Like Facebook for employee use and iPhones in business, data storage policy is a topic that can be a political hot potato within corporate walls.
Implementing policy-based storage management can be so contentious, in fact, that at many companies, it quickly turns into a nonstarter (see survey results in sidebar, below). Rather than haggling with the business about where data gets stored and for how long, IT simply places all data in a single tier -- that which is most satisfactory for the most critical applications. The result is the stereotypical high-cost, overprovisioned storage-area network.
Andrew Reichman, an analyst at Forrester Research Inc., calls this the "highest common denominator" approach to policy-based storage management.
Storage Reduction by the Numbers
Reducing our organization's stored data volume over the next two years is:
* A high priority for our IT shop: 47 per cent
* A low priority for our IT shop: 53 per cent
To what extent is your organization moving toward reducing its stored-data volume?
* We've implemented several technologies designed to reduce stored data in the past 12 months: 25 per cent
* We are just beginning to look at technologies designed to reduce stored data: 48 per cent
* We are not considering technologies designed to reduce stored data: 27 per cent
Source: Exclusive Computerworld survey of 97 IT professionals, August 2010
"Companies really struggle with how to do policy-based storage management right, so in some cases they do nothing. They just don't tier at all," he says.
"Everybody's application is performance-sensitive and important, or you wouldn't be deploying or supporting it. What's needed is the relativity -- how much performance capability is truly needed for this important application to work effectively?" Reichman adds.
"The challenge isn't with the cost of the hardware or the software or in the tools for moving the data," agrees Greg Schulz, an analyst at The Server and StorageIO Group. "It's in getting somebody to define and sign off on what to move, when to move it, how long to keep it, how many copies to keep, when to delete it, and how to handle the deletion. This takes someone who has a business-value understanding of the applications and who can interface with technologists to bring all this together, and who also can get approval from the business that this is what can be done, this is what needs to be done, and also, by the way, will indemnify it."
And when dealing with data files, storage policies are only as good as the level of user compliance, adds Kerry Sylvester, IT director at WaterFurnace International Inc., a manufacturer of heat pumps.
At the Fort Wayne, Ind., company, Sylvester uses FalconStor Software's virtualization technology to move data across its storage tiers, but users don't always store the data as directed, he says. "Some people like to hang on to stuff, and habits are hard to break. To make policy-based storage work, you need your users storing their documents to the right places," says Sylvester.
If politics and culture aren't distracting enough, a lack of knowledge often stands in the way of good policy-based storage management, Reichman adds.
"Performance analytics in storage is pretty binary and often trial-and-error-based," he explains. "So there's often no way to make a confident decision and say, 'This will perform completely effectively on a lower class of performance system.' So a lot of it is guesswork -- and that means there'll be mistakes sometimes, and as a result IT alienates the user community."
But the politics of storage are starting to change with advanced automated data tiering. Available from companies like 3Par (which has agreed to be acquired by Hewlett-Packard), Compellent Technologies and EMC, this technology monitors how data is being used at the block level, determines which data should be on which type of storage, and then moves it there.
Reducing the Complexity
With automated tiering, enterprises still set policies for applications or groups of applications. They can stipulate, for example, that the Oracle production database should run on Tier 1 solid-state and Tier 2 Serial ATA drives. But then they can step back and let the system do its thing, rather than manually moving the data between those tiers as needed.
"Before we went to Compellent-based policy management in the SAN, policy-based storage meant a laundry list of highly technical configurations that needed monitoring daily, if not hourly, and constant planning," says Jack Rahner, assistant vice president of IT operations at AlphaStaff Group Inc., a Fort Lauderdale, Fla.-based employer of record for thousands of clients' professional workers around the globe.
5 Tips for Getting It Right
Experts suggest the following best practices for creating policies:
Conduct a performance study, then use the results to populate storage groups based on the level of performance required for an application, says Rick Hoffman, enterprise storage management specialist for the state of Michigan. Further, preserve your fastest storage only for your highest-performing data, Hoffman adds. "Some applications don't ever need to run on solid-state drives, for example."
Consider automated data tiering from companies such as 3Par, Compellent and EMC, says Mark Peters, a senior analyst at Enterprise Strategy Group. "I don't see much downside to this. It's a more efficient way to do policy-based storage management."
If policy-based storage is a trouble spot, consider other issues to help make the cost justification for automated tiering. "Is performance an issue? Price? Capacity? Support for 10 Gigabit Ethernet?" asks Jack Rahner, assistant vice president of IT operations at AlphaStaff.
Don't attempt to boil the ocean, advises Greg Schulz, founder of The Server and StorageIO Group IT consultancy. "Take a bite-size chunk -- a division, a particular application or function -- and build a policy-based storage management success story for that one operation. Then do it again, and again."
Keep your overall directive in mind. "When I think of policy-based storage, I think of tiering, and the purpose of tiering is to reduce the overall cost of the storage environment by moving data to the most appropriate tier," says Andrew Reichman, a senior analyst at Forrester Research. "If you're not reducing the overall costs by doing tiering, then don't do it."
-- Beth Schultz
"Policy-based storage meant being in touch with the business and knowing what kinds of activities were coming up. It all revolved around tailoring the current storage and the amount of storage we'd be expecting in read/write fashion for each customer," he explains. "It was very complex, and we never seemed to get it quite right, and so what we ended up with was a senior engineer working full time to manage and monitor a card game of storage."
Since deploying Compellent's Fluid Data tiering technology a year ago, the rules have changed and the complexity has disappeared, Rahner says.
At AlphaStaff, which currently has hundreds of terabytes of stored data on its Compellent SAN, the business owners still give IT a heads-up about impending activity, such as open enrollment or tax filing periods, Rahner says. "But that's really more because that's what they've always done," he adds. "So my interaction with the business is still the same, but then we don't have to scramble and whiteboard for three days trying to figure where we're going to put the data."
Of course, you do have to convince the top executives that automated tiering is the way to go. For Rahner, selling the idea to the CFO and CEO was no biggie, but telling the CTO that he wanted to move all the company's production storage to a new vendor -- "now that's a tough sell for anybody," he says.
So in what some may consider overkill, Rahner essentially brought his SAN and ERP data to Compellent's facilities and tested the setup for two months. Tons of metrics later, he was able to make his case easily, he says, noting, "You just can't argue with cheaper, better, faster."
Schultz is a longtime IT editor and writer in Chicago. You can reach her at firstname.lastname@example.org.