This is an entry in my Building a NAS project.
I made the final decision on my hard drive choice and configuration.
I purchased five 3 TB Seagate Constellation ES.2 drives, model: ST33000650NS.
I got them from the big Ebay seller GoHardDrive. These drives are ‘new’ old stock, meaning that they are new and unused drives that did not sell. They are not under warranty from Seagate, but GoHardDrive (the seller) offers a 3 year warranty. Each drive cost $77.
I decided on ‘new’ old stock because of price. A 2 TB drive runs about $50-55 and a 3 TB drive runs about $70-80. ‘Normal’ new 2 - 3 TB drives are either not manufactured (?) or are more significantly more expensive ($100+). The only notable exception is Toshiba’s P300 drive, which runs about $80 for 3 TB. I did consider that drive, but decided against it because I could get a longer warranty period as well as a enterprise-class drive for about the same price.
I expect reliability of these drives be nearly as good as any new drive. These are unused enterprise drives. Enterprise drives are designed for heavy use and long life and their zero use means they have their entire life ahead of them.
I settled on this particular drive because the seller offered a 3 year warranty on them. 1 year is standard and 2 years is occasionally offered. But a 3 year warranty means I can go to the seller for a replacement if one fails for the next three years.
Due to the redundant design of my storage, I originally was comfortable with not having a warranty. If a drive died, I would simply replace the drive. The discovery of drives that have some warranty was an added bonus.
Had I not chosen this particular drive, I would have most likely picked a 3 TB Hitachi Ultrastar 7k4000 or 7K4000 series drive. It’s a ‘new’ old stock enterprise drive made by Hitachi. I chose this drive for the same reasons I picked the Seagate drive. The Seagate drive simply had a longer warranty period.
For the array, I have decided on a 3 way mirror.
I will then add the fourth drive to the array, have it mirror all the data, then remove the fourth drive and store it in a remote location. I will then add/mirror/remove the fifth drive and store it locally.
I will add/sync/remove the local drive on some regular schedule, probably monthly. I will also add/sync/remove the remote drive on a less-frequent-but-regular schedule, probably once or twice a year.
This gives me an array that can suffer two hard drives failing and still be operational. In the case that three drives fail, I have a local copy of all my data. Only the last month’s of data would be lost. (My data won’t change too much and my desktop computers may have a copy of the most recent data.)
And in the event of some large disaster where all three drives in the array and my local backup are destroyed, I still have a copy of my data stored remotely. I will likely lose some data, but I will surely have most of it. Especially considering the magnitude of an event that would cause such a situation.