Evolution of the SAN Market
Thanks to the Internet and the rapid global expansion of computing, humans and their machines will create and store more information in the next three years than in the 300,000 years of history dating to the earliest cave paintings and beyond1. That was what researchers at the School of Information Management and Systems at the University of California at Berkeley forecasted late in 2000, much to the delight of EMC Corporation, the data storage giant that sponsored their work.
EMC was quick to pitch the study to Wall Street, adding it to analysts' projections that spending on data storage products is drawing even with spending on computers themselves and that it will account for 70% of information technology budgets by 2005. EMC also included its own projection that an individual (EMC likes to call him "Tommy" in its advertisements) could easily have a terabyte (the equivalent of 250 million pages of text) of stored personal records, photos, and other data by 2005.
The obvious outcome of such trends, of course, would be mind-boggling growth for data storage products and, more to EMC's point, full-scale storage systems intelligent enough to support the Internet's need for constant access to data. Data needs to live someplace. There's almost no value if it's just put away.
EMC's need to keep Wall Street awed is directly linked to its track record. It became the biggest gainer on the New York Stock Exchange in the 1990s by grabbing leadership of the market from IBM, producing both astonishing profits and sizzling growth. Then, as other technology giants stumbled in 2000, its shares gained another 10% to finish the year at $66.50. From such a pinnacle, convincing the Street the best is yet to come will be no mean feat. But, there are plenty of other companies beating the same drum, from giants like IBM, Compaq Computer, and Sun Microsystems, to fast-growing newcomers like Network Appliance, Brocade Communications, and Veritas Software.
Storage is becoming the heart and soul of all business. What you know about your customers, suppliers, and partners will differentiate you at the end of the day.
Trouble is, even if the vendors and analysts are right about the growth, investors may well have become unrealistically optimistic about how easily it will translate into profits. The publicly traded industry leaders tumbled in 2000, but are still trading at nosebleed levels that leave little room for earnings disappointments.
The impact on data storage of a general slowdown in technology spending is Wall Street's current fear, but storage companies also have to contend with tougher competition. New technology is driving down prices, just as in traditional computer markets. In addition, networking giants like Cisco Systems and discount computer specialists like Dell Computer are moving in, while rising interest from venture capital firms is spawning a steady stream of start-ups scrambling to define niches. They are fighting over a rapidly shifting landscape that International Data Corporation estimates was worth at least $70 billion in 2000, depending on which technologies and services are included.
Storage is getting to be as complex as servers and networks. One sector of the storage market focuses on data used in computations that are stored in caches on microprocessors, on memory chips, or inches away on disks inside a computer. At the other end of the technology spectrum is tape-based storage, which provides a low-cost if somewhat less convenient alternative to disk storage.
But the heart of the action these days revolves around disk-based storage systems outside the computer. The newest hardware building blocks are specialized file servers. Some, not much bigger than a VCR, allow users to add storage capacity directly to Internet networks without buying full-scale server computers. The workhorses for big enterprises, though, are refrigerator-size storage arrays of disks that support one or more mainframes or networks of smaller computers.
Steady advances in the disks and the software that manages them are producing astonishing performance gains. Remember Moore's Law, the longstanding rule that shrinking circuitry allows chip companies like Intel to double the processing power of processors every 18 months? Well, that amazing progression (from the room-size computing monsters of the 1950s to far more powerful fingernail-size chips) pales in comparison with advances in data storage. EMC, for example, says that the volumes of data it will stuff into shoebox-size devices by 2005 would have required covering an area the size of Argentina if 1950s technology were still in use.
For all that, the hottest storage battleground is not storage hardware but software, switches, and other components that meld the storage devices into intelligent networks and keep them online. Brocade Communications' market leadership in Fibre Channel, a specialized protocol for designing such storage networks, drove its shares from an initial public offering price in May 1999 of $2.38, adjusted for splits, to $133.72 in October 2000, though it has since retreated. Veritas's strength in software to manage incompatible storage products from numerous different vendors helped its shares climb from a split-adjusted initial price of 53 cents a share in 1993, to a secondary offering in August 1999 at $22.14 and a peak of $174 in March 2000.
Storage services are also booming as big data users hire consultants, rent outside capacity, or simply turn over the entire problem to technology management experts like IBM Global Services or new specialists like StorageNetworks, a two-year-old start-up based in Waltham, Massachusetts. International Data estimates that the service sector had revenues of more than $24 billion in 2000 and indicates its sales should top $40 billion in 2003.
If there is anyplace where the sometimes conflicting visions of storage's future intersect, it has to be the headquarters of EMC in the Boston suburb of Hopkinton. EMC, like IBM in the past, strives to design equipment that performs best with EMC software, so that customers become locked into it as a vendor. And, like IBM's mainframe business in the 1960s, EMC counts on its reputation for reliability and service support to make it the safe, if premium-price, choice for information managers. But, company officials say, any resemblance ends there and that no one will catch EMC off guard as EMC itself caught IBM in the storage business.
EMC's strategy assumes that information pipelines (bandwidth in the industry's jargon) will become so huge and fast that it will no longer be necessary to store data locally to ensure quick access. Such bandwidth, in EMC's estimate, will allow as much as 90% of data to be centralized in the kind of big businesses that have been EMC's prime customers. From medical files, to movies, to financial records, data consumers would download what they need when they need it, but would not necessarily store it on their own computers.
The best architecture for such data reservoirs is still up for grabs, however. Some data will reside in dedicated, maximum-security systems linked to particular computers. Some will be in cheap file servers (NAS) attached to the Internet. A lot of it is likely to end up in networks of storage devices (SANs) that would be linked to the Internet, computers, and tape storage systems through specialized servers.
How things develop depends on evolving network equipment and software as much as on the storage devices themselves. In areas where the landscape isn't as clear as EMC likes, they're placing multiple bets.
Big bets, too, judging from EMC's vow to invest $2.5 billion over the next two years in research and development, more than 75% of it in software. But what if projections like the Berkeley study prove to be wildly inflated? What if people become smarter about saving only what they really need? The industry's answer is another question: why would they bother? With storage prices headed from about 40 cents a megabyte today to less than a cent in 2005; and the industry moving toward making access to storage as easy as the universal dial tone on the telephone, it's going to take too much energy to throw things away. Besides, the Berkeley figures may well be too conservative, since they exclude any estimates for duplicate storage of information once it is created, one of the fastest-growing segments of the business.