Latest Dispatch from the Unix Server Wars

Published: December 12, 2006

by Dan Olds

The market for Unix systems has become even more competitive over the past year. Only three vendors--Sun Microsystems, Hewlett-Packard, and IBM--are going after a huge market that is pretty much flat in terms of sales revenue growth. So the Unix market can be described as essentially a zero-sum game; any gains made by one vendor will come out of the hide of one (or both) of the other vendors.

The vendors know this better than anyone else, and are thus continually adding to their technology and customer support arsenals in hopes of stealing data center footprints from their competitors. For customers, this results in a high quality, all-you-can-eat Unix server buffet, including a salad bar with all the fixin's. For vendors, it's a never-ending treadmill of constant product improvement and searching for things they can do to add "outside the box" value for their customers and prospects.

We're very interested in the Unix market due to its size and importance to the industry. Just as the Cold War fueled much of the technological revolution from the 1950s through 1990, the Unix wars are driving much of the innovation in server technology and vendor marketing. While advances in vendor marketing might seem less than interesting (and maybe even repugnant) to most data center types, it is the product marketing people who force their organizations to drive prices down, adopt flexible pricing schemes, throw in free or reduced-price services, and generally do whatever they can to get their products into data centers. These folks figure out the tastiest items to put on the buffet line and how to best price the meal to ensure their restaurant is full. . . while the restaurants of competitors are empty. For our part, we want to find out how real enterprise customers feel about the menus put forth by the Unix vendors.

With that in mind, we're releasing results from our second annual 4Q'06 Unix Vendor Preference Survey. The survey is designed to figure out how real data centers rate and rank the "Big 3" Unix vendors--again, that's HP, IBM, and Sun--and their respective products. Our surveys are a bit unique in that we aren't talking to CIOs, corporate IT visionaries, or even bean counter types. The people that Gabriel Consulting Group wants to talk to live and work on the data center floor. In our experience, these are the folks who know most about the day-to-day performance of systems and the support they receive (or don't receive) from vendors. They have direct experience with the platforms, know what they like, know what they don't like, and aren't shy about expressing their opinions. Before we get into the results, a few words on demographics and methodology. . . .

Demographics

We surveyed a grand total of 277 Unix customers (our largest survey to date), focusing primarily on data center personnel--data center managers, IT architects, system managers, and the like. The average survey respondent managed from 25 to 50 Unix systems. About 75 percent of respondents worked in data centers with at least two incumbent Unix vendors, with 37 percent of the respondents reporting that they owned systems spanning three or more Unix flavors. The breakdown between vendors was almost equal, with 34 percent of participants saying Sun is their dominate Unix vendor, and 33 percent each for IBM and HP.

Scoring and Methodology

This survey is a bit different from others you might see in the IT racket in that we aren't trying to figure out the relative market share of the vendors. What we want to gauge is customer perception of the major Unix brands and customer loyalty to their preferred brand. To make the survey fair, all responses are compiled and then normalized. We normalize the data so that no vendor is advantaged (or disadvantaged) by the sheer size of its installed base or the number of respondents in any particular survey. To do this, all respondents were asked to specify which Unix vendor is their corporate standard or the dominant vendor in their organization. Then the total number of responses favoring a particular brand is compared to the number of "votes" for that vendor on a particular factor and scored. For example:

Assume the survey had 1,000 responses, 500 that have standardized on Sun, 300 that are strong HP customers, and 200 that have chosen IBM as their dominant Unix vendor. When asked which Unix vendor had the best dressed salespeople, four participants responded Sun, three picked IBM, and three said that HP salespeople were particularly natty dressers.

While the raw scores favor Sun, the normalized score (which is simply the number of votes divided by the number of respondents who have standardized on that brand) shows that HP is the winner of this beauty contest. HP wins because the company captured a larger number of first place votes than the number of HP respondents. In this simple example, 100 Sun customers defected and voted for HP. We are referring to this normalized number as the Vender Preference Index (VPI).

The VPI computation yields an easy-to-understand score for each vendor and a gauge of installed base loyalty. For quick reference, a VPI score greater than 100 means that the vendor in question was selected by a number of respondents greater than the number of respondents who have standardized on that particular brand of server. VPI scores greater than 100 are very good. A VPI score of exactly 100 means that the vendor was chosen as a leader by exactly the same number of respondents who have standardized on that vendor. VPIs of less than 100 are, of course, bad, and mean that the vendor in question has suffered defections (at least in terms of survey voting) from their own self-selected installed base. While there are certainly more complicated ways to compute the results of a survey such as this, we believe that this method captures the data we are looking for: how customers perceive Unix vendors and who they choose as the winners and losers.

The survey is divided into several components, including questions about their Unix trends, adoption of virtualization/consolidation, along with sections discussing their perceptions of specific Unix vendors. This article outlines the overall results from the survey's "Unix Vendor Face Off" section, where customers are asked to select the vendor they believe is the leader in a specific technology or support category. The "Face-Off" section is split into two categories, the first covering technology issues and the second asking about vendor and customer support topics. Let's take a look at the Technology Criteria category first.

Vendor Face-Off: Technology Criteria

The questions in this category cover the gamut (or at least three-quarters of a gamut) of topics relating to systems, from performance to availability to manageability. Customers are asked to rate vendors on Raw Performance, Observed (real world) Performance, Scalability, Availability, and Observed Availability. Customers also get to grade vendors on operating system features, o/s quality, virtualization capabilities, and out-of-the-box quality. The chart at the left shows the average scores of the vendors on all of the Technology Criteria questions.

As the graph shows, IBM wins the technology section of the survey with an average VPI score that is essentially 100. Sun places second with a score of 87, and HP claims last place with a 78. (Note: Due to the VPI scoring, explained above, row totals on the graphs will not equal 300, it isn't a zero-sum scoring methodology.) Both HP and IBM saw their scores drop, although IBM stays very close to the par score of 100. HP is not as fortunate, as its score of 78 shows that an average of 22 percent of the HP standardizing survey respondents did not "vote" for HP in this section of the survey. Sun made significant gains, improving its average VPI technology score by 10 points.

Vendor Face-Off: Vendor and Customer Support Criteria

A data center can not live on feeds and speeds alone; it needs vendor services and support to survive and thrive. Well, OK, maybe there are some data centers that don't need help from vendors, but most depend on their vendors to a greater or lesser extent to help get things done, make sure needed patches get written, and keep customers informed about the future. If the vendor can add some additional value by advising customers on how to best use the equipment and save money--all the better.

In this section, customers get to rate Unix vendors on topics that are, for the most part, not system specific. Some of the factors include vendor break/fix maintenance, professional and architectural services, sales force/channel quality, and commitment to the Unix platform. Additional questions cover how well vendors stick to their roadmaps, keep their promises, and help customers increase IT efficiency and reduce TCO.

As can be seen on the chart, IBM still leads HP and Sun in the vendor and customer support area. Again, as we saw in the Technology Criteria results, overall VPI scores are down--only Sun manages to improve its score from last year, while IBM drops by about 10 percent and HP loses a bit less.

Vendor Face-Off: Futures

I will close with results from a couple of future-oriented questions. The first question asks customers which Unix vendor has the strongest roadmap. IBM notches a win here, with a score of 97.5. IBM has built a solid reputation for delivering on its Power server roadmap promises over the past couple of years.

Sun posted a score of 86.42, a solid 10 point improvement over its score of 77 last year. I believe Sun's scores on this question have been aided by introduction of "Niagara" Sparc T1 systems earlier in 2006 and a very good "Galaxy" Opteron-based server lineup, which also supports Solaris. Customers may be giving Sun credit for successfully delivering new technology (albeit a bit late), and expecting that Sun will deliver its long-awaited "Rock" Sparc-based systems on time.

HP falls off the table in this category, with a little over 20 percent of its installed base in the survey voting either for another vendor or for the "no difference" selection. On this question, a little over 12 percent of respondents did say that there was no difference between vendors on this question.

In our 4Q'05 survey, we asked customers to tell us which vendor they thought was most likely to drop out of the Unix market. At that time, most customers (about half) thought that HP was most likely to give up on Unix. We asked the same question in this survey and found that much of this feeling has subsided. Now, 50 percent of customers believe that none of the major vendors will leave the Unix market. This is a great object lesson in just how fast customer opinion can change in a relatively short period of time. It also indicates that HP may be getting past the Itanium migration difficulties that have been dogging the company for the past two years.

The last question in this section of the survey asks customers to select the vendor they believe will dominate the Unix market in five years time (2011).

As can be seen on the chart to the right, most respondents feel that IBM will driving the Unix bus in 2011, followed by Sun. If you recall the numbers in the demographics section at the beginning of this article, you remember that the survey base was divided almost exactly equally among the three vendors. In other words, equal proportions of respondents said they have standardized on each of the major Unix vendors. Examining the chart again, you can see that Sun pretty much held all of its base, while a solid chunk of HP standardizers (about a third) said they believe IBM will be the dominant Unix vendor in the future.

So What Have We Learned?

The landscape of the Unix market has changed markedly since our last survey, according to our survey respondents. IBM still wins overall and in most individual categories, although not by as wide a margin as last year. IBM has done a typically solid job of pushing an aggressive technical roadmap, but, in my opinion, did an even better job of talking up the business value of its systems. The virtualization, workload, and system management enhancements in the System p5 line have done a lot to convince customers that IBM truly wants to solve customer problems and not just push hardware.

The real action in the survey was between Sun and HP over second place honors. I was a bit surprised to see Sun finish behind IBM and HP last year, both in overall results and in the majority of survey categories. In many cases, Sun wasn't just third, it was a distant third. However, Sun earned the award for "Most Improved Unix Vendor" this year with its come-from-behind thrashing of HP. Sun has had a good year, particularly when compared with the last five years. Sun has successfully introduced its long promised T1000 and T2000 Niagara systems and has build a fast-growing Opteron server base. Sun re-captured the Unix sales revenue crown for at least part of the year, and managed to significantly increase revenue. All in all, a pretty good job, and I believe customers have noticed the difference in the 2006 version of Sun versus the 2005 model.

HP's results are difficult to explain. The company didn't make any horrible Unix-related missteps in 2006. HP's Unix sales figures held up pretty well--not fantastic, but not awful, either, and HP is right in the thick of things. So why did HP's scores drop so much in our 2006 survey? We think there are two factors that negatively impacted HP's showing:

1. IBM's and Sun's Performance: IBM, as noted above, just keeps rolling along. IBM is on top of the Unix game, is playing aggressively, and will be difficult to beat. Sun is the comeback kid, and has done a good job of executing its plan and have garnered a lot of positive publicity over the past year. By contrast, HP, from a mindshare standpoint at least, looked a little lackluster. HP's "Montecito" Itanium server launch was late--thanks to Intel--and fell a bit flat for various reasons. I am hard pressed to think of any well-publicized HP Unix achievements in 2006. It could be that I didn't pay as much attention as I should and that I missed some HP Unix good news. But, off the top of my head, I can reel off three or four highly public Sun announcements and almost as many for IBM. I can't say the same for HP.
2. Scandal and Skullduggery: Unfortunately for HP, this survey was out in the field during the height of press coverage on the HP pretexting and spying scandal. I am not sure if this would have much of an impact on HP's scores, but in compiling the survey, I saw a significant number of responses that mentioned the scandal. Many of the comments were joking, but others revealed feelings of disgust and disappointment in HP. Is it fair the HP's Unix division should be penalized for what were dumb actions taken by its board of directors? No, of course it isn't fair, and the scandal had absolutely no impact on HP's products, services, or dedication to its customers. But it looks like the bad publicity had at least some impact on HP's showing in the survey. I would bet that the damage is temporary, but only time will tell.

Future Gabriel Consulting Group reports will discuss other aspects of the survey. In addition to the Unix vendor beauty contest outlined above, Gabriel Consulting Group asked customers to share their Unix usage trends and information about how they are using their systems. We also took a deep dive into each vendor installed base by asking them how they believe their vendor of choice will tackle specific challenges. If you are interested in hearing more about this study, or would like to participate in future Gabriel Consulting Group surveys, you can reach us at: gcginfo@gabrielconsultinggroup.com

RELATED STORIES

Who's Ahead in the X64 Server Wars?

Who's Ahead in the Unix Wars?

Stop Arguing About Cars and Start Managing Fleets

Dan Olds is the founder of Gabriel Consulting Group, an IT industry analyst firm that focuses primarily on IT as it relates to business, showing companies how technology can be deployed in a more efficient and effective manner. You can contact him at gcginfo@gabrielconsultinggroup.com.