Nothing about the disk "geometry" means anything anymore. Back in 4.1.1 days I measured the effects of various geometry based disk parameters before building filesystems. But they haven't meant anything for a very long time.
As for something like Thumper, you shouldn't have all the disks in the same vdev anyway. 10 GB of disk space is about $0.50US. So "wasting" a few small s0 slices in a 5-8 disk RAIDZ array is cheaper than using USB flash drives. Also 100% allocating drives leads to the situation that a similar size drive might not work as a replacement. If the replacement is a single sector smaller, you can't use it as a replacement. There's a lot of screwy old logic in format, prtvtoc & fmthard and sd that tends not to play nicely w/ modern drives. Have Fun! Reg --- On Fri, 3/1/13, Jim Klimov <[email protected]> wrote: > From: Jim Klimov <[email protected]> > Subject: Re: [OpenIndiana-discuss] Installing OI on USB stick > To: "Discussion list for OpenIndiana" <[email protected]> > Date: Friday, March 1, 2013, 7:56 AM > On 2013-03-01 14:44, Reginald > Beardsley wrote: > > I'd ask why? There's no requirement that there > only be a single OI slice per disk. The installer and > various other things push in that direction for some reason, > but it's not a valid limitation that I can see. > > Well, approaches good for smaller machines are not so good > on bigger > ones. Say you use this with an array of 48 drives in a box > (Thumper) > and wanted all vdev sized to be equal for balanced IO. Would > you > reserve the rpool size on each of the drives? (Thumper might > be a > bad example, only two disks are bootable anyway, so the tail > of the > rpool disk can be used for something else, like a big temp; > but you > get the overall picture). > > Now, there might be uses for smallish partitions on faster > sectors > with shorter seek strokes - you might use them for L2ARCs, > ZILs, > or just a pool with more speed-critical data. But that would > be a > contrived use for something you didn't intend to have in the > first > place, quite likely :) > > //Jim > > _______________________________________________ > OpenIndiana-discuss mailing list > [email protected] > http://openindiana.org/mailman/listinfo/openindiana-discuss > _______________________________________________ OpenIndiana-discuss mailing list [email protected] http://openindiana.org/mailman/listinfo/openindiana-discuss
