Work In Progress

I've just noticed that in a virtualized environment, where roundtrips are more expensive, read speed differences between a page-aligned partition and a non page-aligned one are even more noticeable:

Non-aligned partition (size in sectors: (41929649-63)+1=41929587):

none:~# fdisk -l -u /dev/sdb

Disk /dev/sdb: 21.4 GB, 21474836480 bytes
255 heads, 63 sectors/track, 2610 cylinders, total 41943040 sectors
Units = sectors of 1 x 512 = 512 bytes
Disk identifier: 0x93dbf2be

Device Boot Start End Blocks Id System
/dev/sdb1 63 41929649 20964793+ 83 Linux

none:~# dd if=/dev/sdb1 of=/dev/null bs=1M count=100 skip=0
100+0 records in
100+0 records out
104857600 bytes (105 MB) copied, 3.80593 s, 27.6 MB/s

none:~# dd if=/dev/sdb1 of=/dev/null bs=1M count=100 skip=100
100+0 records in
100+0 records out
104857600 bytes (105 MB) copied, 3.82304 s, 27.4 MB/s

none:~# dd if=/dev/sdb1 of=/dev/null bs=1M count=100 skip=200
100+0 records in
100+0 records out
104857600 bytes (105 MB) copied, 3.83713 s, 27.3 MB/s
Page-aligned partition (size in sectors: (41929654-63)+1=41929592) :

none:~# fdisk -l -u /dev/sdb

Disk /dev/sdb: 21.4 GB, 21474836480 bytes
255 heads, 63 sectors/track, 2610 cylinders, total 41943040 sectors
Units = sectors of 1 x 512 = 512 bytes
Disk identifier: 0x93dbf2be

Device Boot Start End Blocks Id System
/dev/sdb1 63 41929654 20964796 83 Linux


none:~# dd if=/dev/sdb1 of=/dev/null bs=1M count=100 skip=0
100+0 records in
100+0 records out
104857600 bytes (105 MB) copied, 0.618494 s, 170 MB/s

none:~# dd if=/dev/sdb1 of=/dev/null bs=1M count=100 skip=100
100+0 records in
100+0 records out
104857600 bytes (105 MB) copied, 0.582423 s, 180 MB/s

none:~# dd if=/dev/sdb1 of=/dev/null bs=1M count=100 skip=200
100+0 records in
100+0 records out
104857600 bytes (105 MB) copied, 0.589081 s, 178 MB/s
NOTE: In both tests, the data is cached in the host (so there is no real I/O to disks but traffic through virtualization pipes), but clean in the guest.

While running some tests in a client's environment, we've noticed reading from a partition of a multipath device was considerably slower than reading from its parent node:

[root@none]# dd if=mpath4 of=/dev/null bs=1M count=1024
1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB) copied, 8.92711 seconds, 120 MB/s

[root@none]# dd if=mpath4p1 of=/dev/null bs=1M count=1024 skip=1024
1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB) copied, 17.5965 seconds, 61.0 MB/s

We asked to client support of a well-known GNU+Linux vendor, and they indicated that this behavior was "expected", since this kind of partitions were created by stacking a dm-linear device over the original multipath node. I wasn't satisfied by this answer, since AFAIK dm-linear only did a simple transposition of the original request over an specified offset (the beginning of the partition), so I decided to investigate a bit further on my own.