Are there still reliability problems with high capacity magnetic disk drives?
When 500GB and 1TB drives first appeared on the market, they had significantly lower reliability than their smaller (250GB and less) brethren. I have first hand anecdotal experience, as I bought several 500GB and 1TB drives (< 10 total) over a span of about 2 years (2006-2008) and have had 2 of them fail within 12 months of purchase.
In early 2007, Google released a study of drive failure statistics (http://labs.google.com/papers/disk_failures.pdf) collected from their data centers. This study indicated, during the period of the study, that the high capacity drives had a significantly higher failure rate than the lower capacity drives.
If reliability is an issue I recommend the Western Digital RE3 disks. You pay about a third more than entry level disks, but they are designed for 24 x 7 running. The RE3 range has only been out for a few months, but I have several of the previous RE2 and RE disks in servers and so far none have failed. They're also pretty fast, and they work very well on the Dell Perc5/i controllers.
John Rennie
I was looking into disks and one interesting tidbit is that by the time you get to a terabyte, consumer drives will spend a few minutes trying to recover sectors that start to go bad (too much error correction needed). So if your big drive can't read a particular sector it might start crunching away on that sector for 2-3 minutes until it can (hopefully) get a good read and relocate the data. Yikes!
Terabyte+ drives designed for RAID have a different firmware that won't try as hard to recover bad sectors to avoid timing out the RAID controller.
It depends on the manufacturer and specs of the drives, but on the whole they are getting more reliable. You will have to spend more for higher MTBF.
However, the increasing size of the disks does pose a challenge for things like raid, since statistically speaking even on reliable drives there's a higher and higher chance for a bad sector that could trash a rebuild. This makes it increasingly important to use something like raid 6 that can tolerate 2 disk failures rather than just 1.