Project Proposal
Admission Control in Xbind
Nemo Semret
What is Xbind?
"ATM LAN interoperability has become a major challenge to the
broadband networking community as the various switches available
on the market are only interoperable on the "cell" and not on the
"signalling" level. The signalling technology considered for
deployment reminds us of the telephone systems of the 1910s when
an operator had to connect two users by plugging a cable in the
right jacks. Sadly enough, the situation is not different today
when permanent virtual circuits are set up also by hand. The only
difference is that the operator sits at a computer terminal and
establishes a permanent virtual circuit by typing commands on a
keyboard. The basic user interface (also known as Q.93b/Q2931)
essentially dates back to the late sixties and it is based on the
assumption that the costumer premises equipment is dumb, whereas
the network is intelligent. In a world of PCs and workstations
these assumptions are no longer valid. A change is in order that
requires a concentrated effort before the hardware manufacturers
lock themselves into old technology."
[excerpt from Overview of Xbind]
Xbind
is an implementation of a binding architecture for ATM networks. It's
purpose is to provide hardware independent signalling. In layman's
terms, that means that setting up a "call" (coming from any multimedia
application on the network) should not depend on the hardware of the
involved hosts and switches along the route, which are likely to be
made by different manufacturers.
A version of Xbind is currently up and and running in the Multimedia
Networking Laboratory. The underlying platform is the CORBA
standard for distributed systems, as defined by the OMG (see lecture 1
of this course).
Xbind has an HTML-based front-end. Virtual Circuits for actual
applications (e.g. videoconferencing) are set-up, monitored and
brought down via interactive Web pages.
Admission Control
ATM networks are expected to handle a mix of several different types
of traffic. Data, video, and voice streams with widely varying
bitrates and degrees of burstiness may be transmitted over the same
link. Furthermore, the different types of traffic have different
Quality of Service (QoS) requirements. For example, for a file
transfer, end to end delays are not critical, but zero-percent cell
loss is a requirement. On the other hand, for a real-time video
stream, the delay has to be within certain strict bounds, but
occasional loss of a few frames may be tolerated.
Thus before a virtual circuit is setup, the network has to decide if
there are sufficient resources to guarantee the call's QoS
requirements, based on the characteristics of the incoming call and
the state of the network. This decision is called Admission Control.
Project Objective
Currently there is no Admission Control on Xbind. I propose to
implement a set of admission control algorithms of varying degrees of
complexity. These will be useful as functional elements of Xbind, to
be used in VC setup for multimedia applications. It will also provide
the means for an experimental complement to our on-going theoretical
research in admission control and the fundamentals mathematical and
conceptual issues in resource allocation/sharing in multimedia
networks.
Specifications
The implementation will be in the Java
language. This project will allow us to measure the applicablity
of Java for programming that are
The first is an explicitly stated design objective of the
language. However, if Java it is to become a widespread standard, it's
suitability for "number crunching" will become an important
question. Even if that is not one of it's strengths, Java programmers
will want to do it, at least up to a certain point. Our admission
control algorithms will range from simple to highly computational, and
thus provide means of probing the limits of Java's practical
abilities.
To my homepage.
Wed Oct 4 18:08:55 EDT 1995