ATTO Disk Benchmark + Large Cache = Bad Results

[Stereodude]

Here's why you can't trust ATTO on a RAID controller or disk with a very large cache. The 256MB data length is just too small to get accurate results. Here's 5 back to back runs with ATTO v2.41 on my 6 x Samsung 5400RPM 1.5TB RAID-5 array on a Dell Perc 5/i with 256MB of cache.

First you can see they're not all that consistent run to run. Second if we compare them to an Iometer run set up to mimic ATTO's testing but with a much larger data length we don't see a whole lot of agreement. In the slowest test shown below (.5K read) IOmeter read ~7.36GB of data taking the cache out of play.

 

[LunarMist]

Why should you care about the benchmarks? Are the drives performing as intended for your purpose?

 

[Stereodude]

Why should you care about the benchmarks?

Well, if you want to accurately compare different stripe sizes and other settings, or compare different drives, number of drives, etc to find the best settings for your intended uses it helps a lot if the benchmark is accurate. :nono:

Are the drives performing as intended for your purpose?

I'm sure they will. I haven't yet put any data on them. I have been stress testing the array and making sure it is robust before I load it up.

 

[Santilli]

Didn't one of the links go to a test site on SSD's that agrees with you, but, has some sort of testing tool that they are sold on?