Oh, my. This thread really took a turn. Guess I'll clear up a few misconceptions real quick.
TLER is important for hardware RAID controllers(and many SATA controllers). TLER prevents a hard drive from being "dropped" from a hardware RAID because of a single bad sector. This was a big deal years ago, and is getting worse. The reason being that if you have a RAID6 with 8x4TB drives and one fails you want to drop in a new drive and rebuild the array. Well, because of the URE(Unrecoverable error rate) there's a very good chance that a second disk will encounter an error at some point during the rebuild. If it takes too long to do error recovery(something TLER minimizes or prevents) then you will have a second disk drop from the array. Its kind of frowned upon to have 2 disks missing from an array at the same time. ;) This scenario can and does happen with non TLER drives. Don't like it, but the very expensive Enterprise RAID drives. They simply report to the SATA controller that a hard drive error occurred and let the RAID controller do the data reconstruction with parity data. After all, that is what RAID with redundancy is supposed to do, right?
With non-RAID environments some standard SATA/SAS controllers have a built-in timeout anyway. If a drive doesn't have TLER and keeps trying to read the bad sector forever(some do), the disk is as good as useless after that point anyway. Even if your SATA controller doesn't disconnect the drive from the system. ZFS will start racking up errors, the system may crash if your SATA controller is bargain bin quality, etc. If you are like alot of people, you may have purchased a RAID controller for cheap on ebay or reusing an old controller but have set it to JBOD mode. Every controller I've ever worked with will still drop hard drives without TLER if problems begin even in JBOD mode.
ZFS, being a software RAID(of sorts), has no control over if/when a drive is "disconnected" by your SATA controller. So TLER is nothing more than protection from losing more drives during the resilvering process. Some people find that the extra $10 or so give you a good piece of mind, others don't care. It's like having a seatbelt for your car. You can choose to ignore it, but that one time when you need it, you may have wish you had used it. In a business environment, going with anything besides Enterprise class or NAS class(if there is such a thing) is crazy. At home, it all about how important your data is. If you aren't making backups religiously then going with anything besides NAS-class is taking a risk.
But I will tell you that when it comes time to buying drives, I have 24 WD Greens and aside from self inflicted failures from overheating early this summer, I've had excellent results with the drives. Only 1 failure on more than 3 years of uptime for my drives. If I had to build a new system today I'd probably go with the WD Reds just for the longer warranty and the potential TLER. But, WD Reds haven't been used extensively in the NAS environment yet. They are relatively new to the market being out less than a year.
WD Green's intellipark is easily disabled with the wdidle.exe, so saying that WD Greens aren't good for NASes is rubbish. Those that know better can easily enjoy the lower power and cooler WD Green series of drives just by using wdidle.
One thing I'd absolutely never do is buy 7200RPM+ drives except for situations with amazing cooling solutions. They draw too much power, don't really provide much for latency savings when you consider a file servers total function, and they are more prone to overheating which leads to premature failure. So think before you buy. Your data may rely on it.
