LSI SANDFORCE DRIVEN NVELO CACHED PERFORMANCE
Prior to providing performance results for the NVELO cached system, it is necessary to address the system we are working with. To start, our normal procedure is to test an SSD that is fresh and new. In the case of the ADATA XM13, this would provide for performance of 280MB/s read and 231MB/3 write transfer speeds. The difficulty with this was that our Crystal Disk Information software identified a total of 134GB host reads, 165GB host writes and a power on count of 83 with 15 hours of activity, a bit unusual.
These numbers could account for validation testing and burn in but are a bit high. We also need to recognize that this system contains an operating system and the configuration of the OS is such that it is as easily used, without problems, as can be. Summarily, neither the hard drive or SSD are in a fresh state so we would expect lower numbers than previously attained.
Although significantly better than initial results, ATTO write performance is still very low with 76MB/s read, yet is just about three times faster when measuring read performance. This is not indicative of any previous NVELO Dataplex caching software testing and would most likely be the result of the initial hardware and software configuration.
In this example, we have the results of the NVELO based T14 on the left with that of the SSD alone on the right. Once again, results are significantly better than where we started, but not close to where they should be. The cached system result on the left is still displaying typical write performance that we might observe in the hard drive alone.
We ran Anvil Storage Utilities four times and caching displayed improved performance with each run (1, 2, 3), confirming that our caching software was doing its job. Oddly, we still haven’t had any success in bringing up the write performance. We also had no luck with AS SSD Benchmarks and gave up after 45 minutes with an estimated 35 minutes remaining in the 4K test.