Converged and Unified Storage: The Difference

Modern unified storage relies on a dual-controller storage architecture, and such architectures are sized according to the number of IOPS they’re expected to handle. If spare overhead is bought for future growth, then the buyer is investing a lot of extra (and currently unused) processing power. If the system is optimized for current usage levels, then scaling necessitates downtime while moving into a bigger box or coping with sprawl as more unified storage boxes are added.

In contrast, converged storage is based on three tenets: standardized platforms, federated scale-out software, and converged management. As discussed in our prior article on converged storage design, the blade architecture typically used in converged solutions allows for storage to be only one element in a compute/storage/networking stack. Because the solution is designed from the ground up with this integration and holistic management in mind, there is no cauliflower on the platter that’s been added as an afterthought. Everything is coordinated and optimized into a single unit and managed through a single UI. It not just about disks being in a common pool. All resources are in a common pool. When users need more file serving or block storage, they can simply add more resources as needed without increasing complexity or management. Unified storage hasn’t been able to follow suit.

“Enterprise Strategy Group found that those customers who are deploying scale-out architectures are able to deploy storage about 41% faster because of these architectural differences,” says HP’s Parks. “It just non-disruptively plugs in and you grow vs. those stair step upgrades.

Scale-out storage emerged from the NAS world and is designed such that storage throughput and capacity can expand independently via nodes that run in parallel. As such, scale-out storage is an ideal fit for converged architecture, in part because these added nodes remain internal within the stack. Hence no additional floor space gets consumed and power consumption (and cooling) increases only incrementally rather than doubling with the addition of a new box.

Converged management can yield even greater benefits. Whereas unified storage typically takes days to weeks to get a new application running, converged storage backed by a platform such as HP CloudSystem can leverage automated templates to provision new volumes, plumb them into the network, link that into the storage or server image, and that’s it. Deployments can happen literally in minutes.

Perhaps the key difference between unified and converged storage might be stated this way: Unified storage aims to simplify storage management while converged storage aims to erase it. Ideally, there should be no dedicated storage stack to handle and upgrade. This is sweet news to IT people looking to eliminate as many technology hand-offs and touch points as possible.

“Look at the partnerships that big unified vendors are developing with companies like VMware and Cisco,” adds Parks. “I think they recognize that the industry and customers are tired of dealing with the complexity of technology silos. But marriages tend not to last in business. We think people are going to want simplicity here, too—one vendor who can do everything in-house.”

Less management ultimately means more resources for tackling more profitable tasks, plus it makes scaling even easier—and scaling is something you can expect to see happening to the converged storage space itself. HP made headlines recently with its Project Moonshot and Redstone Development Platform effort, targeting the goal of planting over 2,000 ARM processors in a single rack.

In a silo-based, unified storage architecture, such leaps in compute capability have little impact on storage, but in a converged world the impact can be massive. As much as today’s converged storage represents an evolution of the unified model, even greater capabilities are over the horizon.