What’s your strategy for your Thanksgiving get-together? Happy conversation around a delicious meal doesn’t just happen. It requires intensive planning and preparation to make sure everyone gets what they need (or want, like a third helping of Mom’s special pie).
In some respects, managing a large data storage infrastructure is a lot like making that dinner happen. You have demanding, even quirky, users who insist on doing dishes when there’s a bowl game on, and who have their own preferences about when to eat and where to sit.
Tiered storage is not quite as problematic. Storage professionals can choose storage performance and cost to match data value. But cost savings come with demands on staff time to classify and move data from tier to tier. Virtualization, another resource-saving strategy, brings its own challenges to managing and maintaining the links between virtual machines and data. Mix virtualization with a tiered storage environment, and even seasoned storage managers reach for the antacid.
It’s not surprising that some data storage technology vendors are promising the benefits of tiered storage without the tiers. That’s oversimplifying things, but it’s true that companies can look beyond traditional tiered storage architecture for solutions to some of the technical and strategic challenges that arise, including data availability, data migration and definition of tiers. Using the Thanksgiving analogy, it’s possible for everyone (applications and users) to sit where they want, converse (with storage) without conflict, and get what they want (data), when they want it.
Expect the Unexpected with Your Data Storage Infrastructure
When there’s an extra guest at dinner, you can squeeze in an extra table setting. When the host breaks his arm in a touch football match, dinner has to wait or someone else has to carve the turkey.
There’s a similar problem for data storage infrastructure. Next-generation applications provide huge volumes of information and services, from video to music, but unexpected spikes and even planned peaks in demand can really drag down data availability. Without the means to handle these events when they happen, companies put user experience and employee productivity at risk, and may lose revenue.
To compensate, many companies over-provision nearline storage for files that are likely to generate the greatest read-only demand. This can get tricky for, say, a news Web site trying to serve live event coverage such as Michael Phelps’ Olympic medal junket --and supply archive footage of his past victories that fans want to watch at the same time. If demand isn’t as high as expected, there may be leftovers in the form of unused (expensive) storage capacity.
There are alternatives to over-provisioning and manual migration. Caching appliances are one option; another is to check whether your storage partner offers tools for rules-based automatic migration in response to shifts in demand. With some software intelligence added to the caching process, it’s not hard to see how availability can be improved and storage media such as Fibre Channel can be more fully utilized.
Break Free from the Data Storage Hierarchy
Once initial demand peaks, a file’s availability usually grows less important over time. Information that needs to be accessed frequently or quickly for business-critical tasks -- from online transaction processing to serving up much-anticipated movie trailers -- gets assigned to high-performance storage with a corresponding price tag. Once demand for the clip trails off, the file gets relegated to a less-expensive format with a corresponding reduction in performance.
It’s logical for data to move in a linear fashion down the ranks of storage media, but just as families defy the nuclear stereotype, a file’s lifecycle may break with convention. Storage infrastructure is growing more diverse, requiring complex, non-linear relationships. Consequently, storage professionals may choose to maintain dedicated storage silos for their applications, or park data in an expensive format to avoid the hassle of moving it down and up tiers. Those silos can consume more real estate and use more energy to move data—or dishes -- where needed.
It’s time for storage that frees users from the limitations of linear convention. Consider a leading provider of geospatial solutions and mapping services. Instead of the typical storage path for data, the company finds it more cost-effective to transfer raw aerial survey data to less-expensive serial advanced technology attachment (SATA) disk archives. Then that data is pulled “up” to a Fibre Channel storage tier for technicians to correct, a process that requires simultaneous, out-of-sequence data access.
Despite the fact that its data travels up the tiers against convention, the firm has experienced a 15-fold leap in productivity. Imagine the potential for productivity with tools that automate data migration between pre-assigned pairs of tiers, based on application requirements and user-defined management policies. When data can be moved “down” the chain of storage formats and back “up” automatically, companies can design infrastructure that reflects data’s true lifecycle, for better performance, scalability and access.
De-Duplication Technology
Set aside conventional assumptions about top-down storage hierarchy, and new possibilities emerge for what constitutes a storage tier. In theory, there’s no need to limit storage tiers to task-specific devices. In practice, these devices don’t naturally “talk” to one another. For example, de-duplication technology requires changes to workflow, new protection methodologies and other management tasks that may make prompt storage managers to leave devices as isolated components of infrastructure. If you’re not careful, those devices can start to look like your cousin’s new girlfriend -- welcome, but an awkward last-minute addition without much connection to the rest of the group.
Tools that can cross-link de-dupe and archive devices integrate them into storage infrastructure as additional tiers. Something like automatic de-duplication of live file data onto storage or to solid state cache can boost the efficiency and utilization of other storage assets, and smooth performance through spikes in demand. Meanwhile, the data on those archive and de-dupe devices benefits from snapshot, migration, audit and other important functions provided by the rest of the storage infrastructure.
Everything in Moderation when It Comes to Tiered Storage
The point of taking advantage of advances in tiered storage is not simply overindulgence. New technology based on fresh thinking is leading to simplified planning and delivery, better performance, scalability and access, better resource utilization, and even lower cost of ownership. These new developments promise to take the guesswork out of planning your storage infrastructure get-together, so that availability and accessibility levels are just right, without loads of leftover capacity. That’s something that storage managers can be thankful for.
Jon Affeld is senior director of product marketing and business development, BlueArc Corp.
