ylfink@water.waterloo.edu (ylfink) (03/29/88)
DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES
DISTRIBUTED SYSTEMS SEMINAR
- Wednesday, March 30, 1988
Mr. Sanjay Radia, a graduate student of this
department, will speak on ``Autonomy and Transparency
in Naming Communication End-Points in Distributed
Systems''.
TIME: 11:30 AM
ROOM: E2-3324
ABSTRACT
We address the problem of naming communication end-
points (processes in our case) in distributed systems.
Autonomy is important in our dynamic environment where
machines and networks dynamically connect and
disconnect, possibly to and from different systems,
while continuing to function with respect to local
resources. Coherence in the form of network transparent
interprocess communication is also important.
Global, fully qualified, hierarchical names are
commonly used for communication end-points; such names
can be freely exchanged across machine and network
boundaries. Although global names are useful for
achieving network transparency, they limit the autonomy
of machines and networks and hinder dynamic
connections.
We propose that, instead, partially qualified names be
used. A name is qualified only as far as needed---
machine and network identifiers are used only when
messages cross machine and network boundaries. A simple
rule for mapping names embedded inside messages is
presented. The scheme is extended using source routing
to permit interprocess communication between autonomous
domains that do not share a common name space. Our
solution supports a high degree of network transparency
without limiting autonomy.
Naming communication end-points is an essential step in
our research on naming and autonomy in distributed
systems. A high level name of a resource is typically
resolved to a low-level name of a communication end-
point through which the resource is accessed. In our
approach, autonomy and network transparency in naming
at a high level are based on autonomy and transparency
at the lower level.