On Mon, 03 Mar 2014 17:36:10 +0100 Jean-Christophe Deschamps <j...@antichoc.net> wrote:
> > >It's how RAID5 works. Check this page docs http://baarf.com/ about > >it. > > This is utter BS. No. > Serious RAID controllers perform parallel I/O on as many drives that > are making up a given array. Of course I'm talking of SAS drives here > with battery backed-up controller. > > Kid sister RAID5-6 implementations using SATA drives and no dedicated > hardware are best avoided and have more drawbacks than are listed in > cited prose. > > I run 24/7 an Areca 1882i controller with 6 SAS 15Krpm drives in > RAID6 and a couple more in RAID1 and I've yet to witness any problem > whatsoever. RAID3-4-5 was great when disks were expensive, in 80's an 90's. Now not. A minimal RAID5 needs 3 disks. A minimal RAID10 4. An enterprise disk SAS 15Krpm 146 GB 6G is $350, and a not enterprise grade cheaper and bigger. Now RAID1E and RAID10E give more flexibility and variable security, from "paranoid" to "i don't care" grades. When something goes wrong: RAID 3-4-5-6 When one of your disk brokes, replace it. Then rebuild the RAID3-4-5-6. You need read from all disks to recover the lost blocks. All disks are busy recovering it and your R/W performance drops. Recovery reads the same block on and the parity data, makes some computations and writes the lost block. If any of the RAID disks is near its MTBF and fails, you lost everything. RAID 10 When one of your disks brokes, replace it. Then rebuild the RAID10. You need read from mirror disks to recover lost blocks. Only the mirror disks are busy recovering and your R/W performance drops only when accessing data in those disks. Recovery reads the same block and directly writes lost block. If all disks that mirrors to broken one are near its MTBF and fail, you lost everything. The time to recover a RAID 10 is less (lot less) than recreating a RAID3-4-5-6. > It's just like talking BS on a language because of some obscure bug > in a non-conformant compiler. No, it's talking BS on language that is bad designed for your actual needs and no matter which compiler you use because is not an implementation problem. --- --- Eduardo Morras <emorr...@yahoo.es> _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
