On 2008/01/03 16:59, Tobias Weingartner wrote: > Stuart Henderson wrote: > > > > It wouldn't be more likely that the disk _crashes_ by doing this, > > and it may give _some_ protection against _some_ failure modes. > > It also gives new and exciting ones to take their place. > > Actually, since you'd be mirroring to two different portions of the > same disk (assuming a non-flash device), chances are you would be more > likely to crash.
Indeed, let's try that again ... By writing to two areas of the same disk, you can do something to protect against one failure mode (unreadable physical sections of the disk) at the expense of increasing the risk of others. > 2) You'll be writing everything to two parts of the same disk, making > the disk continuously seek 1/2 a disk distance. Likely not something > you want to promote. I mentioned software complexity but definitely overlooked the additional seeks which are not a good thing. Thanks for the correction.
