Rules for max_coalesce and queue_depth for ssa RAID-5 disk array

 This is the maximum number of bytes which the SSA disk device
 driver attempts to transfer to or from an SSA logical disk in a
 single operation. The default value is appropriate for most
 environments. For applications that perform very long sequential
 write operations, there are performance benefits in writing data
 in blocks of 64KB times the number of disks in the array minus
 one (these are known as full-stride writes times the number of
 disks in the array minus one, or to some multiple of this number.

 Specifies the maximum number of commands that the SSA disk device
 driver dispatches for a single disk drive for an hdisk. You can use
 the chdev command to modify this attribute. The default value is
 correct for normal operating conditions.

max_coalesce is (N-1)*64k
queue_depth is from 2*N to 3*N

N = number of disk in the ssa raid-5 array
0x10000 is 64K

N max_coalesce queue_depth
3 disks 0x20000 6 to 9 4 disks 0x30000 8 to 12 5 disks 0x40000 10 to 15 6 disks 0x50000 12 to 18 7 disks 0x60000 14 to 21 8 disks 0x70000 16 to 24 9 disks 0x80000 18 to 27 10 disks 0x90000 20 to 30 11 disks 0xA0000 ( not 0X100000) 22 to 33 12 disks 0xB0000 ( not 0X110000) 24 to 36 13 disks 0xC0000 ( not 0X120000) 26 to 39 14 disks 0xD0000 ( not 0X130000) 28 to 42 15 disks 0xE0000 ( not 0X140000) 30 to 45

To change max_coalesce or queue_depth: 1) umount all filesystems on the raid-5 vg 2) varyoffvg the volume group. 3) Use smitty chgssardsk or chdev Example: chdev -l hdisk2 -a max_coalesce=0xb0000 Example: chdev -l hdisk2 -a queue_depth=3 4) to verify the change: lsattr -El hdiskX
Created: Nov 2003 by Bruno Croft