The Do's and Don'ts of SSD Implementation

Arthur Cole

Advances in enterprise-class solid-state disk technology are coming so fast that the question of whether to deploy SSDs in the data center is quickly being replaced by when to deploy, and how.


While the storage industry at large continues to generate increasingly dense systems capable of cramming more data into tighter spaces, the same trend is hitting the SSD market. The only difference is that more dense SSD formats not only deliver increasingly compact storage modules, but go a long way toward evening up the still-significant price discrepancy between solid-state and traditional spinning media.


Just last week, Intel and Micron announced plans for a 25 nm NAND flash device, which stands to halve the size of its current 8 GB die built on a 34 nm lithography process. The new device is expected to measure 0.35 x 0.74 inches and will be hold only half of the 64 Gb chips needed under existing designs. That means a 256 GB drive can be built with only 32 of the 8G dies, which should go a long way toward cutting the price of everything from tablets and mobile phones to full SSD drives.


There is no doubt that enterprises will see a net benefit with SSDs, particularly when it comes to high I/O applications and environments. But as Storage Switzerland's George Crump explains, there are myriad ways of implementing SSDs, and success or failure will depend largely on your expectations and your existing storage infrastructure. Deploying SSDs in conjunction with some of the new automated tiering systems and cache-type appliances, for example, can go a long way toward reducing disruption to your overall storage environment.


It also helps to have a clear idea of what kinds of applications lend themselves best to solid-state storage, according to Datamation's Henry Newman. Most databases, naturally, will benefit from the extra IOPS that SSDs provide, but even then, the ones that rely on small block I/O requests probably will show the best results, considering a standard SSD can deliver upwards of 30,000 IOPS compared to perhaps 250 for a 15k mechanical drive. And you can leverage SSD performance even more with upgrades to underlying server and networking hardware.


It also couldn't hurt to consider SSDs for applications like business intelligence (BI), data mining and virtual desktop infrastructure (VDI), according to John Hope-Bailie, technical director of systems distributor Demand Data. You can also target them at specific uses like cache support for data file and Web servers, real-time data processing, CAD/CAM and 3D animation and rendering. It's important to note, though, that many SSDs don't come with tools like error correction and native encryption, so it's wise to deploy them with appropriate backup mechanisms, especially if they are being considered for mission-critical data.


Despite all the headlines they've generated over the past year, SSDs are not poised to "take over" the enterprise. Most IT execs should prepare for a mixed mechanical/solid-state environment for some time to come.


And since SSDs will come at a premium price for a number of years, it's wise to make sure you have a detailed deployment and implementation strategy in place before you swap in your first drive.



Add Comment      Leave a comment on this blog post
Feb 5, 2010 10:16 AM Daniel Feller Daniel Feller  says:

I know many people are asking about SSD in regards to desktop virtualization.  When talking 1,000-10,000+ users, there is a need for greater IOPS so many are looking at SSDs as the answer.  And because many of the writes are small in size, the fit makes sense.  But because of the cost and the newer technology, most are reluctant to move forward in massive amounts.

Reply

Post a comment

 

 

 

 


(Maximum characters: 1200). You have 1200 characters left.

 

null
null

 

Subscribe to our Newsletters

Sign up now and get the best business technology insights direct to your inbox.