Review of "Implementing Remote Procedure Calls"

From: Joanna Muench (joannam_at_spro.net)
Date: Sun Feb 01 2004 - 16:18:34 PST

  • Next message: Steve Arnold: "Review: Birrell & Nelson, Remote Procedure Calls"

    Birrell and Nelson (1984) describe a package supporting remote procedure
    calls across a network. While not the first attempt to implement RPC,
    the ideas contained in their paper form the basis of currently
    implemented RPC systems.

    The goals of their initial system were to make distributed computing
    easy, implement an efficient communication protocol and provide secure
    communication. To achieve the first goal, they made a fundamental design
    decision to make a RPC as similar as possible as a local procedure call.
    This principle strongly influence many design decisions, such as not
    having a time-out mechanism.

    Like any system, most decisions were about trade-offs. Communications
    are optimized for simple calls, cases where all the arguments fit into a
    single packet buffer. The communication protocol is arranged so that a
    new call cannot be initiated until the results of the preceding call
    have been received. While this may not be the optimal design for speed,
    it does remove the needs for additional consistency checks. Another
    trade-off is providing notification for every argument package. This
    creates a significant extra packets for arguments too large to fit in a
    single packet, but allows simpler and faster response for single packets
    which are presumed to be the majority of the traffic.

    While listed as an important feature, the authors deferred discussion of
    security to a later paper. I had hoped that would be the Grapevine paper
    we were assigned to read, unfortunately that doesn't seem to be the
    case.

    This was a very interesting paper to read, especially given the
    widespread use of RPC's today through tools such as CORBA. Unfortunately
    the performance results were all fairly theoretical, and I would be
    interested in reading about how the system actually performed under an
    active network and the size of packets that a system under full use
    would actually transmit.


  • Next message: Steve Arnold: "Review: Birrell & Nelson, Remote Procedure Calls"

    This archive was generated by hypermail 2.1.6 : Sun Feb 01 2004 - 16:14:58 PST