5/16 paper evals

Yod (h13nguye@ieng9.ucsd.edu)
Tue, 16 May 2000 07:49:19 -0700 (PDT)

Henry H. Nguyen
h13nguye@ucsd.edu
(858) 587 - 7046
Title: Performing Remote Operations Efficiently on a Local compuer Network

This paper discusses efficient communication techniques for very high
speed local networks, specifically the Xerox Alto computers. The
differences between the proposed communication subsystem describes in
this paper and the typical network communication subsystems are: the
designers of this system advocate an integrated implementation approach
rather than the layered approach, because the crossing of layer boundaries
results in decreased efficiency. Secondly, more emphasis on ensuring
that the functions provided by the communicaiton subsystem are suitable
for implementing high lebel primitives.

The paper went on to descibe the communication model called the remote
reference/remote operation model in which a taxonomy of communication
primitibes is defined. The argument is that this communication
subsystem can provide powerful primitibes yet be implemented efficiently.
These primitives includes: Utility of primities, support for reliability,
potential for protocol simplification, and protential for efficient
implementation.

In another section, this paper presents the implementation considerations
for this subsystem, and also exemplifies a highly reliable communication
primitive that it could provide. The performance studies provided in
this paper demonstrates how much faster and efficient, which are
questionable, that the communication subsystem under this design is
better than that of other communication subsystems. Title: Implementing Remote Procedure Calls

This paper addresses the issue of Remote procedure calls(RPC). It
describes about the overall structure of our RPC mechanism, including
facilities for binding RPC clients, the transport level communication
protocol, and some performance measurements.

The advantages or attractive aspects of using RPC includes clean and
simple semantics which make it easier to build distributed computations,
and procedure calls seem simple enough for the communication to be
quite rapid and efficient. But the best part is that of the generality
notion, RPC behave the same as procedure calls that are a well-known
and well-understood mechanism for transfer of control and data within
a program running on a single computer.

Some of the major issues facing the designer of RPC facility include:
the precise semantics of a call in the presence of machine and commu-
nication failure; the semantics of address-containing arguments in the
absence of a shared address spaace; integration of remote calls into
existing programing systems; binding between caller and the callee;
and how to provide data integrity and security in an open communication.

Then the paper goes on to describe some of these issues in greater
length, and provides some of the decisions chose by the designers of
the system. It also provides run through of simple procedure calls and
complicated calls. Security, exception handing, and performance were
also covered. According to the performance chart from the paper, one
of the conclusions stated in the paper is that for tranfers in one
direction, protocols other than RPC have an advantage, since they can
transmit fewer packets in the other direction.