-- [ Picked text/plain from multipart/alternative ] Raid-0 is not random, but you are right, there is no redundancy.
On Thu, 09 Feb 2006 00:56:55, Dan Sorenson <[EMAIL PROTECTED]> wrote: > > At 12:44 AM 2/8/2006 -0800, m0gely wrote: > > >> The quickest way to > >> boost disk read speed is to run a raid-1 config, because both disks can > >> be read from alternately -- whichever one has the sector needed closest > >> to the heads gets used. > > > >Did you mean to say RAID-0? 1 is for mirroring. As for the closest > >sector remark, heh, what the heck are you talking about? > > Be warned, this is going to be a little geeky. Raid-0 is no > raid at all, it's merging two or more drives into one volume. Where > data is read or written to is random. Raid-1 is mirroring. You've > two disks with the same information on all tracks and sectors. Tracks > are the outward-to-inward portions of a disk, sectors are the segments > around those disks. So let's say I have a Raid-0 with some data on it, > I might be reading all of it from disk 1 or disk 2, or maybe half > from disk 1 and half from disk 2. That's not efficient because I > have to read that data in an order, so I may read it as disk1 disk1 > disk2 disk1 disk2 etc... > > Raid 1 is two disks mirrored. Here's where it gets really > pretty. Let's assume they're running in synch, disk 0 and disk 1 have > 10 tracks and 4 sectors per track covering 90 degrees of the platters > and are spinning together. Let's say I need to read a map from sectors > 1 and 2. With Raid 1 I can read sector 1 from disk 1 and sector 2 from > disk 1 on the next revolution. That's no better than having one disk. > Let's say the disks aren't in synch. I can then read sector 1 from > disk 1 and sector 2 from disk 2 that's only 1/4 revolution behind. > Or, I can read sector 2 first from disk 1 and cache it and grab > sector 1 from disk 2 as it's coming around. > > Oh, if I have to write data I have to write it to 2 disks. > So that's two write operations waiting for the disks to come > around, but again it they're out of synch I can always write to > them in the order they come around. > > Raid0 means I can't predict which disk is going to be the > next one I can read from or write to. Raid-5 means I have to do > at least 2 reads or 2 writes before that data is considered valid, > which means I might have to wait for three sectors to pass by > and heads to move to the proper track before I get my data. > > Raid-5 means (disk MB * (disks - 1)) is your volume size, > so 5 x 100GB disks is a 400G redundant volume. That's 80% of > disk available vs disk purchased. Raid 1 is (disk MB * disks/2)), > so you buy a gigabyte of disk and you get half that usable. > Naturally, this makes Raid-1 expensive for storage but the fastest > available. Raid-5, at 66% to over 80% capacity depending upon > the number of disks, is slower but a more efficient use of disk. > > >If he had an 8 year old 1GB IDE hard drive it *wouldn't* cause the map > >changes to take 30~40 seconds. > > Agreed. I was giving a lesson in isolating the limiting > factor and how best to work around it. Still, I felt the Raid-0 > vs raid-1 vs raid-5 advantages and trade-offs deserved a little > more elaboration. > > - Dan > > * Dan Sorenson DoD #1066 A.H.M.C. #35 [EMAIL PROTECTED] * > * Vikings? There ain't no vikings here. Just us honest farmers. * > * The town was burning, the villagers were dead. They didn't need * > * those sheep anyway. That's our story and we're sticking to it. * > > > _______________________________________________ > To unsubscribe, edit your list preferences, or view the list archives, > please visit: > http://list.valvesoftware.com/mailman/listinfo/hlds > -- _______________________________________________ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds
