Remote Procedure Calls

Question 1

(7 steps)

What are the steps involved in Remote Procedure Calls (RPC)?

Answer 1

1. The calling system creates an RPC call and the client RPC service:
   
   • Creates a unique ID for the call
   • Marshals the argument into some suitable format
   • Sets a timer
2. Data is passed as a message through the lower levels
3. The server RPC service:
   
   • Unmarshals the arguments
   • Makes a note of the ID for this RPC call
4. The called system calls and executes the required method
5. The server RPC service:
   
   • Marshals the return arguments, ready for transport
   • Sets a timer
6. The client RPC service:
   
   • Unmarshals the return arguments
   • Disables the client RPC service timer
   • Sends an acknowledgement to the server RPC service that the RPC with the specified ID has arrived safely
7. The server RPC service timer is disabled

Question 2

(5 answers)

Describe how RPC timers work and how they provide robustness.

Answer 2

• The timers will expire if there is a problem
• Failure of the client is not noticed until no acknowledgement arrives, meaning it may not be possible to undo certain state changes at the server side
• Failure of the server will always be noticed due to the first timer expiring and when this happens, the client may decide to repeat the call using the same ID
• If the called system detects a repeated RPC call, it can just resend the message
• If the called system does not detect a repeated RPC call, this may result in changes to the data being made twice