A Novel Server-Based Measurement Infrastructure for Enterprise Networks

Date of Submission: 
October 16, 1998
Report Number: 
98-031
Abstract: 
Despite many proposals on building an Internet measurement infrastructure from the research community, we believe that it will not be in the near future for such an infrastructure to be fully deployed and operational, due to both the scale and the complexity of the Internet. In the meantime, to take advantage of the enormous benefits that can be offered from such a measurement infrastructure, we propose to build a measurement infrastructure on enterprise network servers with the specific aim of optimal server selection and QoS monitoring. We present the architecture and algorithm used in our server-based measurement infrastructure. A novel design feature of our measurement infrastructure is that we are able to generate most of the performance metrics of our interest for each user traffic flow at a server through non-intrusive measurement technique. In particular, we build a data collection and performance metrics generation software called Woodpecker, which is capable of obtaining most of the traffic metrics on-line. On the other hand, intrusive measurement is used only to complement the limitations of non-intrusive measurements in Woodpecker and its use is kept at minimal.