SANs may be more efficient, but can be much more complex to manage than direct-attached storage, says Pat Martin, chairman and CEO of StorageTek. He argues that combined with the fast growth in storage volumes, this means businesses must find new ways of organising their data before it spins out of control.
His choice is information lifecycle management (ILM), an approach which builds upon hierarchical storage management (HSM) techniques that were developed in the mainframe world some 20 years ago. HSM shifted little-used data off onto cheaper storage, typically tape. ILM extends this with multiple classes of storage, and adds mechanisms such as data profiling to determine what should be stored where.
"The problem is we tend to treat all information the same," Martin says. "A new perspective is required, where we look at the data we manage and recognise that we don't have to treat all databases the same.
"We've known for 20 years that the probability of information being accessed is very high if it's just been created, but less than 20 percent after 30 days, and very low after 90 days.
"The current tactics are not sustainable. The cost of managing storage is increasing faster than the management complexity is simplifying, with the result that many customers are saying it's already not possible to effectively manage their storage environment."
Adding to this is the problem of declining budgets, which he identifies as the number one priority for CIOs (chief information officers) this year. "The low-water mark was the CIO of a large US financial organisation who was ordered to reduce his budget by 50 percent," he says.
And of course, while falling storage prices allow you to manage rising data volumes and decreasing budgets, they also mean you end up with a whole lot more storage to manage, which ties in with his second priority for CIOs, namely the need to support ever-increasing service and application demands.
"Right now, technology officers are under siege," Martin adds. "They don't have the same political power as the people running the applications. Application owners just say 'disk is cheap' - that to me is the biggest barrier.
"When we talk of cost per MB, we're not recognising the cost of software, services or management. Today it costs 5.1 Euro to manage every 1 Euro of storage we buy, and that'll be 5.3 Euro by 2006."
And there is another hidden cost - electrical power: "I know one California data centre which cost as much for the electricity to run the disk subsystems as they cost to own," he says.
Martin has turned StorageTek around since he joined the company three years ago. He did it in part by telling demoralised staff that no, tape wasn't dying, and there was nothing to be ashamed of in selling it, and by recognising that StorageTek should build on its existing technologies and its strength in the market, rather than abandoning them.
Now he has turned his attention to the next stage, which is where ILM comes in. The idea is to bring together disk arrays, tape drives, management software and new layers of the storage sandwich such as cheap ATA disk systems, to ensure that data is stored and managed according to its importance and the likelihood of it being accessed any time soon.
"HSM has been well accepted - MVS and everyone else in mainframes does it. But open systems evolved as point solutions, until server farms were created and handed over to professional data centre managers, who were appalled by what they found," Martin says.
"You need a hierarchy of storage, and you need to ensure your data protection matches the criticality of the data. ILM doesn't solve the problem and it won't give you a flat budget, but it is probably the biggest thing you can do right now. Everybody's talking about it - HP has it on slides, EMC and Hitachi are working on it."
However, it could be a double-edged sword for those vendors, as it lets data managers postpone buying more high-end disk: "I think what we're doing is disruptive to their market," he notes. "But we only have a 2 percent market share in enterprise disk so if we disrupt that, it doesn't hurt us a lot."
Future extensions to ILM, such as a hardware-independent storage format, could address other issues such as data longevity. "We all know that the devices you use today will not be around in 20 years," Martin says, "so what is your policy for media management that will let you read today's data back in 2025?"
Looking ahead, Martin sees more and more of StorageTek's effort going into software, such as its recently announced EchoView semi-synchronous data replicator. "It won't stop us developing next-generation hardware, but an awful lot more of our intellectual energy is being focused on software," he says.
And he adds, "We will see more consolidation [in storage hardware]. For example, following IBM's sale of its disk manufacturing to Hitachi, it would not surprise me to see more commonality between IBM's Shark and the Hitachi products.
"When the dust settles, 6 percent annual market growth may be sustainable - and that's not to be sneezed at."