RIP Alpha

by Nicholas Penne "Nicholas Penne" is a computer professional with over thirty years in the industry, in such diverse fields as Software Engineering, Education, Services, Operations and Marketing. Since 1990, Nick has been a competitive, strategic and industry analyst. In the industry analyst community, excessive candour can sometimes impede the ability to make a living, so Nick chooses to remain anonymous at this time. But these days, that doesn't preclude direct communication. You can reach him here.

THE COMPAQ ALPHA DIVESTITURE and its implications for Compaq's future has been a favourite topic for industry analysts and the trade press. Like the blind men describing an elephant, they get parts correct, but miss the big picture.

Rules for Speculation

Although many will speculate on past and future courses of action for Compaq, there are certain ground rules that time and again have proven themselves mandatory in order for any speculation to be deemed reasonable.

Rule #1. Is it going to cost anything?

Although Compaq has been known to buy businesses, they have never significantly increased their investment in an existing business. Any speculation that requires Compaq to increase funding in any substantive way is beyond the realm of possibility. Note that even groups internal to Compaq have run afoul of this rule.

Rule #2. Do they actually have to *tell* anyone?

At Compaq, what people commonly refer to as "marketing" is allowed only in the PC division. The levels of approval required for any other Compaq organization for something as simple as a press release are astounding. So give up any speculations that require "marketing."

Rule #3. Not in Houston?

At Compaq, membership in the executive "good old boys/girls" club is limited to those in Houston. A part-time office is not enough. If the executives who would be in charge of your speculation reside elsewhere, they will be powerless to assume any role other than that of caretaker. (They will also be very busy removing the knives from their back.)

Rule #4. "Software is something Partners do."

A commonly held belief by those in charge from Compaq Classic (and actually articulated by Enrico Pesatori) is that "software is something partners do." If your speculation requires Compaq to get into a software business, you are engaging in wishful thinking.

RIP Alpha. What's next?

Given recent history and Compaq's internal prejudices, the decision to divest Alpha was inevitable. This decision has its roots as far back as Compaq's original reason for acquiring Digital, which was to acquire a world-class services organization. Although there was some noise made about acquiring an enterprise capability, there was then, (and still today) no understanding of what it takes to be a player in the enterprise market by the executives controlling Compaq's strategic direction.

Those executives can only mindlessly follow Microsoft's "lead" for a future enterprise direction. Besides, there's that pesky rule #4 to consider. The word "Solution" has been much in vogue, but upon even cursory examination, these "solutions" are found to be collections of components. Since there's yet to be a software ISV who will a priori refuse to make their product available on other vendor's hardware (especially when they're all Intel-based), this form of "differentiation" is illusory at best. Given the timing, it seems likely that Alpha was on the table during the HP talks, but that HP was solidly behind the decision to divest, and may have made it a prerequisite to the deal.

Compaq has now completed divestiture of:

* Windows NT on Alpha
* Commercial software for Windows NT
* The Alpha processor and its associated engineering groups
* Compiler technology and products

Each of these events can be seen to be small steps in a larger plan. Each step perturbed the customer base, but each in itself was small enough to allow the corporation to survive, and to mislead the customer base that each step was the last. Ignoring (for now) the potential HP acquisition, the process will not be complete until Compaq has divested itself of the rest of the Business Critical Server Group (BCSG), which includes:

* Tru64 UNIX
* OpenVMS
* NSK/Himalaya
* Custom Systems

Although it was inevitable that Compaq would/will eventually divest these groups, this deal with Intel at least allows as many as 1000 engineers to be gracefully fired under the pretext of transferring them to Intel, without having to incur the not-insignificant expense of laying them off. But if these engineers had wanted to work for Intel, they already would be. Consider that nearly every member of the StrongArm design team eventually found employment elsewhere. Some analysts have said that there are limited alternatives for these designers in the New England area, but perhaps they haven't seen the half-page ads in the Nashua Telegraph for design positions with Sun Microsystems, notwithstanding Sun's recent layoffs in other areas.

Intel has given itself some time to try and keep these engineers, as wholesale defections are unlikely until the "stick-around" bonuses vest.

The divestiture of Alpha has little or no real impact on NSK/Himalaya. The Himalaya customer base was already facing a port to a new architecture, so nothing changed for them. The previously announced port to Alpha was targeted at EV8 (the now-defunct next generation of the Alpha architecture) because previous versions lacked the required processor lockstep functionality, so substantial porting effort had yet been expended.

Changing to an Itanium Product Family (IPF) base is no worse, and probably a better solution for Tandem, as it would allow them to take advantages of synergy in their low- and high-end development streams. Tandem's previous port from a proprietary architecture to MIPS has already favorably positioned their code base (the first port is the hardest), with most if not all of the problem areas already identified. Tandem was never really incorporated into Compaq, and still operates as a mostly independent company. Previous to the possibility of the HP acquisition, the most likely scenario for the future of Tandem was to follow in NCR's footsteps, and be spun-off or sold.

A port of VMS is feasible, although as is been shown to be typical Compaq operating procedure, no VMS engineers were consulted before the decision was announced. The work to isolate the hardware-aware components of the operating system has already been mostly done. The VMS engineering community is certainly up to the challenge of working around any architectural limitations of IPF they may encounter, as it's problematic at this point how open Intel would be to any IPF architecture enhancements.

But it seems unlikely that Compaq would actually spend the money required to do a complete effort (see rule #1), especially on such areas as conversion tools, and emulation of binaries. If Compaq were serious about VMS on IPF, rather than wait until the decision was announced to start the process, they would have previously negotiated and simultaneously announced agreements with key ISVs for important customer functionality, like Rdb, spun off to Oracle by Digital (but there's that pesky rule #1 again).

Instead it was left to the business groups to play catch-up. At first blush, it seems much more likely that Compaq would sell or spin off the VMS business. This sold or spun-off business would not be encumbered by any requirement other than the port itself, making it easy to push all application requirements on the customers of the new business.

But divesting the VMS business is not an easy problem to solve. VMS service revenues, presently amounting to over $2 billion per year, are still important to the Compaq services organization. In recognition of this, the VMS group has already been "divested" to a certain extent, with the alignment of VMS with that services organization. But this is more of a shotgun marriage. Over the years, the services organization has dabbled in engineering, but its track record has been dismal. Revenue and profitability pressures have historically resulted in shortsighted cutbacks to the implementation of new functionality in these services-run engineering groups, and only enough maintenance done to barely satisfy customer contractual commitments.

What little new functionality is haphazardly incorporated from customer-funded projects, and results in a patchwork product over time.

A port of Tru64 Unix, while feasible, would be a waste of effort. What little differentiation Tru64 Unix offered (mostly in the areas of performance and RAS, or reliability, availability and serviceability) was mostly due to the underlying AlphaServer system. As yet another version of Unix available on IPF, there is little to distinguish it. (In absolute terms, Tru64 clustering technology is not all that better than the competition.) There previously was no competition with Linux at the high end; this protection evaporates along with Alpha. Compaq will promise, but when it comes time to commit resources, in the absence of an HP acquisition they will most likely spend a much smaller amount of money on porting key Tru64 technology to Linux, and on helping Tru64 customers move to Linux.

Digital never invested enough in Unix development and marketing to be perceived as being committed. Compaq didn't change this, and won't now. Even if Compaq were to invest in an IPF port rather than merely redirect existing development resources (but see rule #1), it would be an uphill battle to carve out a new customer base, as any semi-conscious customer will realize (and the competition will delight in pointing out) that any vendor who has abandoned the Unix customer base as many times as Compaq/Digital can certainly not be trusted yet again.

What little acceptance Tru64 has gained in the ISV community will be completely lost by this move. Previously, it was possible to get some ISVs to grudgingly port to Tru64, mostly on the basis of Compaq's showing in the High Performance Technical Computing (HPTC) and elecommunications (TELCO) markets. But with Tru64 as yet another UNIX on IPF, ISVs (who have a choice) will wait to see if Tru64 on IPF garners any following. Since generating a following for Tru64 flies in the face of both Rules #1 and #2, ISV support would be limited to those who have no alternative customer base.

But what ever happened to Compaq's UNIX/ALPHA market strategy?

Although there has been much posturing and saber rattling, the only rational way to infer a strategy is to go by events and trends (and follow the money).

At Compaq, there have been only two markets keeping the lights on in BCSG. HPTC (which includes life sciences) and TELCO.

The HPTC market is loyal to only the fastest iron out there. Lately, that's been Compaq. With Alpha, Compaq has had the fastest iron for a few years now; and as a direct result, they now are number one.

This market has been telling Compaq that they want RAS, but this is a second-order consideration. This market's over 40-year track record tells us they will not stay long with any vendor that does not have the fastest iron.

(There are those in Compaq who thought RAS would keep this market, even if leading-edge performance were missing. They were wrong. This market's traditional behavior will not change overnight to accommodate this opinion.

Digital's own experience with the VAX underscores this.)

Compaq's existing investment in Alpha was not enough to maintain this lead. IBM's Power4 announcement was a wake-up call for Alpha engineering, which realized that existing projects would not maintain a convincing performance lead. The EV8 (next-generation Alpha) project was in substantial disarray.

Managers were making unaffordable demands for resources. The project was too big and too complex. Proposals to simplify were ignored (there was a proposal to split EV8 into two projects, a single-stream performance project and a on-chip SMP project; these projects would have been affordable, and might have allowed Alpha to maintain its industry performance advantage); but middle management was either unaware of Rule #1, or convinced themselves that Alpha was sacrosanct. In a year, maybe two, other vendors would field the fastest iron, and Compaq's HPTC market would evaporate as fast as it appeared.

Even aside from IBM, even IA32 architectures were on a trend to overtake Alpha. The only way out of this dilemma presented to Compaq's upper management was an infusion of more resources. When pressed, they said this was the only way to meet their goals. They were believed. Rule #1 remains intact. EV8 is history.

The projects kept (Marvel, the follow-on to the GS series; Privateer (now announced as the ES45); and EV7, the faster Alpha) were all substantially complete, and well understood. There would be no savings in canceling these projects, and would allow at least a bit more time to gracefully cut off the customer base at its ankles, rather than its knees.

There is no particular tie to the Alpha architecture in the TELCO market.

For them, much of Compaq's value-add is in its AlphaServer system engineering. But without HPTC, there isn't enough critical mass in TELCO to maintain a high-end system engineering line. Although possible, it doesn't seem likely that systems engineering in ISSG (Industry Standard Server Group) will be improved to make these systems as desirable to Telco's, as this would break rule #1. There are also initial signs of a coming shakeout and consolidation in the TELCO base as well.

Therefore, Compaq's Alpha strategy has been shown to be the concession of the HPTC market, and to go out of that business. This decision merely made it official. If Compaq wanted to continue to win in the HPTC marketplace, a more rational play would have been to make a deal with AMD. An "AlphaHammer" would have been a clear performance (and therefore HPTC) winner. But again, no one in HPTC was consulted prior to the decision being announced.

The show is run from Houston

Another clue to the inevitability of the divesture of the rest of BCSG can be inferred from their choice of executives. From day one of the Digital acquisition (and considering Rule #4), it was obvious that running the Alpha businesses as an ongoing concern was not a goal, from their choice of Enrico Pesatori (a master of consolidation, but never a choice to grow a business) to most recently Rich Marcello, a widely acknowledged expert at keeping the lid on the VMS organization and customer base, all the while extracting as much cash from the cash cow as possible.

These ex-Digital groups always were the stepchildren, with no presence in Houston. Mary McDowell, vice president of ISSG, actually even considered the ex-Digital groups to be her competition. The most externally visible sign of this was the ill-considered and ill-advised Unisys deal, whose only merit was to provide Mary with a "competitive" answer to the AlphaServer GS-series (code name "Wildfire").

But the key to differentiation is software, isn't it?

Analysts and others say this. Analysts and others are right, especially when you consider the razor-thin margins on hardware these days.

When Compaq's new CTO was first hired, he was perceived to be a brilliant man. He initially commanded great respect, especially among the engineering community. One of his first talks to the engineering community included his belief that Compaq had some of the best software engineering talent and software technology in the industry, and that Compaq should and would leverage these strengths for differentiation in the market. Perhaps in preparation for this latest round of divestitures, take note that this new CTO was given responsibility for Compaq's troubled internal IT organization.

He will be so busy fighting alligators he won't even think of mentioning the "S" word. Note that software is conspicuous by its absence from any of the initiatives he's been publicly talking about. Given the current situation among Compaq's research community, it wouldn't be surprising if he were to quietly leave as soon as he can negotiate a new position. The proposed HP acquisition gives him an even better rationalization to leave, should he want one.

The divestiture of the compiler groups and technology made no real business sense whatsoever. They were a source of significant differentiation for Compaq, especially in the context of the Alpha divestiture. They were perfectly positioned to make Compaq the source of the best compilers available for IPF. (And IPF sorely needs excellent compilers to survive. Intel has had to have its engineers hand-optimize applications to realize even their dismal performance numbers to date.)

But remember that Rule #4 mandates that "software is something partners do," so the compiler groups had to go.

Didn't Compaq say their customers are okay with this?

Promising support until 2007 shows how little Compaq upper management understands the enterprise (in the sense that nearly everyone in the industry except Microsoft and its lackeys use the term) customer, and shows how little they value the enterprise experience of the ex-Digital (or even the ex-Tandem) crowd. As any of them could have told Compaq's upper management if they had been consulted, no enterprise customer would even consider buying a platform today that will be obsolete in six years, as although the investment in hardware might be amortized by then, investments in software and especially infrastructure have a much longer horizon. This move is perhaps the last straw for the ex-Digital enterprise customer base.

Compaq claims their customers are on-board and okay with the decision. When Compaq got rid of Alpha NT, they said the same thing. Unfortunately, that announcement had to be postponed a week, one reason for this was because they hadn't actually talked to any customers. At the last minute, they dug up some customers who would say what was desired. No key customer lacks the presence of mind to publicly denounce such a decision and lose whatever leverage they may have left with Compaq. But customers will of course vote with their feet when they decide it is in their best interests to do so. The most likely scenario for these customers is to freeze new development on Alpha, and pursue a strategy of coexistence and long-term migration to their new platform of choice. This will be reflected in a near-term loss of off-base sales, and a longer-term downward trend in upgrades.

What about the proposed HP acquisition of Compaq? How might it change things?

The decision process fits right in with the rules for speculation. Even the knowledge that talks were going on wasn't shared with many high-level vice presidents, as shown by how unprepared they were at the employee briefings to field questions or enlarge on anything not already in the press releases.

Indeed, given the near-unanimous negative reception of the proposed acquisition by Wall Street and the financial and industry analyst communities, it's obvious the negotiations were conducted in a sealed room with a dwindling oxygen supply. It would have been easy enough to run such a proposal by key analysts, but this wouldn't be in keeping with the Compaq way of doing business. (Note that security isn't really an issue. Analysts know how to keep our mouths shut when we have to; and multiple scenarios can be evaluated in parallel. Well-run corporations conduct contingency planning all the time.)

It's also not likely the acquisition will go through as proposed. Even assuming stockholder approval, it will have to get approved by the Federal Trade Commission and the European Union (those same folks that blocked HP's acquisition of the consulting arm of PriceWaterhouseCoopers). The resulting company with perhaps 70% of the U.S. Consumer PC market can hardly be positioned as anything other than a monopoly. After the regulators get through with this deal, it's anyone's guess what it will look like.

But what if?

The operational situation for VMS would remain the same. As to environmental issues, HP certainly has the experience and ability to properly run a business like this, as the appropriate model would be nearly identical to HP's MPE-iX business (although recent events in the MPE-iX arena would tend to cast doubts). Unlike Compaq, HP appears to have an understanding of the Enterprise market, and the desire to keep (and expand) this customer base. HP might even annul the marriage to services.

The situation for NSK/Himalaya is very similar to VMS. No changes in operation, but HP makes for a much more hospitable host for a business targeted at enterprise customers.

The acquisition would put the final nail in Tru64 UNIX's coffin. It would be sheer madness for HP to invest in a second IPF port. The only difference in this scenario is that HP-UX is the beneficiary of the Tru64 technology HP finds interesting. Although immediately painful, the situation for Tru64 customers could be substantially better in the long run, as they could be provided a reasonable migration path to an enterprise-class UNIX at a company that understand how to deal with an enterprise customer. µ