Developing data hierarchies is the foundation of ILM. The idea is that less critical data can be moved to cheaper, less available storage. Less sensitive data can have easier access requirements.
The first step, though, is to do a data inventory and then determine value. Enterprise search tools like those from Google, FAST, or Endeca will dig up whatever data you’re looking for, while tools from Kazeon, StoredIQ, and EMC with its Infoscape product will help you figure out how to classify that data.
That’s the easy part. The hard part is determining value.
“This is the biggest ILM abyss,” said Brian Babineau, analyst for The Enterprise Strategy Group. Departments will bicker and personal agendas may trump true value. There are few types of data that everyone will value similarly. “Confidential customer records and things like credit card numbers will have a high value,” Babineau said. Beyond that, it’s up for grabs.
“If it’s just a file server where people can keep old Powerpoints, you don’t need expensive infrastructure,” Babineau said. Objectively, this is true, but Babineau noted the trouble with this logic is someone in marketing or some C-level executive may disagree with archiving these files. All of their data is especially important, they believe.
One shortcut that avoids intra-organizational feuds is to rate the data by how frequently and how recently it has been accessed. “The data resides on the optimal tier of storage based on usage and cost requirements,” said Bruce Kornfeld, VP of marketing for Compellent, an enterprise storage vendor. “You want the lowest tier possible without impacting performance.”
This makes good sense except in those rare instances when infrequently accessed data is of the utmost importance. National defense and nuclear launch codes are good examples, said Dan Cobb, CTO of EMC’s Information Management Software Group.
“This is an extreme example,” he said, “but if I’m in national security, I want the launch codes on the highest level possibly, and I hope that I never actually access them.” A less extreme example is your customer’s personal information. You can move it to less available storage, but never less secure storage.
Kornfeld noted that a frequency- and time-based system can adapt quickly to changing values, dynamically moving newly important data to higher tiers. Additionally, most systems have overrides, allowing you to lock certain information to specific types of storage.
Assuming your organization can agree on some data policies, what’s next? Cobb reiterated the need for articulating concrete business goals. “The most important thing to do is to meet the needs of the business, in terms of cost, service level, IT efficiency, risk management, security, and recoverability.”
If ILM makes data more available and secure, while lowering your costs, then its value is obvious. However, Kornfeld cautioned against ignoring the implementation pains that some solutions necessitate.
“Often, ILM doesn’t get implemented due to labor issues,” he said. Many solutions in the industry require the manual movement of storage among tiers, which can push ROI well off into the future and place too much of a burden on IT.
Cobb warned against looking at ILM too narrowly. “Parts of the industry believe that ILM equals tiered storage. That’s not the most valuable thing to business. A holistic approach to policy is where the value is,” he said. “You need to ask ILM vendors how they store, protect, and optimize information. All of these are important and none can be taken in isolation.”