Maybe leslieann has better info on how to config a system for your use. [/size][/color]
Personally, I never let them spin down for faster access.
I'm not sure I'd want windows siting idle without sleep controlling a drive spin up/down for the reasons TP4 mentioned. Either make the whole system sleep, or leave it all on. Modern Windows is really geared towards SSDs at this point.
Let's say Person A buys 8 new drives and builds an array. 2 drives worth of space are reserved for redundancy and 6 for actual data. If 1 or even 2 of the drives die at the same time, your data is safe while you rebuild the array.
Now let's say Person B buys 10 refurbished drives and builds an array. 6 drives are for data and 4 are for redundancy. While the odds of a refurbished drive failure goes up, the odds of 5 drives dying simultaneously brings the failure rate of the overall array way down. Don't forget you'll need more space for those extra drives and they'll consume more electricity etc. But in the end, you should get better reliability for a lower cost.
You want your drives from multiple batches for any array. The thinking goes that drives from a batch will typically die at a similar rate, so once one fails the rest are almost sure to follow, and I have seen it happen. So you get some drives from one batch, others from (several) others. Admisn will often buy from different retailers from different parts of the country to get a mixed batch.
Another issue is safety.
If several drives have failed, I don't care how much redundancy you have, you're on the tipping point because of batches, even if mixed, you're rolling the dice if more than one has failed. More importantly though, with the transfer speed of spinner drives and capacity of todays drives, it can literally take DAYS to rebuild an array at 100% load. Todays drives s*ck for this reason and something needs to change for them to really remain viable in the future for any use really.
Lastly...
Backblaze should never be used for reliability ratings on desktop drives, the duty cycle is completely different. Backblaze fills a drive (quickly) then it mostly just sits, whereas a desktop drive gets an initial modest load (windows, maybe a backup) and then lots and lost of start/stops, add remove... for years. Even with a home server, it's likely not run like what the Backblaze duty cycle is. No one should bother with those reports except backup service providers.
The report you want to look at was the old Google drive report where they found drive brand and size was less a prediction of failure than several key S.M.A.R.T. functions, (If I remember right) the most important ones being if a drive showed lots of corrected/read errors (can also signify a sata cable failure) or almost ANY reallocated sectors, events or uncorrectable errors. I have seen some with a few bad sectors live a long life, but the moment I see more than about half a dozen I replace the drive. Beware, once that number crosses a dozen or few dozen, it starts to climb VERY fast and I've seen drives fail as I was copying off the data more than a few times.
BTW, these days, shucked drives may not have a sata connector and over a USB connection a drive may not have S.M.A.R.T. functionality.
