nfs acting weird for different size partition

[abhig]

Hi guys,
I am setting up a server and trying to access its partition over the nfs on a client machine.
Our server has 4 1.5TB disks and the idea was to make a RAID 5 out of it.
Now here is the problem:
When I used only 1 disk that is nfs partition was of 1.5TB, I got fine speed(30MBps), but with multiple disks whether I am making 3TB (Raid 10) or 4.5TB (Raid 5), I get a really poor speed of 5-6MBps. I/O just waits
But for all above cases, disk I/O locally works fine.
Does anyone knows any NFS limitation that is causing this problem?
Thanks,
Abhi.

[Rubberman]

Is this hardware or software RAID? Also, is this read speed, write speed, or a mix? How are you testing?

[abhig]

Hi Rubberman,
This behavior was with both soft as well as hard RAID.
The speed is writing speed. Reading speed was just fine.
On server:
/etc/exports
/mnt/bioexp 172.16.0.0/16(sync,rw,no_root_squash)

On client:
/etc/fstab
fs-bioexp:/mnt/bioexp /mnt/bioexp nfs rsize=8192,wsize=8192,timeo=14,intr 0 0

On client when I run dd command as:
dd if=/dev/zero of=/mnt/bioexp/1gb bs=1024 count 1000000
the time taken to run this command is 200secs

But on server locally, the dd command runs in about 4-5secs

Thanks,
Abhi.

[Rubberman]

Most RAID implementations (especially RAID 5/10) slow down considerably for write operations. This is likely what you are experiencing. After all, the array has to have the parity bits computed and written as well as the data, and parts of the data has to be written to multiple drives (striping). Reads will generally be as fast, or faster than a single drive or logical volume set, but as you see, writes are another story altogether.

[abhig]

But even RAID 0 is behaving the same way. Logically RAID 0 should be faster than single disk.