CSR: DMSS, TM: Quantifying the Impact of Latency on Network-Interactive Computing

Carnegie Mellon University, Pittsburgh PA

Investigators

Abstract

There is growing academic and commercial interest in models of interactive computing where code segments that lie in the critical path of user interaction are split across a network boundary. We use the term "Network-Interactive Computing" to characterize such models. Thin client computing and Ajax-based computing are two examples. Unfortunately, introducing network delays into the critical path of user interaction can degrade the crispness of system response, and hurt usability. Today, we lack the scientific framework to quantify the impact of latency on network-interactive computing. This research will establish the scientific underpinnings of network-interactive computing with respect to performance and usability. Our work will span three areas: (a) Development of metrics and benchmarks; (b) Calibration of metrics against user experience; and (c) Classification of real-world applications based on their user interaction characteristics. We will develop the intellectual tools and evaluation techniques to help answer questions such as the following: ``Will a legacy interactive application that performs well on a local computer degrade unacceptably in a specified network-interactive computing environment?''; ``When designing a new environment for network-interactive computing, can one predict which applications will perform acceptably?''; and, ``Can one develop a taxonomy of applications that allows us to easily understand and predict the impact of networking changes on interactive performance?'' Our goal is to establish network-interactive computing on a sound, well-validated and replicatable experimental foundation.