papers for Tuesday

John-Paul Fryckman (fryckman@SDSC.EDU)
Tue, 16 May 2000 04:49:46 -0700 (PDT)

Performing Remote Operations on a Local Computer Network

This paper centers around a communication model that is idealized
as remote procedure calls--that is the ability to issue commands
on another node. The fact that these calls are conducted on remote
machines are abstracted from the user as much as possible. Furthermore,
this paper addressed some of the major performance concerns on high
speed networks of the time.

Several key points are worth mentioning. One is that they did not
use a layered approach, but instead an integrated the entire
communication system. Another design goal is that the interface is
suitable for high-level languages. Some the implementation concerns
involve minimizing message generation and simplification of the

The paper was very detailed in the taxonomy of the RPC communication
structure including how it was implemented and what protocols were
used. It was very readable and thought provoking.

Implementing Remote Procedure Calls

As the title of this paper indicates, it's all about implementing
RPCs from the design specification on down to the protocols used.
Given location abstraction, that is to say that as far as the programmer
is concerned the RPC just acts as if it is running on the local
machine, they provide a simple means for distributing work and data between

The user inserts a stub in his code that in turn calls the underlying
RPC mechanism which is the RPCRuntime package, and finally the server
stub and the server itself. The user only specifies the interface
and not the instance on RPCs (this is done dynamically via the software).
They also use low-level protocol that bypasses the TCP/IP stack.
Another key feature is that they try to reduce the overall message
count by piggybacking receipt acknowledgements with the actual data
and other message mergers.

This paper was easy to read and addressed the key points about RPC
that any paper on the subject should address.