilizing PC hardware to its fullest potential are available at very low prices. And the last piece of the puzzle, the software that makes it all worthwhile, has been migrating in large numbers to the PC platform – both to Microsoft’s NT and to Linux, the most popular UNIX of all time.
What does this have to do with Sun? Everything. If nothing else, the “Solar System” management fiasco at Sun proved that their hardware and software revenues are inseparable. Take away the SPARC system revenue, and Sun can’t make money on their software. Take away the software, and Sun can’t sell their hardware.
Sun absolutely relies on sales of hardware systems, and those system sales are about to get demolished by PCs, just as the VAX hardware sales were demolished by the SparcStation.
No wonder Bill Joy was making nervous jokes.
Reacting to the inevitable
None of this is news to Sun. The first obvious signs of their reaction to the personal computer threat were visible in 1994 when they started to once again diversify their operating system support – porting Solaris first to the Intel, then to the PowerPC. If the SPARC market collapsed, Sun wanted to be sure that the software was already there to support a hardware jump to whatever was necessary.
Unfortunately Sun drastically underestimated the impact of two different market forces. The most obvious was NT; in the last two years NT has gone from minuscule sales to outselling all commercial UNIX systems combined, making huge inroads in engineering, business, and software engineering environments where UNIX has been dominant for almost a decade. In 1997 NT will come very close to selling more units than UNIX has sold in its entire history, and 1998 will see NT right in the middle of the mainstream business market – tens of millions of units. That has turned NT into an irresistible market to many traditionally UNIX-oriented companies.
NT, however, isn’t the most serious problem. Sun could probably compete effectively with NT purely on the grounds of the head-start they have in operating systems performance technology; they wouldn’t dominate any longer, but they could still compete. The most serious problem is the same as it was for DEC – the high end market, the market that Sun calls their own and everyone has assumed is untouchable, is about to fall prey to the same inexorable march of technology that killed the VAX.
For all its faults, Microsoft’s “Scalability Day” did prove one thing: “PC” hardware and operating systems are on the verge of competing with the best that Sun has to offer. The combination isn’t quite there yet, but it’s probably not more than a few years off on the outside. When that happens Sun’s lucrative hardware business is going to evaporate and the whole business will come crashing down in a financial free-fall.
Sun knows that this is going to happen too; PCs have been slowly climbing the capability ladder for years, and Intel’s creation of CPUs which are easily grouped in recent years has accelerated the trend. Unfortunately Sun doesn’t have the resources to compete against Intel in the hardware spectrum – they’ve been losing ground for years. What they can do is try to generate as much demand for their large systems, the ones that PCs cannot compete with yet, as possible.
This is what the Network Computer is all about. It’s not about making desktop systems cheaper to administer, although that’s a nice side effect. It’s about making the central computer indispensable. The funny thing is, this is exactly the same tact that DEC tried.
DEC used to make a lot of their money selling terminals to connect to their systems. Every terminal you sold was actually two sales: you sold the terminal, and you sold dependency on the computer it connected to. Unfortunately the PCs and workstations that were becoming ever more popular provided one thing that the terminal could not – interactive graphics. When DEC realized that their character-based terminals were doomed, they started looking around for alternatives. MIT had a distributed graphics system they were working on, and DEC (long an MIT supporter anyway) started funding it. This software was the X Window System, a software system designed to be able to run on even a minimal terminal – a terminal DEC would later call the VT1000, and the industry at large call X terminals.
The X terminal was supposed to provide a much more manageable, lower cost graphics desktop. If this story sounds familiar, it should: this is exactly the same story that is being given for the new Network Computers. Few would call X terminals a big success. It stands to reason, then, that we ought to give some thought to why the X terminal failed before we bet too much on Network Computers.
So why did the X terminal fail? The primary reason was the ubiquitous PC. A PC is not a dedicated function device; if you have one, you can adapt it to do most anything with new software and/or hardware. By the time the X terminals hit the market many character terminals had already been replaced by PCs which were acting both as terminals and as local processing systems. This left the X terminal at a threefold disadvantage: 1) the PC could be equipped to be an X terminal for a lot less money than replacing it; 2) if you replaced the PC with an X terminal you also had to replace the other software it ran; and 3) the volume of the PC market gave it economies of scale such that allowed PCs to sell for prices only slightly higher than X terminals.
These market forces have not changed in the last seven years. In fact, PCs are far more numerous now than ever, having replaced terminals in all but the most limited areas. With that growth has come even more diversity of function and ever falling prices. But the final bullet is the software: the very same software that has been promised for Network Computers already exists on the PC. Every PC which runs NetScape, or which could be upgraded to run NetScape, is already a Network Computer.
New terminals didn’t save DEC, and they’re not going to save Sun.
Survival Strategies
Not all of the minicomputer vendors were killed by UNIX. Hewlett-Packard in particular managed to make the transition from minicomputer systems to UNIX workstations, keeping step with market demand. Today HP is the #2 supplier of UNIX workstation hardware behind Sun, the only previous-generation computer company that is even close.
The reason why HP succeeded in this transition and others, DEC in particular, failed was that they were not afraid to support more than one technology, letting their customers choose which was right for them. DEC bet the farm on their VAX/VMS technologies, and they lost. HP went with the flow and flourished. HP customers got to get the new style hardware they wanted but keep their traditional vendor, a comfortable situation all around.
Today we can see the same contrast in mentality between HP and Sun. Sun is steadfastly refusing to work with PC technologies – they’re not building PC-style hardware, and they’re adamantly against supporting PC operating systems. HP, in contrast, has a large presence in the PC market already, is investing in high-end PC technology, and is supporting both UNIX and NT solutions. HP is ready and waiting for the next market transition; Sun is standing steadfastly against the tide.
If Sun is going to survive they must do two things immediately: they must start investing heavily in the production of PC technology, and they must support Windows NT on that hardware. Customers will shortly start demanding it, and if Sun cannot offer it they will go somewhere else.
If Sun fails to do these things, and all indications are that they aren’t even considering them, then we’re going to see a spectacular sunset.
to top