Yada yada yada, of course, lightning never strikes, computers never get
wet, etc, etc, etc... You still have to back things up, although raid5 does
save you lots of time and frustration when a single drive does fail, which
will be the majority of cases. Of course while that drive is out, unless you
have a replacement sitting on your shelf, you're operating at raid0. The
cost of raid5 then is not just the extra parity drive, but its replacement.
Go ahead and rely on raid5, and BTW, MTBF is a bullshit number and you
know it. You rolls your dice, you takes your chances. Besides, MTBF is
not based on how long a single drive lasts. It's not a hard and fast
number, that can be used to calculate failure rates for a multidrive
system, at least not so easily. It's based on failure threshold and
percentage of drive failures. 80% of the drives will fail after 250,000
hours, based on 200 drives for example. Each manufacturere is free to come
up with their own MTBF numbers, so it's a totally bogus number. Higher
MTBF's only suggest your drive is less likely to fail immediately.
A dead drive in a raid array doesn't stop you from rebuilding the array
with what's left either, it's just more time consuming.
> If want hardware with a MTBF around 3.5 years then talk to PackardBell.
Sorry, it's a crapshoot these days, and that's it. Your a lot better off
getting rid of this notion that you can buy good hardware that will last the
ages, because it's only going to burn you until you do.
- Steve
P.S. This thread doesn't belong on the kernel list, you're welcome to
move it over to server-linux.
.------------------------------------------------. # * # # # # # #
| Steve Baker | Barely Working | # ## # # # # #
| ice@mama.indstate.edu | System Administrator | # # # # # # # #
| Red-Hat Rulz! | Will work for hardware | # # # ## # # # #
`-- SYS-ADMIN FOR HIRE, HAVE UNIX, WILL TRAVEL --' #### # # # ## # #