Features - Enterprise Data Insights:

IS THE DATA STORAGE FIELD FINALLY LEARNING TO COOPERATE?
By Justin Pope

What do railroads, computer networks and laser printers have in common? They all started out as neat, but unwieldy, ideas -— until they adopted uniform standards and the industries took off.

The same may someday be said of data storage, the unglamorous but critical industry that holds and manages the world's gargantuan and ever-expanding collection of electronic information, everything from bank accounts to airline reservations to the human genome project.

But only if the multibillion-dollar industry's rivals can figure how to do something goes against their nature: cooperate.

It's starting to happen, but it hasn't come easily.

Data storage is a relentless challenge for businesses, which will need 30 times more in 2004 as they did in 1999, the Gartner Group estimates.

Still, the industry is slumping.

So cost-conscious customers are piecing together hardware from different vendors, rather than buying big new systems. The piecemeal approach requires software that "talks" to all the hardware, whether it comes from EMC, IBM, Hewlett-Packard or others.

And until they get more software that works across platforms, many customers aren't buying anything at all.

"Users are afraid to buy anything right now because they're afraid it won't work together," said Steve DuPlessie, senior analyst at Enterprise Storage Group, a research firm. "The whole industry is killing itself with this proprietary fiefdom."

Prisoner analogy

Competition among storage software and hardware makers resembles what economists call the prisoner's dilemma.

In the long run, companies' interests lie in cooperating to make sure their products work together. That's what customers want, and for vendors it's easier than wasting time and money reverse engineering rivals' products to try to force them to work together. Instead, that energy could go to making better, cheaper software.

But in the short-run, the vendors' interests lie in competing. It's hard to break the habit.

The first step has been sharing "application programming interfaces," or APIs, the codes that allow machines and software to communicate with one another. EMC and HP announced the first major API swap a year ago.

They've since expanded the deal, and HP has swapped with IBM and Hitachi Data Systems. EMC hasn't reached deals with those companies, but it has put out its own de facto standard, called WideSky, that works with their products.

Meanwhile, amid the flurry of API swaps, more than 20 vendors used an early version of an industrywide standard, called CIM, to link their machines last month. Cooperation can work.

When CIM is finalized next June, EMC and HP will have compatible products ready to go.

"The movement forward we have seen over the last six months has really removed the need for these API swaps," said Don Langeberg, HP's director of marketing for networked storage solutions.

Difficult to cooperate

Still, the cooperation has been grudging. At a recent industry conference, EMC Chief Executive Joseph Tucci declared it "stupid" some rivals weren't yet sharing their codes, while HP complained EMC wasn't sharing enough.

Last month, the companies traded patent infringement suits, prompting an HP spokesman to say, "It's not cool when a competitor essentially rips you off."

And DuPlessie warned against assuming the CIM standard will produce perfect interoperability.

"We think it's good moving towards standards, but at the end of this day these guys can't agree on lunch," DuPlessie said.

Customers like Montreal-based CGI, North America's fourth largest information technology outsourcing company, hope the vendors put their disagreements aside.

CGI takes over IT tasks for huge companies like banks and insurers, and has a staggering 40 terabytes -— or trillion bytes -— of data stored on its storage area networks. That's about four times what would be needed for the entire Library of Congress. EMC estimates a terabyte equals about 1 million books or 250 digital movies.

CGI uses EMC software to manage a variety of non-EMC platforms but still has some problems when it takes on clients with certain hardware.

With the CIM standard, senior consultant Serge Dansereau estimates, his staff could manage twice as much data for the same cost and effort.

Advantage of standards

The hope for both vendors and customers is that standards will do for data storage what it has done for other industries.

When the railroad industry lined up its gauges in the 1800s, trains could suddenly move seamlessly around the country and the industry exploded. And computer networking mushroomed after adoption of Ethernet and TCP/IP.

But there are always losers when standards emerge, such as companies with proprietary technology that is suddenly worthless when everyone adopts the same rules.

And some worry that if standards are too all-encompassing, then companies that build a better mousetrap will be punished and innovation stifled.

That's why EMC says it supports standards for a base level of functionality, but no more.

"There's a misconception in the market that standards will drive innovation," said Mark Lewis, EMC's chief technology officer. "Standards are more a way to inhibit innovation. ... You want to have some standards at some levels, but if you're going to have value in a product, you can't have standards everywhere."