Originally posted by: Madwand1
Originally posted by: Minerva
The two Raptors RAID0 are right about where I was with a pair on NVRAID on the A8N32SLI with a 3GHz FX60. With a pair on the ARC-1280ML with 2GB cache it jumped to about 36k. As more drives were added it scaled perfectly. With 8 drives it tops 200,000!
I hope I don't sound jealous and bitter (although I kinda am..), but I think these numbers are somewhat hokey. The tool uses 64k access, which is fine for typical use, but not all that's seen, and some RAID arrays will give real-world non-cached performance that is better than what's reported by the tool. In addition, the 2 GB controller cache part likely tricks the tool into producing results that are far better than what would be seen -- here I assume that the tool deliberately turns off the system file caching to get more meaningful results, but this doesn't work for the controller cache.
Edit: Also, as a file-system level tool, the performance would depend quite a lot on what's already on the file system, so a new/empty drive / array will measure quite a bit better than an old/full one.
Edit2: Please take the above with a grain of salt. This is a random utility with no documentation or source code that I can see, so it's hard to say what it's actually doing. Just a couple of guesses I've thrown out; could easily be wrong.
Edit3: It looks like the tool doesn't work at the file system level, so ignore edit (1) (but still consider edit (2)). I've re-confirmed that the tool significantly under-reports performance compared to other tools and actual usage in some cases with RAID arrays. I guess now that Minerva's high-end setup is able to overcome these aspects of the tool. 500+ MB/s is not unreasonable for such a setup, provided that enough/large enough requests get to the drives in time.