Unix RPC programming

Question

I needed some help with ONC RPC programming.

My task is to create a two-tier client-server architecture wherein one main server (something like a directory) keeps a track of level-two servers and acts as a lookup; the level-two servers expose some trivial functions, and finally, the clients for level-two servers.

The clients ask the directory where a server is located, and then communicate with it. Using RPCGEN, we can create a pair of client-server code; however, the clients in this case need to have stubs for the directory as well as the level-two functions. Being a newbie to RPC, I'm having trouble conceptualizing the way I should code this. How can I call a function from a different server if a client is generated using a different IDL?

Answer 1

I suppose you could check out NFS 4.1, which also uses ONC RPC and supports parallel NFS (pNFS), where the client first contacts a metadata server, which then gives the client info about the data servers that the client then contacts to stripe data over.

You can check out efforts to implement pNFS for Linux here.