Proposed/Revised Testing Methodology

This is really the first time that I am testing a Hard Drive / Controller. Before that it was just a simple run of HD Tach (which doesn't work in Vista - I have a BSOD to prove it too LOL) to see my godly RAID 0 speeds. My forte still probably remains in CPU/GPU/Cooling though.

In the process of researching what type of controller would be ideal for me, I have significantly read up a LOT on the intricacies of the workings of storage.

Initially, I set out to run some simple tests (low level tests using Winbench99, HD Tune, and a tad fancier IOmeter - of which I have never used).

Though I had to ditch some things (again, in the previous post), I think it is time to revise my methodology. Primarily this has to do with the (to me) uncharted domains of IOmeter.

I initially set to test out some results using both Sequential and Random Read and Writes (take a minute to digest - public service announcement) of two Transfer sizes which are 4KB and 512KB sizes. This is mainly to find some congruence with the currently established TechARP testing methodology (if I can have a better understanding and conjure a more 'thorough' test - I will - still a WIP) but at the same time, I threw in a different mix. I chose to use two levels by manipulating the number of Outstanding I/Os. Low level benchmarks (such as HD Tune and HD Tach) utilize a single I/O request and that doesn't really show all too much load. So I decided to run two versions of the test. One with 1 outstanding I/O and the second with the max number of outstanding I/O per second - 256 (which simulates a heavy load situation - the 3ware hardware RAID controller should shine very well in this regard).

Some explanations of the jargon (mumbo jumbo):
http://www.storagereview.com/articles/200003/20000313OSandBM_5.html

So as is, I will run some tiers tomorrow (hopefully my younger brother will forgive me as I totally dominate his PC again for the better part of the day).

Tentitively, the tiers are as below (# of Outstanding I/Os):
1 -done?
4
16
32
64
128
256 - done?

I will double check my results tomorrow as well as test various things (if the length of testing has a difference or not).

Now if there are any of my audience out there (imaginary or otherwise) who would like to chime in, I am all ears/eyes. :)