Robust Routing in Dynamic Networks. Part 2.
Abstract
The goal of various control schemes used in a network is to assure a continued good performance as the load and other network characteristics change. Many of the flow and congestion control schemes are designed with assumed network dynamics. In order to get a better understanding of the network dynamics a tool was designed that can be used to monitor the detailed end-to-end behavior of a path of Internet for extended periods of time. The transit times and loss characteristics are observed by probe packets that are sent at regular intervals. Traditionally the flow control techniques used in the networks have been designed assuming a stochastic behavior for the load as well as the components of the network. In this research the approach has been to model a network and its connections using deterministic modeling techniques as far as possible, invoking stochastic assumptions only when deterministic assumptions cannot be justified or the lack of information makes it essential. A new version of Maryland Routing Simulator (MaRS) with improved front end has been completed and distributed through FTP. It has been copied by over 250 sites. Algorithms, Distribution, Dynamics.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 1994
- Accession Number
- ADA282396
Entities
People
- Ashok Agrawala
Organizations
- University of Maryland