Red Hat Knowledgebase

Release Found: Red Hat Enterprise Linux 3

Symptom:
You are aware generally that there are differences but are not sure of the specifics. You would like to know what some of the specific differences are with regards to NFS performance.

Solution:
Think of the Linux kernel as being like an onion with concentric layers encompassing sub-layers. One of these layers is called the Virtual File System (or VFS) layer. The VFS layer abstracts filesystem access from applications and the rest of the kernel.

Within the VFS layer, is the code which manages NFS. This code is separated within the layer, from other code used to access physical disks and other storage devices.

With Red Hat Enterprise Linux 3, from a client application perspective, I/O over NFS is only "talking" to the VFS layer within the kernel. The NFS code within the VFS layer then is responsible for passing the data to the NFS server.

A Red Hat Enterprise Linux 3 NFS server will receive the data by way of the Portmap NFS daemon process. This process will interface with the VFS layer within the server's kernel. In this case however, it is likely that the VFS layer will further send the data to the code responsible for managing the physical storage device.

Therefore, NFS client applications will benefit from any optimizations which apply to the NFS client code within the VFS layer. The same is true server side, though you also have to consider the performance of the server's physical storage code within VFS.

Generally speaking, the largest performance improvements with NFS can be had by optimizating network communications. For further and more specific information on this, see the relevant section in the NFS HOWTO guide.