On Wed, Sep 26, 2007 at 11:17:54AM -0700, David Rees wrote: > Your BackupPC server's disk is completely maxed out. Looks like it is > doing a lot of seeking. To get more throughput, you'll need more disk > spindles. RAID1 will improve random read IO performance, but you'll > need RAID10 w/4 disks which should get you 2-4x read performance and > 2x write performance.
So I've been of the opinion (not backed up by experimental data) that a concatenation (what linux md driver calls LINEAR; similar effects can be realized with LVM) of two RAID1's would be better for BackupPC than a RAID10. My rationale for this is that in RAID10, the disks are generally seeking to the same spot, unless you have a write that doesn't span across multiple raid stripes. This certainly happens, but i suspect most writes span multiple stripes. i guess this really depends on the RAID stripe size, bigger would be better. IF i'm right and your disk heads are seeking more or less in tandem to one another, then you don't really buy much random i/o write performance by going to RAID10. On the other hand, with the concatenation, over time, as the disks fill up, you are likely to have the disk heads seeking totally independently of one another so you should get more random IOPS. You don't have any control over which spindles get which I/Os, however, so it could easily be imbalanced, especially at the beginning before the disks fill up. database guys always say "raid10 raid10 raid10". Presumably, backuppc's random I/O is similar to a big database. So maybe I'm just wrong and there's something better about RAID10 that I don't understand. On the other hand, if they're just comparing to raid5, well, yeah, raid10 is better. David, do you have any data comparing the two scenarios? Do you see flaws in my logic? > Stay away from RAID5 unless you have a good > controller with a battery backed cache. Even then, performance won't be great, especially on random small writes (look up the "RAID5 write hole" and "read-modify-write" to understand why). danno ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ BackupPC-users mailing list BackupPC-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/backuppc-users http://backuppc.sourceforge.net/
