Setting Up the Correct Environment to Evaluate the ANX 1500 By Tim Kellogg On July 11, 2011
Now that you know what to look for when evaluating cache appliances, we can look at the ANX 1500 NFS Throughput Acceleration Appliance.
The ANX 1500 is designed to both improve throughput of NFS data and reduce latency when responding to client data requests. The SPECsfs®2008_nfs.v3 results are a testimony to the ANX 1500’s ability to achieve these two goals. Included in the Alacritech published results is the ANX 1500 overall response time (ORT) of .92 ms, the best ORT rating of any network-attached storage (NAS) device listed.
These results are achieved by both caching active data on the ANX 1500 and accelerating data movement in and out of the caches with maximum efficiency. Naturally, to achieve the best results assumes a high hit rate, that is, most of the NFS requests are fulfilled by the ANX 1500. This is accomplished when the ANX is caching a large percentage of the working set. A high cache hit rate not only delivers great performance, but it reduces the load placed on the NAS filer, allowing the NAS to have more time to perform other operations and support more clients. In order to observe similar performance results in your own NAS infrastructure, the test environment should be configured appropriately.
A dedicated NAS filer or cluster should be allocated for the test. It’s best to mount most if not all of the clients being served by the NAS to the ANX 1500, which will cache most of the working set, achieve a higher hit rate and benefit the NAS from decreased load. If the NAS continues to directly support clients that are already overloading it, the ANX 1500 can’t do much to offload the NAS or accelerate data for those clients.
While the ANX 1500 is optimized for reads, it also reduces latency for passing writes back to the NAS. The latency associated with passing a write to the NAS is comparable to switching latencies. The ANX holds a copy of the write but doesn’t commit it to client available cache until the write has been acknowledged by the NAS. As demonstrated by the SPECsfs2008 results, overall latency is improved even with writes in the mix. Of course, it stands to reason that as the percent of write operations increases relative to reads, more will be expected of the NAS. Taking it to an extreme, if 90% of the mix was consistently writes, the benefits that the ANX could provide would be less. Fortunately, this is not the case for most NFS environments. Before an ANX 1500 evaluation takes place, Alacritech reviews the customer’s NFS operation mix, to ensure that the best results can be achieved.
Use parameters to tune the ANX 1500 performance and leverage the appliance. For example, a read ahead parameter allows the ANX 1500 to cache data prior to it being requested. Also, several time-to-live parameters can be used to reduce the number of requests made to the NAS controller, off-loading the NAS. Details describing how to set these parameters may be found in the ANX 1500 User Guide.
The ANX 1500 can significantly improve NFS performance of your existing NAS environment. By setting up the right testing environment, you’ll find that it’s easy to obtain clear and accurate test results that represent the benefits your company will receive.
Related posts:
Leave a Reply