Why It’s Important to Know All Your Storage Options

by Chris Preimesberger

Storage architecture options with and without NAND flash are available in an array of options. Make sure you’re educated on the pros and cons of your architectural options before making a choice on what is best for your application environment.

Storage architectures built in the pre-virtualization era were not designed for the random I/O profiles associated with virtualized workloads. Bolting on NAND flash as a cache is not an effective way to deliver more performance.

The monolithic storage arrays (storage-area networks, network-attached storage) were designed to meet the needs of environments before the requirements for virtualization, scale and high performance of today’s enterprises emerged. If you have an older system, some new hardware is probably in the cards at some point.

Host-side flash has become a popular choice to quickly remedy local host performance issues, but this limits the performance fix to applications located on the single host and places a heavy burden on the local CPU to drive both applications and flash performance. This is fine for a single-host solution but not for shared storage as your enterprise scales.

All-flash arrays deliver high performance to your applications, but they also can carry a hefty dollar-per-gigabyte cost. Plus, they have limits in scaling capacity with performance due to their architecture. This can work well for a small batch of identified applications that require performance all of the time but not for the bulk of applications that need that performance periodically.

Software-defined storage has three key characteristics: 1) software stack is decoupled from hardware for flexibility in using various commodity hardware options; 2) use of object storage to enable Web-scale performance for petabytes of data, structured and unstructured; and 3) business/application-level policy enforcement of performance and durability SLAs. Many existing and emerging storage options fall into this category, so be sure to dig into the details and be ready for some steep pricing.

A storage system needs to be modeled in parallel to the type of business it is serving. Is your business data usage static or dynamic? If it’s static, perhaps virtualized servers and cloud services aren’t immediately necessary. If it’s dynamic, will the storage architecture support simple capacity scaling?

An important question to ask a potential storage vendor is this: Will the storage architecture be able to automate application performance management? This goes for applications used in-house and in mobile devices. You may need to look independently for help in managing corporate data in mobile devices, including smartphones, tablets and laptops.

Storage historically has always been about clearing and avoiding bottlenecks, no matter what size the system. A key question to ask a vendor is this: Will the architecture scale application performance as a new capacity is added without causing or running into bottlenecks? Will you have to manually load balance the system as you scale?

Aside from capacity growth, do you anticipate performance need growing due to new projects, seasonal spikes and other factors? It’s a good idea to look at the enterprise’s storage capacity history for the last five years and make some projections based on that plus company sales metrics that can project growth of the enterprise itself.