Imagine driving a car without a fuel gauge. You have no idea how much fuel is in the tank and worry that you're about to run out, so you stop at every petrol station you see to buy more. Storage managers in data centers have much the same problem.
They purchase additional storage capacity and then assign portions of it to various purposes -- the database program, the e-mail program and so forth.
Once they've apportioned all the storage capacity, it looks like it's all used up (though they set some aside as a reserve). The storage vendor, of course, is happy to sell them more.
"This is sort of the dirty little secret of the storage industry," said Ray Villeneuve, president and CEO of MonoSphere. After a storage vendor delivers the customer's newest array, "the sales team is hovering, already working on the next US$2 million deal."
Privately held MonoSphere itself offers storage resource management software designed to give customers an accurate picture of not just how much storage they have but how much they are using up.
Many storage administrators endure a laborious process to figure that out, said Dianne McAdam, a storage industry analyst at The Clipper Group. They often use monitoring applications that place software agents on servers to track how much data is sent to storage. But they have to add up the numbers from each of what may be hundreds or thousands of servers, then manually match the results to the amount of storage provisioned for each application.
"It's ugly, it's painful," said McAdam, who once worked in the industry doing that very task. "I was doing queries to each storage appliance and just putting them into a spreadsheet manually."
Said John Kelly, product marketing manager for the Storage Essentials management software product at Hewlett-Packard: "There's a joke in the industry that Microsoft is the leading storage resource management vendor because everyone uses Excel."
Smaller, niche players like MonoSphere and Onaro offer specialized storage resource management software, McAdam said. Similar firms have been acquired by larger vendors, such as the 2005 HP acquisition of AppIQ Inc., to integrate into their own management software.
MonoSphere's Storage Horizon software runs on its own server on the customer's network. Instead of using agents deployed on each server, MonoSphere views the volumes of data generated by the application servers and correlates them with the storage capacity allocated to those applications.
If one portion of storage is being underutilized, it can be reallocated to another area that needs it.
Brent Ozar was surprised at the results after he took inventory using Onaro software of the 80T-bytes of storage he manages at Southern Wine & Spirits, a beverage distributor in Florida.
"We had about a half dozen servers that we had equipped with [storage] space that have never even used it. We were able to free up several hundred gigabytes of space by just going through and starting to ask questions," Ozar said.
On average, only 35 percent to 40 percent of storage capacity is being utilized, usually because various departments overestimate their storage needs, said Steve Norall, senior analyst with Taneja Group.
Overprovisioning helps push storage spending higher. By the end of 2006, the storage industry recorded its 14th consecutive quarter of revenue growth, according to a report from IDC, which also projects worldwide storage spending to grow to US$32.4 billion by 2010, from US$29.1 billion in 2006.
But if enterprises are spending those billions on storage and using only 40 percent of it, they're wasting money that could be better invested elsewhere in their business, Villeneuve said. By more accurately measuring how much storage they are using and reapportioning it within their data centers, they could bring utilization rates up to 80 percent.
If, for example, a company is actually using just 1 petabyte of storage, at 40 percent utilization they end up buying, managing, depreciating, powering and cooling 2.5 petabytes. If they bring utilization up to 80 percent, they would only be buying 1.25 petabytes. If storage costs, say, US$20,000 per terabyte, the capital expense savings would be US$25 million, not to mention the savings on operational costs and depreciation, Villeneuve said.
In general, it's better to do storage planning up front, when the storage is being provisioned, because reprovisioning it afterwards can be expensive and disruptive to the IT system, said Bryan Semple, vice president of marketing for Onaro.
"It takes a lot of work to reclaim and reallocate storage and it's not worth the cost just to reclaim 50 gigs," Semple said. If a company wants to use software like MonoSphere's to continue to monitor storage, it should use it only for the largest portions of storage.
Ultimately, storage overprovisioning is more a behavioral issue than a technology one, he said. Even if overall storage spending is rising, the unit price is declining so it's still considered a bargain. Storage administrators buy more than they need because they figure they won't get fired for having too much storage, but for having too little.
"No database manager wants to get paged in the middle of the night that a database is out of storage space," Semple said.