Improving the quality of real -time applications through path switching
Shu Tao · Scholarly Commons (University of Pennsylvania) · 2005
Although the current best-effort Internet cannot readily provide the service guarantees that real-time applications often require, path switching can potentially address this problem without resorting to the more complex network QoS mechanisms. The idea is to allow the application to dynamically select the path that carries its traffic, leveraging the robustness to performance variations provided by multiple connectivity options (e.g., multihoming or overlay networks). In this dissertation, we design and evaluate path switching mechanisms that embody this idea. In particular, we take an application perspective and devise a path switching scheme that is aimed at delivering improved quality to real-time applications. We first study the feasibility and performance benefits of path switching. Specifically, we investigate two common approaches offering path diversity---multihoming and overlay networks---and investigate their characteristics in the context of a representative wide area testbed. We explore the end-to-end delay and loss characteristics of different paths, and find that substantial improvements can potentially be achieved by path switching, especially in lowering end-to-end losses. Based on this assessment, we develop a simple path switching mechanism capable of realizing those performance improvements. Next, we investigate how to construct diverse paths with uncorrelated performance. A promising approach for acquiring path diversity is to leverage the capability of peer-to-peer systems, in which many nodes can act as relays for others, therefore offering a large number of candidate paths to select from. However, due to the sheer number of alternate path choices and the fact that not all choices are equally good, we need mechanisms to identify the relay nodes that yield good alternate paths. In this dissertation, we formulate and evaluate such mechanisms in the context of a large peer-to-peer system. We develop several heuristics that allow the selection of good alternate paths, relying only on the routing information available to the source. We validate the resulting solution by comparing it to a number of benchmarks, using both real network topology data and measurements on the PlanetLab testbed. Last, we assume the availability of sufficient path diversity and design path switching mechanisms to improve the quality of specific applications. As a first step, we present an online probing-based approach to estimate the loss performance of a network path, and extend this estimate to infer the performance that an application using the path would see. (Abstract shortened by UMI.)