Do you remember when hard drives were 10 MB and bad sectors would eliminate 20% of the drive? Hard drives were not very robust (nor were the PCs, the OSes, …) but that was pretty outrageous. Not only would the sectors go bad which wiping out large percentages of the drive but the drives crashed pretty often as well. Over the course of four years I had to replace at least three drives.
That sets the stage for something I thought was a thing of the past—hard drive failures. I’m not suggesting that drives never crash but with the exception of a x-coworker who used to claim his HDD crashed in order to get out of doing a demo/presentation (more than was even remotely reasonable) the MTBF must be pretty low. Even with the five – seven laptops and Tablet PCs I’ve had which at least have a reason for failing do to all of the banging around in my bag I’ve had zero HDD problems.
In the past three months I’ve had two hard drives crash. One drive was in my Ultimate TV (upside: I now have 3x the capacity I had—room enough for recording almost all of the SciFi Channel’s Twilight Zone New Year’s Day marathon!) and just this past Tuesday the 160GB HDD in my Windows Media Center Edition PC (I popped in the drive that it replaced and it worked fine so it is a drive failure). The UTV was two or three years old and that drive was always being written or read from because that’s just how it works. The latter drive was only three months old. It was manufactured by a well respected, name brand OEM. They have a great warranty/return policy all completely done via their web site. They even give the warranty holder the option of using a credit card so you don’t have to wait until they receive your drive before sending you the replacement. Kudos for the fully functional web site. However, maybe drive crashes happen a little too often which is why they have the return section really nailed. <g> I’m joking, sort of. What a bummer though if I hadn’t of had a backup of my data.
We never turned off the MCE PC and it has far too little memory (see earlier RDRAM post) so it swaps constantly. I always felt bad for the HDD as the light indicating disk activity looked like the one on my firewall that indicates packet traffic—it was constantly on. Did that contribute to its failure? It might have but I’ve got another PC that is six years old that is always on and it’s humming only nicely (but with 756 MB of SDRAM it never swaps).
I have to find an advocate within Microsoft to get my PC RAID5 idea on the feature list (unless it is already being developed and I just don’t know it which is quite possible). Because I’d gone so long between crashes it almost seemed like it wasn’t worth pursuing but now I know it has merit. If it weren’t for doing very regular backups I’d been hosed. Fortunately, between the script (which several of you were kind enough to help me with only last week—see earlier post) which distributed our music and photos to all of the other PCs and my external HDD I’ve not lost anything (well maybe my recently updated shared OneNote file containing future blog ideas).
I’m really trying to dive down into the details of Longhorn’s WinFS because it’s a fascinating idea. How do you build redundancy into the “database” engine such that the inevitable corruption (unexpected outages might not happen on servers because of UPSes but they certainly do on PCs) doesn’t mean data loss? I’m glad there are some really smart people thinking about it because I certainly don’t fully understand how to make it 100% fault tolerant. I’ve tried to grab some of the WinFS PM’s blogs but I haven’t read anyone discuss that issue (though it could be a trade secret).
BTW, doesn’t the capacity of the drives fifteen years ago seem preposterous compared to the capacity we have today? I’m sure that’s exactly what we’ll be saying in ten years when terabytes of storage are the norm. <g>