Conductor: Distributed Adaptation for Complex Networks

˜ Computer networks are becoming more complex and diverse. Increasingly, an end-to-end connection will traverse several links with orders of magnitude differences in characteristics such as bandwidth, latency, error rate, jitter, and monetary cost. At the same time, most applications assume a level of network characteristics below which they either provide no service, or service at a cost higher than the user is willing to pay. Conductor is an adaptation framework that moves the responsibility for network complexity out of the application and into the network. Conductor allows an application’s use of the network to be tailored to the needs of the user in a manner that is transparent to the application. Conductor allows adaptation to be distributed to the points in the network where it is required. Further, Conductor allows arbitrary adaptations without compromising reliability. Finally, Conductor requires minimal changes to existing systems. All of these characteristics are incorporated into a single, integrated framework. Conductor has been shown to be effective in important classes of problems in mobile computing, where network complexity is most prevalent today. This paper describes the Conductor architecture and presents experimental results indicating that transparent, distributed adaptation can significantly improve the user’s experience in complex networks.