CA Exec Offers Vision Of On-Demand Storage

Turner said the goal of storage management is to make storage as simple as possible, however, enterprise storage is still very complex in terms of underlying hardware, technology and management. "Customers feel the ROI of storage is low," he said.

Business continuity risk is still high, and storage quality of service is seldom delivered, Turner said. Storage security is almost non-existent, and storage utilization is on the average less than 30 percent, he said, referring to a CA customer survey. "In fact, more data which has not been used in the past 12 months is stored than data which has been used more recently," he said.

The problem is that storage solutions have failed to focus on the business needs of storage, Turner said. "Everybody, every solution provider, has been responding to the growth in the demand of storage. ... We don't have the set of tools to manage this environment," he said.

For that reason, CA is moving to develop an on-demand storage strategy under which IT resources are made available as needed from a pool of platform-independent and vendor-independent storage resources, with such resources being centrally managed, Turner said.

Sponsored post

On-demand storage will allow capacity to be allocated from a shared storage pool in a platform-independent manner and will allow for a quality-of-service guarantee, Turner said. "We have to be able to provision storage as needed," he said. "But we must also provide security, charge-back policies, and so on, automatically."

Anthony Nguyen, director of network services at Pomona College, Claremont, Calif., said he has been keeping an eye open for software to allow automatic provisioning of software but has yet to find any suitable tools.

The concept of virtualizing storage into a storage pools is still surrounded by hype, Nguyen said. "At this conference, I'm getting mixed signals," he said. "Some people say virtualization is still a lot of hype, while some say we are close to it becoming a reality. At least it's better than last year when there was definitely a lot of hype."

Pomona College has pooled its storage into a SAN, but it handles that implementation in much the same way it would handle direct-attached storage, Nguyen said. "It's pooled, but there's no easy way to virtualize it or grow it easily or make it available to other servers," he said. "There's no way to partition the storage easily. Adding in a new server or array requires a lot of manual work."

The main enabler for the centralization of storage management is the ability to abstract the behavior of the data from the data itself, Turner said. "Today, how data is protected depends on policies set on a backup engine," he said. "In the future, it will be based on where the data exists and on the value of the data. This will allow the data to be protected with the most effective mechanism."

In order to move forward with better automated storage management policies, data will need to be described as storage objects based on the attributes of that data, Turner said. For instance, the storage network could be directed not to put information related to supply chain applications or online transaction applications on the same spindle or be directed to store mission-critical data on its own spindle, protected with snap copies, he said.

To move forward, storage must be more automated, Turner said. Today, the provisioning of storage takes 15 different steps; this should be reduced to one, he said. Other requirements for centralized management include a single point of management and the implementation of different security procedures based on the requirements of the data, he said.

One critical step enterprises must take is to determine the business value of their data based on how critical it is to their operations, how accessible it needs to be, and the cost associated with recreating the data, Turner said.

"You have to start to understand what data you have in your environment and how it changes over time," he said. "I'm always amazed how many enterprises don't understand what applications work with what data. As a result, they treat all data the same. That's not the best way. You must understand the value of the data to manage it properly."