Most Read This Week
The Future Is Now: Why Flash Storage Will Transform the Data Center
By using software-defined storage, data center architects can design a flexible, efficient and powerful framework
By: Stefan Bernbo
Apr. 4, 2014 12:30 PM
For an example of just how dramatically storage has changed over the past fifteen years, consider your music collection. At one point, you had a collection of cassettes that stored the song files on tape. As the years went on, your hairstyle changed and you bought a CD player that used a spinning disk to store more song data at a higher quality than tape could. Spinning disks flourished well into the MP3 player era, surviving even the initial introduction of flash storage due to its competitive cost. Eventually, however, your newest smartphone or iPod shipped with flash storage instead, as manufacturers bowed to its improved performance over disk storage and its increasingly competitive price point.
This is an example of a sea change taking place at a much bigger scale as well. Instead of gigabytes, think petabytes.
The data center infrastructures designed by telcos, service providers and major enterprises to store massive quantities of data have lately used predominantly disk storage in their servers, sometimes blending in flash storage for performance-intensive tasks. While the speed and performance of flash storage has tempted data center architects to deploy it more widely throughout the data center, it has only been recently that the price of flash has decreased enough to make its broader use a viable option.
To understand why flash storage has suddenly become a practical choice for data center architects across industries, it is helpful to examine the differences between flash and disk storage.
The Next Big Thing, Again
However, new improvements in flash have slashed its price significantly, positioning it as a true data center hardware alternative whose benefits - speed in throughput and latency - have dramatically increased at the same time. As an added plus, flash is highly energy efficient, needing only a fraction of the power needed by disk storage, sometimes at the ratio of one to 16. Flash drives still break down at a faster rate than does disk storage, but its boosts in performance and drop in price in recent years have made flash a realistic and highly attractive option for data center architecture and design needs.
Making the Switch
While the high-performance-demanding industries of telco and service providers are diving into flash straight away, vendors in other vertical markets have made cost-benefit calculations and have elected to hold back until the price of flash storage drops even further. For example, a Dropbox-style file hosting service for consumer cloud storage isn't as likely to be motivated by fast performance as it would be with ensuring the availability of cheap storage at scale. Companies like these are making the usual tradeoff in storage: finding a comfortable place between price and capacity. However, when the price of flash finally descends to that of disk storage, the last barrier will be removed for those companies that want to remain competitive. When this last milestone finally happens, the market shift will be as significant as when disks replaced tape storage by beating it on the same markers: higher performance and better pricing.
Advancements in Software
As background, the concept of software-defined storage seeks to move functions and features from the hardware layer to the software layer. This approach removes the dependence on expensive and annoying redundancies that solve issues based in the hardware layer. Data center architects must also plan for the inevitable failure of hardware. Flash storage, in particular, currently has a faster time-to-failure rate than disk does. In storage environments that don't use RAID cards, the failure of a disk prompts an error that will impact the end-user's experience. To solve this, architects will build in expensive and redundant RAID cards to hide the errors. By using the right software-defined strategy, these problems can be absorbed and made invisible to the end user. Since software-defined storage is hardware-agnostic, it can run on any hardware configuration.
There are a number of additional benefits that telcos and service provider data center architects can achieve by combining software-defined storage with flash hardware. For instance, the organization could still utilize a single name space spanning all its storage nodes if it were to use a software-defined storage approach. In addition, it could also run applications in the storage nodes as well, creating new "compustorage" nodes instead. As a result, the storage hardware wouldn't need to be big or costly, but could still have very high performance and speed. Organizations can start with a small number of cheap servers instead of building a large, expensive and traditional installation, and still scale linearly as needed.
Reader Feedback: Page 1 of 1
Subscribe to the World's Most Powerful Newsletters
Today's Top Reads