NSF:
WebTP (1999-2002): A new protocol for web
browsing.
WebTP
is a project envisioned at the EECS department at University of
California, Berkeley to take on the problems of the current world wide web and
formulate a comprehensive protocol for the next generation of the internet. The
primary purpose of this protocol is to bring the user into the transaction
loop, and to optimize any web transfer to ensure maximum satisfaction for the
user.
The
project embodies four primary design goals:

NSF: Online
Statistics of ATM Networks (1996-1999)
Providing guaranteed Quality of Service (QoS) on the Internet requires that the network be able to measure traffic and negotiate access to network resources. As QoS-demanding applications are more widely used, the network will eventually need to provide users with statistical guarantees of QoS, such as the loss rate, average packet delay, and delay jitter. In this paper, we present a collection of estimators that use on-line measurements of user traffic to provide call admission for a single-server first-come-first-served (FCFS) queue. Our approach infers the QoS parameters from measurements of the buffer occupancy at packet arrival times. A nonlinear optimization routine is used to fit these measurements to a buffer occupancy distribution, from which we can infer the loss rate and delay statistics. Furthermore, we extend these estimators to call admission by combining our fitting algorithm with virtual queuing systems to estimate the additional capacity within the queue. As a result, we are able to improve the performance of call admission algorithms by providing fast and reliable estimators of QoS.

NSF: Control of
High-Speed Networks (1993-1996)
The
objective of this research is to clarify the possibilities and methods of
control of high-speed networks. These networks include not only the
asynchronous transfer mode networks (ATM) but also high-speed networks based on
the TCP/IP protocols. The high-speed creates new problems related to the large
rate-delay product that tends to make traditional window-based congestion
control mechanisms ineffective. Our work has led to the following concepts:
1.
Effective bandwidth for general sources;
2.
Decoupling bandwidth;
3.
Sufficient condition for low cell loss ratio;
4.
Fast estimators for bandwidth required per connection;
5.
Admission control and routing strategies for multiclass
traffic;
Loss in utilization caused by on-line estimation.
Last Edited 11/16/2004