At my company, we are about to do a major software upgrade, and our contact for this software has given us some benchmarks to hit for our new server that the software will be put on. Specifically, he has suggested we make sure our servers hit 2,000,000 KB/s throughput using CLIBench. However, our new servers (which are very powerful) can only seem to hit around 700,000 KB/s using this benchmark software. I am no fan of it, since it is nearly as old as I am, and other benchmark software such as CrystalDiskMark give results that are much higher and more in line with what I would expect to see.
I have been pulling my hair out trying to figure out how to get this number to go higher (my boss, coworkers, and the contact insist I do so) to satisfy the recommended speed for the upgraded software. One thing I have tried is changing the stripe size of the array, but every benchmark software I run seems to spit out higher numbers when I set the stripe size to a much lower size, when every site says that higher stripe size is faster. I'm at a total loss here.
The server I'm testing all this on has 4 1.6TB SAS SSDs set in a RAID1+0 array. It's all new hardware, and from what I've seen of the specs, it should all be up to snuff. Any help would be appreciated.
