Mike Wright wrote:
Bill Davidsen wrote:
Alan Cox wrote:
The raw read speed on a RAID1 array is the speed of a single drive.
The
No. A smart raid1 driver will use both source disks. The raw write speed
is that of the slowest drive (plus overhead).
Does this mean that the Linux software RAID is not smart, or there's a
way to make it actually do this? Current FC[678] doesn't seem to be
"smart."
RAID 10 gives you better write throughput which on a modern PC normally
means that with PCI Express your bridge/memory bandwidth becomes the
limit.
This is what I measure running an E6600 CPU and 3xSeagate 320 with
Recent FC7 kernel. All reads and writes to the raw array using dd, 1MB
buffer, 1GB i/o to/from /dev/{zero,null} for raw speed. Units are
MB/s, 64k chunks, speed as reported by dd.
RAID lvl read write
0 110 143
1 52.1 49.5
10 79.6 76.3
10f2 145 64.5
raw one disk 53.5 54.7
Sorry to butt in but what does 10f2 mean/stand for?
Array creates with two "far" copies. See "man mdadm" and search the "-p"
option for near, far, and offset disk layout. The "far" layout has been
giving me great read speeds, which is (in most cases) more important
than write, since writes can be cached but when the program is waiting
for data it stops. Readahead isn't a substitute for throughput if you
read BIG data.
--
Bill Davidsen <davidsen@xxxxxxx>
"We have more to fear from the bungling of the incompetent than from
the machinations of the wicked." - from Slashdot
