raid 5 : partitioned array VS lvm

[Fab <azertyzero@free.fr>]

Hello, 

Lately I tried some different configurations (lvm & partition) to divide my raid
5 array. 

I noticed that partitioning the array was resulting in far better performances
than using lvm. 

I have the following configuration : 

-------------------------
# mdadm --detail /dev/md0

/dev/md0:
        Version : 00.90.03
  Creation Time : Sun Sep 23 12:29:14 2007
     Raid Level : raid5
     Array Size : 1435825152 (1369.31 GiB 1470.28 GB)
    Device Size : 478608384 (456.44 GiB 490.09 GB)
   Raid Devices : 4
  Total Devices : 4
Preferred Minor : 0
    Persistence : Superblock is persistent

    Update Time : Thu Oct 18 22:42:12 2007
          State : clean
 Active Devices : 4
Working Devices : 4
 Failed Devices : 0
  Spare Devices : 0

         Layout : left-symmetric
     Chunk Size : 128K

           UUID : 0d69d64b:89ec4fb0:e4564657:5e4ed718
         Events : 0.80

    Number   Major   Minor   RaidDevice State
       0       8        3        0      active sync   /dev/sda3
       1       8       19        1      active sync   /dev/sdb3
       2       8       35        2      active sync   /dev/sdc3
       3       8       51        3      active sync   /dev/sdd3
-------------------------


I tried a lot of different configurations when I made my logical volumes,
but none worked better. I don't know could be wrong. 

Here are my benchmarks :

-------------------------
# dd if=/dev/md0 of=/dev/null bs=100M count=20
20+0 records in
20+0 records out
2097152000 bytes (2.1 GB) copied, 9.58494 seconds, 219 MB/s

dd if=/dev/raidvol1/data1 of=/dev/null bs=100M count=20
20+0 records in
20+0 records out
2097152000 bytes (2.1 GB) copied, 12.396 seconds, 169 MB/s

# dd if=/dev/raidvol1/data2 of=/dev/null bs=100M count=20
20+0 records in
20+0 records out
2097152000 bytes (2.1 GB) copied, 13.4758 seconds, 156 MB/s

# dd if=/dev/raidvol1/home1 of=/dev/null bs=100M count=20
20+0 records in
20+0 records out
2097152000 bytes (2.1 GB) copied, 17.7541 seconds, 118 MB/s
-------------------------

If I use a partitioned array, then I get the best performances (same than
/dev/md0 : about 220MB/s). With LVM, it has been impossible to get more than
170MB/s in "any" configuration. 

I'm not quite sure why I'm getting these results. Any ideas about
what could be wrong, or what to look for? 

Thanks in advance, 

Fabien