Mad Dog 21/21: Clouds Gather Over The Server Business

March 18, 2013 Hesh Wiener

Aggregate revenue in the server business didn’t grow last year. As we previously reported, based on Gartner data, it might not grow next year either. This is tough but tolerable for vendors with market breadth, but it is punishing for players with a tighter focus. Moreover, in an ironic twist worthy of an O. Henry plot line, the effort by vendors to boost cloud computing have exacerbated matters, draining funds that might otherwise be spent in customers’ glass houses.

This is particularly so in the midrange, which, like Newport, once attracted The Four Hundred but now often lives in ignominy, like a corporate Claus von Bulow.

What the market researchers say is that worldwide server revenue rose during last year’s fourth quarter after three quarters in doldrums. The year-end improvement was attributed to a leap in mainframe shipments as IBM began volume deliveries of a new generation of System z processors. Sales of other kinds of servers, including Power boxes and the vastly more popular X86 machines, declined year-on-year.

O. Henry: The twists in William Sydney Porter’s personal plot line led him to adopt a pen name on his way from prison to fame.

Depending on which guru’s slicing and dicing of the data one wants to examine, there may have also been an upturn in high-end X86 servers even if the market for more ordinary X86 iron skidded. The market analysts failed to split sales of various processor types by destination, by which we mean whether they were sold to end users or to computer services companies, but other reports indicate sales to providers of cloud services is pretty substantial, particularly where X86 systems are concerned. Cloud services might turn out to be the thing that gives the X86 business, which is going through a midlife crisis, a whole new era of growth.

That kind of midlife change is what William Sydney Porter went through in 1902 when, at the age of 40, he got out of prison after serving about three years for embezzling a bank. Under a new name, O. Henry, he became a writer, and a very successful one at that. By 1906, when he published his second collection of short stories, his prominence allowed him to poke fun at and bury the American cultural phenomenon Mark Twain called the Gilded Age about the time he said, “What is the chief end of man?–to get rich. In what way?–dishonestly if we can; honestly if we must.” Porter, also a wit but light in mood where Twain was dark, stuck his blow for a new public sense of social self-worth by entitling his second volume, largely about ordinary people in New York City, The Four Million.

Ward McAllister: Believed, as did other privileged folk of the Gilded Age, that only the elite were truly important.

The title referred to “The Four Hundred,” a phrase used by social commentator (and arbiter) Ward McAllister, who declared that, during the post-Civil War era, there were only 400 people in all of New York social society who counted. As it turned out, this number coincided with the capacity of the ballroom in the home of the prominent socialite, Caroline Astor. McAllister is credited with (or blamed for) the rise of Newport, Rhode Island, as the pre-eminent out-of-town destination for New York’s elite.

A hundred years later, Newport would be in the spotlight when one of its residents, Claus von Bulow, was controversially arrested, tried twice and ultimately acquitted of attempting to bump off his wife, Sunny, using an injection of insulin.

O. Henry felt about The Four Hundred pretty much the way the press, even organs aimed at the rich and famous, felt about von Bulow. By naming his book The Four Million, O. Henry was declaring that it was not the few who made New York (and America) so wonderful, but rather the many. Had O. Henry written about computers rather than people he might well have paid little attention to mainframes, IBM i platforms, and other relatively exotic machines in favor of those serving with a more common touch, such as garden variety X86 boxes. It was O. Henry’s wit mated with IBM’s name for the child of its System/38 that begot the name of this very publication back in 1989.

Servers: Gartner’s sketch of the server market shows it was flat last year despite a good showing in the fourth quarter.

Just as O. Henry’s stories often end with a turn of events the reader is unlikely to expect, so, too, do chronicles of industrial development. The disk drive business pioneered by IBM became a model for disruptive technological change used by Clayton Christensen in his studies of industrial culture. Disk hardware has long since ceased to be part of IBM’s product line. Were it not for storage management software, IBM (and very likely EMC and others, too) would no longer be viable participants in the data storage industry. IBM’s production of server hardware based on Power and mainframe chips has only survived thus far because Intel failed to develop an effective strategy for its Itanium chips.

By contrast, IBM’s creation of new information technology practices embodied in its services and software operations has proved to be an excellent source of revenue and profit. But even for an institution as strong and smart as IBM, the growth of cloud computing, which replaces portions of customer premises computing, including some of the most profitable parts of that trade, may prove to be a step too far. And if the change is difficult for IBM to negotiate, it will be even more challenging for many of IBM’s customers.

Gartner, whose numbers we cited earlier, has looked at cloud computing and,as reported by Forbes, believes the cloud business is growing at nearly 18 percent per year. Moreover, a key part of cloud computing, infrastructure services, which is a fancy way of saying remote whole system image hosting, seems to be growing at twice that rate.

All told, cloud computing is, Gartner says, generating revenue in excess of $100 billion a year, making it twice the size of the server business.

Rackspace Hosting is a relatively pure cloud company bringing in about $1.3 billion a year in revenue. In its most recent financial reports for 2012, Rackspace says it spent about $218 million, or not quite 17 percent, a sixth of its revenue, on what it calls customer gear, mainly servers. (Switches, routers, and some other hardware figure in here, too, but the lion’s share of this money is for server hardware.) Let’s guess that servers cost Rackspace about 15 percent of revenue. If that is true across the field, cloud computing accounts for $15 billion a year or 30 percent of the server market. But suppose Rackspace is very heavily dependent on infrastructure services and not a huge player in software-intensive or personnel-intensive value added pieces of the cloud business. Would it be reasonable to guess that 10 percent of the cloud’s intake is for servers? If so, the cloud is providing the server business 20 percent of its revenue.

Now if 20 percent of the sever vendors’ output goes into the cloud, the end user market for on-premises computing is not flat, it is down and it is only 80 percent the size it seemed when the cloud segment was counted in with the end user in-house market. (If the cloud’s buying spree comes to 30 percent of the server market, as it does at Rackspace, the situation is even more dramatically different than the headline “flat” assessment of the Gartner market estimate.)

The Cloud market: Growing very quickly, cloud computing is already twice as big a business as server hardware.

The cloud is not only a large and growing part of the server business, it is also a powerful, probably just about irresistible, force for standardization, the antithesis of the kind of computing that gives the on-premises data center market its personality.

When a user has a glass house, even a small one, that customer’s choice of hardware, software, and operational characteristics can be specifically tailored to perceived needs. These needs don’t have to be logical. In fact, they probably aren’t. But the technology team at the user company is familiar with and probably pretty happy with the shape of things and might even take price in some unique ways of getting the job done. Outside contractors with expertise in the user’s operating environment, middleware, and application software cater to the user company’s quirks. In that service lies a bond. Whatever is unique at each site may be seen as optimization or tailoring, but it is also a barrier to migration. User organizations are loath to give up their special way of doing things. They are attached to their computing culture.

But if they move to the cloud, there is a clear advantage to giving up unique choices and that advantage makes itself visible in the form of price quotes from cloud vendors offering standardized computing power at lower rates than custom hosting configurations. By moving to a standardized cloud the user company may well lose uniqueness, some of it valuable, in return for improved economy, ease of migration (should a particular cloud vendor’s service turn out to be disappointing), better access to backup facilities and other shared resources, and elimination of a lot of the contractual wresting associated with acquiring, maintaining, and supporting on-premises systems.

IBM might find some opportunities to profit by providing cloud access to proprietary systems, but if it wants to go for the big bucks and the fastest growing part of the cloud opportunity, it must make its biggest and best efforts in the provision of standardized capabilities to customers who are able to take advantage of the emerging cloudy ways of computing. To get the best of both proprietary and standard worlds, IBM may have to do something that it has been able to avoid for years: migrate its proprietary environments to standardized systems or to layers of code that can ride on standardized systems and provide the features offered by proprietary environments.

Some of IBM’s would-be rivals, now vanquished, proved that IBM’s flagship mainframe software could be supported on hardware running a Unix variant along with special firmware. Solutions with less efficiency but capable of running on stock standardized hardware and software, such as X86 with Linux, are only a few steps away from emulators that seemed to work pretty well. IBM has an X86-based mainframe environment for developers, but it’s not widely seen as a commercial grade offering.

The mainframe market is not the only one IBM must address if it wants its proprietary offerings to have a life in the cloud. IBM could undoubtedly port IBM i and AIX to an industry standard environment if it wished, but if it takes a while to do this, ironically that industry standard might be ARM and Chrome rather than X86 and Linux. Slowly moving Power-based software to clouds where LAMP lives as the light dims would make the move a Gift of the Magi.