|HOME||PUBLIC LIBRARY||ANOTHER PERSPECTIVE||INFOPERSPECTIVES||CONTACT|
You would think that computer companies would eventually learn to spot areas where standards make so much sense they are obviously inevitable, and develop their products accordingly. But they do not. A current example of this foolishness is the blade server market. Blade servers are an awful lot like the racks of servers everyone already has, with one basic exception: Instead of the customer wiring all the gizmos in the rack together and managing the mess, blade servers have the wiring built in and include system management tools.
With a rack configuration, a customer can mix different processors, even different brands; add routers and firewalls and storage boxes from another bunch of suppliers; and in the end get a lot of computing into a pretty small space. If a particular component becomes inappropriate, it's not too hard to replace it with one that's more suitable for the task at hand. This is why rack-mounted
It looks like blade servers are just the next evolutionary step. But wait; there's less. So far no vendor offers a blade server that lets the customer plug in another vendor's storage module or router, let alone a processor. Hewlett-Packard may have come closest by building one of its lines of blade servers to an industry standard developed within the telecommunications industry, but because nobody else in computing likes that standard (including the former Compaq, which has been eaten by HP), it might as well not be a standard at all. All the other players in, or threatening to enter, the blade market, such as IBM, NEC, Dell, RLX Technologies, and Sun Microsystems, seem to see blade servers as a way to lock up users who might otherwise buy racks that take standard equipment modules.
As it stands, each of these vendors is trying to glom the whole market segment by locking out direct competitors. For most, if not all, of them, this will be a recipe for disaster. The vendors' greed will not only turn off customers but also get in the way of the real value that blades can deliver much better than racks: servers with functional modules. By functional modules, we mean blade servers that have things like database engines, Web engines, e-mail engines, and even more specific engines such as ones that perform some (or maybe all) of the work of an ERP suite.
It won't matter whether hardware companies, software companies, systems integrators, or just clever marketing outfits offer these products. The day a customer can buy any of several basic accounting packages in a can, that can run them for, say, 100 seats, is the day the blade server business will have it made. Anything short of this, which is where things stand right now, and blades just cut the cheese.
We're not surprised that every one of the blade server vendors wants a total lock on the market: Nothing makes money the way a monopoly does. But, with few exceptions, monopolies self-destruct. And good riddance to them, too.
Eli Whitney once thought monopoly was the answer. He might have expressed his sentiments like this: "Keep your cotton picking hands off my gin."
It didn't work out for Whitney. He was, arguably, the third most-important inventor in U.S. history after Edison and Bell, and he went bust in the cotton gin business. A cotton gin is a machine that picks cotton out of bolls and leaves the seeds behind. The "gin" in cotton gin is a shortened form of the word engine.
Whitney eventually learned his lesson, as Yale graduates do, and later prospered by bringing the concept of interchangeable parts into manufacturing, a concept that opened, rather than closed, markets. But between 1793, when he invented the cotton gin, and 1800, when he revolutionized the musket business, Whitney paid dearly for his doomed attempts to control and viciously exploit his invention.
Perhaps it is all just as well. Clones of the Whitney cotton gin, which would be illegal now but back then fell into a gray area created by flaws in the relevant patent laws, proliferated. At the turn of the 19th century, Congress straightened out patent law, but by then Whitney was bust, all his potential rewards for the gin eaten by years of fruitless, costly litigation.
Meanwhile, Whitney's purloined invention initiated a chain of events that eventually, but quite directly, led to the huge rise in slavery in the American South and, finally, to the Civil War. If Whitney had had his way with the gin, he might have remained in the South, amassing bad karma along with a ton of money. Fate, however, kept his hands empty and clean in the 18th century and amply rewarded him in the 19th.
We wonder what, if anything, two hundred years from now, people will say about greedy people in the computer business, and what they will say about equally ambitious but more enlightened ones. We wonder if they will spot those that started out as greedy as can be but somehow grew wiser and rose to greatness.
Maybe you should put copies of this newsletter in some kind of time capsule to give our descendants a little help figuring things out. But even without such excellent material, they will undoubtedly be able to see how things work.
Every time a new wrinkle in computing comes along, a bunch of companies, some old and some new, go into the hot new market saying (but not really believing) that somehow their situation is different. They all imagine they will come up with products so good that nobody will mind if one great company controls the whole market.
Maybe these computer companies know too little about the cotton gin and too much about the liquid kind, which, by the way, gets its name from the juniper berries that give it a characteristic flavor.
Gin, the beverage, is a lot older than the cotton gin. It was invented in the middle of the 17th century by a professor of medicine at the University of Leyden named Dr. Franciscus de la Boe, who was also known as Dr. Sylvius. The good doctor's recipe was for a cheap, effective diuretic, and it was used to treat kidney problems. Juniper is called genever in Dutch, the language of its inventor, but the English word gin might, like many other words in English, have come from the French cognate genièvre.
Gin's medicinal value as a diuretic soon lost importance to its power to intoxicate. British troops fighting in Holland during the Thirty Years War liked the stuff, nicknamed it "Dutch courage" and brought it home, where it became quite popular. Soon, cheap gin was taking such a toll on the working classes (who were sometimes paid in booze) that the Industrial Revolution was in danger of failure. In 1736, untaxed gin became the subject of a steep levy imposed by the Gin Act. The English rioted! These riots were illegal under the Riot Act of 1715, but that law was not so easy to enforce. It could only be brought into play if it were actually read aloud to the rampaging crowd by a magistrate, something that was not always as healthy for an official of the courts as, for instance, downing a shot of gin. In the long run, however, the Gin Act was enforced and untaxed beer became the drink of the British working classes, and a few other people, too. The beer culture seemed a lot kinder and gentler than the gin culture it displaced, at least according to the artist William Hogarth. His 1751 illustrations quite harshly depict the evils of gin and assert the salubrious effects of beer.
In order to overcome the economic barriers imposed by taxation, gin producers turned to technology, although it took them a while to sober up enough to get anywhere. Eventually, in 1831, the English developed what is today the most highly regarded variant on the gin theme, which is still classified as London distilled gin. Although anyone can make this kind of gin these days, the English still dominate the market.
By 1831, Eli Whitney had been dead for six years, while King Cotton was alive and well, and was providing the raw material that was spun into thread and woven into cloth in, among other places, industrialized Britain. America's cotton exports to beer-guzzling, gin-sipping England brought in the money that was used to buy slaves. America exported some long staple cotton, which only grows along seashores and has seeds that are easy to remove, but mainly sold green cotton, which is easier to grow in more places but has sticky seeds that can't easily be removed from the rest of the boll. The cotton gin picked out these seeds and made green cotton, cheaply harvested by slaves, a viable crop. No other crop depended on slave labor like cotton did, and none was nearly as huge, which is why historians say that the cotton gin, by enabling cotton to become America's biggest export, was instrumental in the rise of American slavery.
Like the English gin distillers, computer companies can dominate their markets with superior technology, but that takes a lot more work than merely developing proprietary interfaces that restrict what can be plugged into a blade server. Even when such companies fail, their ideas can live on. The standard parallel interface found on most workstations is the Centronics interface. Centronics was once the leading manufacturer of small printers; for years it held the starring role that today is played by HP in the blade server market. Centronics, like so many computer companies, and Nelson Rockefeller, too, came and went. The latest version of Microsoft Word still uses keyboard shortcuts promulgated by WordStar. WordStar was never as big as Centronics or, for that matter, Rockefeller. It just went. It was a half Nelson.
The idea that eventually made Eli Whitney a financial success, and which really shows his importance in the development of technology and industry, was the concept of interchangeable parts. Until well into the 18th century, most manufacturing was done by craftsmen, even if it was done by groups of them working in factories. In each trade, every tradesman learned all the skills and processes of a particular craft. Even when a craftsman made the same thing again and again, each item was unique. Then a Scotsman name Adam Smith noticed that some industrial processes could be broken into steps. Moreover, he pointed out, using a pin factory as an example, that when workmen, even those lacking the skills of craftsmen, divided labor and concentrated on just one or two steps, the total production of a team was vastly greater than the output of the same number of craftsmen, each performing the whole manufacturing task. Smith's book, The Wealth of Nations: An Inquiry into the Nature and Causes, published in 1776, explained it all.
Smith's ideas went over well in Great Britain, which was at the vanguard of the Industrial Revolution. They also either caught on or were contemporaneously discovered in the colonies that had just broken free of English rule. Eli Whitney had a factory that made pins and nails using the methods Smith advocated. According to some pretty good sources, Whitney's factory was the only one of its kind in what would become, when the Constitution was ratified, the United States of America.
After the cotton gin debacle, Whitney returned to his roots in Westboro, Massachusetts, a town that would much later be the home of Data General and its successor EMC (which has its headquarters a couple of miles away in the reservoir village of Hopkinton). In 1800, Whitney managed to get a contract to manufacture 10,000 muskets for the armed forces of his young country using processes that were, at the time, unique. Whitney's guns were to be made like Adam Smith's pins, by workers of modest skills who each made one part. The parts were to be identical, allowing, for the first time, a working musket to be assembled from parts selected at random from the batches produced by specialists.
There was a catch: The workers could not do their jobs to Whitney's standards without special machines. Whitney had to invent all the machines, including the milling machine, descendants of which are still in use today, so his employees could make parts to the required close tolerances.
It took Whitney 10 years to fulfill the musket order he initially promised to deliver in two years. But by the time he got the first order completed, his factory, in Springfield, was the best around. The plant took on and fulfilled an order for 15,000 muskets. This time Whitney got the job done in two years.
The division of labor, combined with properly arranged marriages of skills and machinery, was a revolution within the Industrial Revolution, and a very successful one indeed.
These ideas, now roughly 200 years old, are implemented in the factories that make computer hardware and, to some extent, in the organizations that make and test large software packages. But the computer industry apparently does not want to help its customers to do the very same old thing with computers. And the users have not caught on, although an exceptional group can be found in the merciless ISP business, where rack-mounted server farms quickly attained a dominant position and where blade servers have a good chance of catching on, too.
The last time corporate users had a rational and organized scheme of information processing was when they used systems of unit record equipment that resembled in many ways the system of individual machines used to make things like Eli Whitney's muskets. Then along came the computer, and the central processing system became the sole emphasis of computing and computer operations.
Companies using computers took to glass house computing the way the English took to gin, and wealthy companies have stayed with the old way much as the rich Brits could afford gin, even after the working classes switched to beer.
One of these days, the big users will realize they can't afford outmoded computing systems any more than anyone else. The realization might arrive slowly, but it is more likely to be the result of some cataclysm, just as the Civil War, not any more gentle social or political process, put an end to slavery in America. Chances are, nothing will change as long as big business is making big money.
That former emphasis on big central systems is not only no longer necessary but also at odds with computing done via networks. Today, the technology that provides the best value and the greatest flexibility is found in well thought out rack-mounted systems. The next step could well be blade servers, but not if the blades must plug into proprietary cabinets and not if the blades are just ordinary old computers in new cases.
Proprietary blade servers would not only be unattractive in themselves, they would also suffer from two kinds of competition. Rack-mounted systems that allow customers to mix components from several vendors provide more flexibility and, because competition is quite strong at the component level, very good value, too. The cabling and management headaches common to rack-mounted systems are a detriment, but computer services companies are becoming very good at running rack-mounted installations. Now that the dot-com excesses have been wrung out of the server farm business, hungry heirs to the Exodus empire are knocking on corporate doors. They are bound to get better at this; they have pretty much run out of ISPs and pure Internet outfits.
Another contender is the large partitioned system. This concept began as a defensive action in the mainframe market, and to a very impressive extent it has worked. The flow of work from mainframes to satellite servers seems to have slowed as the partitioning capabilities of mainframes have improved. Similarly powerful technology is now springing up in the OS/400 and Unix worlds, and partitioning is also coming to the Wintel and Lintel worlds. Partitioning could have a similar impact in these markets as it does on the mainframe. Nevertheless, partitioning is limited in scope. It only applies to central processor power. While a software-based firewall can run in a partition, a dedicated machine feels like a safer place for a vital system to do its nasty security work. No partition can replace a router. And no central system can do the kind of caching job a network appliance can do without imposing large penalties on the performance of every other application in the box.
Ultimately, blade servers will have the edge. And when they are properly honed they will be made of well-matched hardware and software components. All the components will work cooperatively with many other kinds of blades developed by a diverse society of vendors. That's when blades will be a compelling choice.
The evolution from rack-mounted systems to blade servers is going to happen very quickly. The products announced already and the handful that will come out by the end of this year will all fail in the market if they give customers less choice than rack-mounted systems. The vendors that are in this segment will be out a lot of money very quickly at a time when none of them can afford costly mistakes. They will have to accede to the inevitable.
The race for real leadership and tolerance of foreign blades in a rack will most likely be led by Dell or HP, although IBM could still be an intelligent contender. Sun will either be a joiner or a suicide; it's a bit soon to say which. NEC will get a nice lead at home in Japan, but it has a strong tradition of failing to sell its products, even the best ones, in overseas markets.
The companies with an edge in manufacturing blades are not yet visible in the end user market, at least not in the server end of the business. This is because the companies that are by far the best at making compact processors that use low power and survive in brutal environments are the Taiwanese outfits that make laptops for just about everyone. They make these laptops in their local plants and also in factories that are in China.
The disadvantage these companies have stems from the regional disregard for intellectual property rights. No sane software company is going to get into bed with a hardware manufacturer based in a region rife with pirates.
If the governments of China and other East Asian industrial powers cannot provide credible assurances that the blade business will be properly and effectively governed by law, the computer industry will have to build blades elsewhere, in places where intellectual assets are secure. For the blade business to succeed, it will take tough lawyers as much as skilled manufacturers.
That's another 200-year-old lesson from Eli Whitney.
— Hesh Wiener May 2002