Network Storage Protocols Discussions
I am in the process of replacing an older 3210 with a much newer Unit, it has more CPU,Memory,IO and Faster drives.
The 3210 is running 8.0.2P5 [7-Mode] the new unit is running 9.1P2 [Cluster-Mode].
There is a process that we run on FreeBSD servers with an NFS mount to the netapp that manipulates a bunch of small files.
That process on the 3210 roughly takes 30s. using the same mount options just changing the ip address that same process on the new Netapp takes 2-3 min.
Large file transfers appear to be faster on the new netapp.
This is the mount option we have used for years on the BSD Client. /vol/data nfs rw,tcp,intr,nfsv3,rsize=32768,wsize=32768,rdirplus,nolockd 0 0
Both Netapps are plugged into the same 10G switch and the Client is on the same subnet.
According to Harvest and Performance monitor the new netapp has almost no latency and very little impact to performance.
Seems like there is a difference between 7-mode and ClusterOntap in regards to NFS, there must be a parameter or setting that we could change to get the performance on par with the old system.
MTU is set correctly
And I did not see any dropped packets as a result of overflow.
How are your aggregates configured on each system? What type of disks? Flash pools? Flashcache?
All those factors come into play too - especially for small file operations.
see this article that appears to match your issue.