Flash or SDS? Finding the Right Path for your Infrastructure
Software defined storage has tremendous potential. The ability to decouple storage from the applications and systems that use those resources can yield huge benefits for businesses, allowing them to quickly scale capacity when it’s needed and maximise the utilization of infrastructure.
That’s the theory, at least. But while SDS has taken off quickly in test and dev environments, its adoption for production use hasn’t been as swift. According to Gartner, last year just 14% of enterprises were using SDS broadly in production.
As it turns out, software defined storage is hard to do right. Deploying any new stack can be a long, complex journey, and decoupling software from hardware introduces complexities that are hard to manage, particularly in a DIY deployment. It’s easy to get tripped up by driver and software compatibility issues, and the architecture needs to be just right to guarantee performance. Integrating flash in a way that maximises its benefits is even harder.
In short, customers are having a hard time identifying the unique pain point that makes overcoming these challenges worthwhile.
The “Apple” Infrastructure Model
This difficulty leaves a lot of what software defined storage has to offer on the table. But there’s another route that allows customers to leverage the benefits of both virtualized storage and flash without the heavy investment in man hours required to build out a DIY SDS architecture.
A converged appliance that integrates software and hardware in a single, pre-configured system removes a lot of the uncertainty from virtualizing storage, while still leveraging the plugins and APIs that automate many of the traditionally cumbersome provisioning tasks - one of the key tenets of SDS.
Tightly coupling hardware and software yields greater efficiencies, such as being able to aggregate storage resources into a pool that can be scaled and allocated virtually to applications (think of of the efficiencies with Apple’s hardware-software coupling approach to their consumer electronics). And an appliance can be designed and configured to specifically support the evolving needs of the business
Moreover, the best appliances offer software tools for intelligent management of the storage architecture, including cloud-based analytics that provide visibility across the entire IT stack and at the virtual-machine level.
The Hype of Hyper-converged Systems
There’s another path to storage virtualization in the form of a hyper-converged system, where storage, network and compute are tightly woven together. These systems are easy to set-up, which helps to explain why some analysts see hyper-converged as the fastest-growing sector of the software-defined storage market.
But anyone who’s gone this route can tell you that demand for compute and storage rarely scale linearly. If your compute needs outstrip the capacity of the system, you need to buy a new box and existing storage is left stranded. Hyper-converged systems also tend to rely on proprietary protocols and interfaces, which makes them hard to integrate with existing infrastructure.
On top of that, servers refresh every two to three years, storage refreshes every three to five years and networking gear refreshes every five to seven years. With hyper-converged systems, it can be possible that technologies can quickly become out of sync with best of breed products. This can be a problem given that IT is now leveraged as a competitive differentiator.
Conversely, a good storage appliance can integrate with your existing infrastructure. So while the ease-of-use gains from a hyper-converged solution can be significant, the trade-offs mean it won’t be a good fit for everyone.
You Can’t Do It Alone
As storage needs evolve further, particularly with the growth of real-time analytics for marketing and other needs, flash becomes a greater necessity. But most customers simply can’t expect to build flash arrays in house and expect the type of efficiency achieved by specialist vendors. Managing the various storage tiers and moving data around based on performance needs is another task highly suited to converged systems.
Before you set out on any new storage project, get a clear sense of the pain points in your organization now. Look at the applications coming over the horizon and determine which might be suitable candidates for virtualized storage. Then determine which solution is right for you, based on what you’re trying to achieve.
(About the author: Rajesh Nair is co-founder and chief technology officer of Tegile Systems, where he leads technology and product development of Tegile's flash storage portfolio. He brings over 19 years of engineering and product development experience.)