On Thursday, 5th May 2005, Chuck Robey wrote:
Stephen McKay wrote:
And I'm fairly certain that you're right on this. I'm too busy right this minute to give the project enough research to get it right ... I have two disks, each 145G, and which I wouldn't mind giving up up to 1/3 of the space, if I could, to reliability, I really want the speed, and I would NOT give up the entire disk, I don't need that sort of reliabilty, I just don't. So, would you mind tellme, what do you think I ought to do?
Hmm. If you split each disk at the 1/3 point and mirror 2/3 of each disk, you end up with 66% of the disk space usable: 33% striped, 33% mirrored and 33% wasted (in the mirror). In your case, you would have about 96GB of fast disk, and about 96GB of "safe" disk, instead of your current 290GB of fast disk. It would be up to you to determine what goes in the "safe" partition and what goes in "fast" partition.
Is that what you were asking for? I started with something like this and pretty quickly decided I needed "safe" more than "fast", everywhere.
Maybe if you split them in half instead (resulting in 50% striped, 25% mirrored and 25% wasted) you'll find that you don't need so much speed and space after all and can then easily mirror the other half later, resulting in two equal sized mirrors.
I like this quite a bit, you're going to have to flag me down if I get too garrulous, a discussion like this is exactly my cup of tea.
Anyhow, breaking up one single disk area into two areas is something I basically dislke, because I have found having a single area to be very very useful, in unpredictable circumstances, but often enough so that I wouldn't give it up without a stronger reason.
What I was saying, if the resulting disk area were to devolve down to as little as the vague area of about 200G (from the current 290G), if this would result in markedly greater reliability, but without too great a loss of speed, I would just jump at the chance to do that. BUT if I were to lose a great deal of speed, I would duck out. If I did it, and it didn't result in really greater reliability, I wouldn't consider it a good deal either.
What I *think* I have now is the twpp disks, 145G each, striped together so that both disks are accessed in parallel for data. I set it up so that each disk is on a separate controller bus (dual scsi controller) so that it would maximize the transfer rate, and hopefully result in double the effective transfer rate, and as far as rotation rate controls transfer rate, double the efective rotation rate (no change in access times, too bad) so that the effective rotation rate rises from 10K to 20K, as it affects transfer rate. On scsi3 ultra160 disks, this seems a good thing. You've seen the dump from my disks (from vinum) right? Do I have the setup I describe? If not, ok, what do I do to get it?
Stephen.
_______________________________________________ freebsd-chat@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-chat To unsubscribe, send any mail to "[EMAIL PROTECTED]"
