This project is designing new application and network transport protocols as well as improved workload characterizations and models for evaluating the protocols. In the domain of workload characterization, the project is evaluating and extending a new method for network traffic characterization called Bayesian Block (BB) analysis that is significantly more accurate than previous methods for estimating the time-varying rate of bursty event processes, such as packet losses in a network flow, packet or flow arrivals to a gateway or end host in the network, or requests arriving to a network service. Anticipated results include (a) more precise and accurate characterizations of these network event processes, (b) more precise and realistic synthetic workload models for network flows, and (c) extensions to the BB analysis method to improve its accuracy. The project is also investigating more precise and accurate peer-to-peer workload characterizations that can be used in the design of improved peer-to-peer applications. In the domain of protocol design, the project is investigating (a) new near-optimal download protocols from multiple (peer-to-peer) servers, and (b) exploratory design of new transport protocols. The new transport protocols employ forward error correction to recover from packet loss as well as BB analysis at the receiver to determine when to send feedback to the sender regarding observed packet loss rate for use in regulating the rate the packets are sent to the receiver.
