Internet Protocol (IP) Network Measurement, Characterization, Modeling, and Control for Self-Managed Networks
Abstract
IP network technology cannot continue on an ever-increasing course of technological complexity and yet require the kind of human intervention that is necessary today for network management. Networks must be self-managing. This can only be done by a system of measurement that copes with the dynamics of packet movement. This system must process the packet-level measurements into variables that characterize network behaviors, which then form the basis for control algorithms that react to the variables. Such packet level measurements can lead to characterization and control at the application layer, at the transport layer, at the network layer (including overlay networking), and in some cases at the link layer. The research documented in this report used tools of statistics, data mining and machine learning to (1) determine network variables derivable from the measurements that characterize network behavior; (2) develop models of the critical network variables that characterize performance, usage, security, and early onset of problems; and (3) develop automated control methods based on the variables.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 2005
- Accession Number
- ADA439560
Entities
People
- Bowei Xi
- Hui Chen
- Jin Cao
- William S. Cleveland
Organizations
- Purdue University