Lots of vendors claim high performance for their NAS platforms. As a solution provider in the channel, how do you define high-performance NAS?
We actually define a network-attached storage (NAS) product to be "high-performance" once it crosses certain performance thresholds in terms of file system, file system limitations and theoretical maximum throughput. The theoretical maximum throughput includes NFS operations, IOPS (input/output operations per second), as well as file system metadata caching. We utilize both cluster-based and ASIC-based technologies to deliver high performance in a NAS storage platform, but lean more toward the ASIC-based technologies, because we know what the sustainable throughputs and IOPS and other performance attributes of that solution would be regardless of the data set.
 |
| About the author |
| Jason Sparks is VP of storage and systems at Xiologix. Listen to more from Jason in our high-performance NAS podcast |
|
|
 |
 |
It's important to note that each NAS vendor applies their own definition of "high-performance," which can vary substantially from our definition as well as the definition used by other vendors -- this can be extremely confusing for our clients. Vendors tend to define high-performance NAS in terms of what they do particularly well. Just look at any vendor's website. You'll usually see that vendor touting one specific feature of performance rather than an overall well-rounded performance specification.
For example, some vendors focus on throughput, but that is only good for certain types of applications, such as streaming media or online media presentations. Their product specifications will continuously talk about X gigabytes per second of total sustained throughput, but that means nothing to the average person. Another NAS vendor handles file systems extremely well. They talk about NFS operations or IOPS consistently throughout their product brochures, but their sustained throughput is really very low -- on the order of megabytes per second. This means the NAS product handles file systems in short I/O bursts very well, but doesn't handle large block I/O or large files well at all.
As a solution provider, you need to define what the particular vendor does well, and then put it into layman's terms that clients can understand based on their data sets. I think it would be great if there was a standard set of specifications that every NAS vendor would report against.
Return to the high-performance NAS FAQ guide and read the rest of Jason's expert responses.