Introduction next up previous
Next: Latency profiles Up: Monitoring the Performance of Previous: Monitoring the Performance of

Introduction

We broadly define a wide area application (WAA) as an application that runs using public infrastructure and involves a federation of hundreds of servers and tens of thousands of clients, all with disparate characteristics, e.g., dissimilar connection bandwidth or network congestion en route from server to client. Several challenges must be addressed to adequately support WAA development in information-centric networks. First, it must provide a scalable methodology to monitor and predict end-to-end client-side performance. Second, it needs to validate resources on servers and monitor the staleness (obsolescence) of cached (replicated) resources. Third, it must differentiate delivery of services depending on application semantics or requirements.

In this paper, we address the first challenge of performance monitoring. We present Latency profiles (LPs) that model end-to-end latency (delay) experienced by a group of clients as they access digital content from repositories or content servers. Related networking research on IDMaps, points of congestion, and BGP routes support the hypothesis that one may identify network level characteristics (feature vectors) of paths between clients and servers that impact latency [3,6]. Based on this hypothesis, we further develop the concept of coverage of clusters of client-server pairs by an aggregate LP.


next up previous
Next: Latency profiles Up: Monitoring the Performance of Previous: Monitoring the Performance of