Achieving Low Latency Non-Disruptively By Doug Rainbolt On February 4, 2011

When we set out to design the ANX 1500, our objective was to create a product that would sit alongside the NAS that customers already have and synergistically deliver better performance, better scalability, and easier to manage infrastructure. All while saving the customer huge sums of money and floor space.  Keeping it really simple, think of the ANX 1500 as a giant shock absorber for the NAS that you currently own. It takes care of responding to a great deal of your NFS data requests so that your NAS is freed up both in terms of controller cycles and disk-drive dependencies. The ANX 1500 not only delivers very high system throughput, it also substantially reduces latency. Think of this as being the average time it takes from the receipt of a request to the fulfillment of this request. Obviously, the lower this number the better. The ANX 1500 can achieve latency of .2ms based upon a typical load consistent with industry leading benchmarks.

Alacritech’s synergistic approach, working with the NAS you own, means taking advantage of the NAS’ strong points. For example, many of today’s NAS solutions have ample NVRAM and are very well-tuned for supporting NFS writes. NetApp is an excellent example with its WAFL file system. So why not take advantage of the NAS’ write capabilities while the ANX 1500 focuses on caching reads?

A 0.2 ms latency was achieved with the ANX 1500 acting as a write-through cache supporting a mid-range NetApp Filer. This assumed that NFS writes hovered right around 10% of the overall NFS Operations mix, which is fairly typical and consistent with leading file server benchmarks. The 10 μs latency associated with the ANX 1500 passing a write thru to the Filer was overshadowed by both the Filer’s ability to use its write-back cache to quickly satisfy client requests and the ANX 1500’s very low latency in handling client requests for cached data, be it payload or NFS metadata. In other words, the 0.2 ms latency incorporates both reads and writes, utilizing the ANX 1500 and the NAS it supports. The respective products work together — they don’t compete. Alacritech is more than happy to have ownership of data reside with the NAS vendor.

So just what are the benefits of generating only 0.2 ms in latency? While storage administrators are sometimes removed from revenue generation for their companies, they are often asked to look for ways to reduce response time to the user community, be they internal or external. This can mean reduced build times, faster project completion, and often, more revenue opportunity for the enterprise. It can also translate into taking some of this latency benefit and generating cost savings in capital, space, and power. By targeting response times to service clients, administrators can dial up or dial down infrastructure. A 0.2 ms latency could suggest, assuming that 0.7 to 1.0 ms is the target, that there is opportunity to use more capacity on disk drives or even use SATA drives in place of high performance SAS drives. Of course, administrators have to plan for peak times and occasional issues that could impact performance in the data center. The point is: having a significant latency cushion provides strategic advantage, including flexibility.