With all of the talk about the onslaught of Windows in the data center and the explosion in Linux server sales – mostly at the expense of Unix platforms – the most obvious question on the minds of the people who use or sell Unix platforms is this: Is Unix dead, or is it just entering a new maturity that comes with middle age? As you might expect, the answer to that question depends on what data you use to measure vitality.
There are two things that matter most in IT land: revenue market share and shipment market share. Some vendors (like Microsoft, Intel, and now Sun since it has been born again into some common sense) have always had an intuitive grasp that dominating shipments can eventually lead to domination of sales. But measuring the vitality of the Unix market strictly by market share statistics – particularly during major technology transitions and economic turbulence – is a bit risky. Things may not be as bad as recent history might have suggested they could be.
In the 1990s, Sun dominated the entry Unix space, where it was the vendor of choice among telecom companies, Internet service providers, and financial services firms who needed rock-solid Unix platforms. HP, with its vast installed base of HP 3000 minicomputers, decided in the late 1980s that it was going to be a big player in the commercialization of the Unix market, and it did just that. Sun took the low-end of the Unix market, and HP took the midrange and tried to create a high-end market. IBM jumped in 1990 with its Unix boxes, but was an also-ran except among True Blue shops until the late 1990s, when it sought in earnest to knock HP and Sun, which by then had mainframe-class Unix boxes, from their perches. Across all system types and customer types, the Unix market has become a three horse race between Sun, HP, and IBM, as they all chase shipment and revenue share. But each of the Unix players has a collection of fiefdoms that they protect. They also have customers with unique pseudo-legacy applications (yes, C and Fortran programs tightly compiled for one particular RISC/Unix platform give vendors a modicum of legacy lock-in, like the mainframes and minis of old), ISVs with unique application suites, and customers with unique IT experience, opinions, and skills. There is a tremendous amount of inertia in the server market, and companies do not jump platforms easily.
That is not to say that customers will not jump. Many will, and history bears this out. It is perhaps not a coincidence that the proprietary minicomputer boom in the late 1970s followed on the heels of two recessions and an oil crisis; PC servers followed a recession in 1982; and Unix servers took off just as a recession got a foothold in the early 1990s, and companies started using them instead of proprietary mainframes and minicomputers. Windows and Linux have surely got a boost from the adverse economic conditions of the late 1990s and the full-blown recession of the early 2000s. While mainframes have had a resurgence in recent quarters, with IBM selling around $1.5 billion of machines in the second quarter of this year, that is about a third of what IBM used to sell each quarter 15 years ago. All share gains are relative, and no one would argue that the mainframe is going back to a $12 billion a year hardware business any time soon. No one will argue that the Unix market will comprise more than half the revenue share for servers again, as it did in the late 1990s. But there is a floor to the Unix business, and the market is perhaps reaching it.
That certainly seems to be the opinion of Don Jenkins, vice president of marketing for HP’s Business Critical Systems unit, which is responsible for the HP Integrity, AlphaServer, and NonStop server lines – all high-end Unix or Unix-capable machines. In general, we’ve seen a gradual increase in the Unix business over the past year, he says. He says that HP, Sun, and IBM have all made strategic moves in terms of pricing and technology to make their Unix boxes more competitive against Wintel and Lintel iron in terms of both initial sticker price and overall price/performance. Even though HP had a tough fiscal third quarter ended in July, the company’s Unix business was up 8%, says Jenkins. That contrasts pretty nicely with a Unix market that itself saw revenue declines of 3% or 4% in the second calendar quarter of 2004 (which shares the May and June months with HP’s fiscal second quarter).
As for the future of Unix, Jenkins unhesitatingly says that HP expects – as do the big IT consultancies such as IDC and Gartner that measure market share – that sales of Unix servers will grow at 1% to 2% a year for the planning horizon, which extends to around 2008 right now. Overall server sales (across all geographies, types, and industries) will probably exhibit a growth rate that is four to five times this level, to be sure. And Unix vendors would love to grow at this rate, or even beat it. But the Unix market is still a bit shaky at the moment, still recovering from the dot-com hangover and only now getting in shape to compete against Wintel and Lintel iron. There is hope, particularly in the entry and midrange markets. The Big Three of the Unix space are, however, all transitioning to new technologies and slashing prices like crazy as they compete. By my own analysis, the newest entry and midrange Unix machines (IBM’s Power5, HP’s Integrity, and Sun’s V20z and V40z Opteron machines with two or four processors) give the best bang for the buck for OLTP workloads – even better than Lintel and Wintel iron. (Those comparisons include reasonable base configurations with an operating system and database on top of them.) Unix zealots can now make the economic case for keeping Solaris, AIX, or HP-UX, even if the sticker shock for the initial hardware can be sickening.
Jenkins says that the real determinants and drivers of the Unix market – and indeed, any market – are having budget money available and the perception that acquired technologies are delivering good return on investment. Our Unix customers have a largely insatiable appetite for IT projects. Time and again, we have seen that the market can consume all of the processing capacity we can deliver. Our business is limited by dollars and staff, not by appetite. He says that HP is pushing Unix not just based on price/performance numbers, but on the idea that the time for getting back a return on investment can be shortened by sticking with Unix and getting a complex project that hopes to deliver either profits or new revenue done sooner is better than saving money on Wintel or Lintel iron on the front end and delaying the project as an IT shop learns new skills. With Unix having 15 years of reference examples in every industry and on every application under the sun, with more mature systems management programs and virtualization features, and with IT personnel in abundant supply with lots of experience in Unix, making the ROI case is pretty straightforward. Any system where skills and tools are better translates into ROI, and that is the value of Unix in the market today, says Jenkins. If this sounds suspiciously like a page out of the IBM playbook that Big Blue uses to sell mainframes and proprietary midrange gear, it is. And, for many customers, it is the truth.
During the downturn in Unix revenues in the past couple of years, Unix shipments have held up pretty well, and Jenkins says that HP expects that Unix shipments will continue to grow in the coming years. The drivers behind these shipments, at least in HP’s core midrange and enterprise markets, will be the growth of basic ERP workloads combined with the addition of supply chain management, customer relationship management, data warehousing and data analytics, fraud detection, and other CPU-intensive workloads. These are also, incidentally, the main drivers of the Windows market, but Windows machines (at least in terms of revenues) have the most appeal to small and mid-sized businesses that use uniprocessor or two-way servers, sometimes four-way servers. In a Unix shop, a four-way machine is really an entry box. That said, all of the major Unix vendors have entry Unix machines that can and do, technically, compete with Wintel and Lintel iron.
Unix is not going away because of Linux, says Jenkins. Linux will take its rightful place, and so will Unix. That place seems to be where a level of sophistication and scalability that Linux cannot yet offer is needed, or where companies have neither the time, money, or inclination to learn a new platform to roll out a new application. Those with good Unix skills and good relationships with more than one Unix vendor are going to stick with Unix as much as is economically feasible. That said, Jenkins expects heated competition in the two-way and four-way server market across Linux, Unix, and Windows platforms. And, as far as he is concerned, by moving to the Itanium processor and a consolidated Integrity platform that can support all of these operating systems, HP is pretty well placed to deal with those competitive situations.