Using Dependencies to Assert Application and Network Configuration

Craig E. Wills
Associate Professor of Computer Science,
Worcester Polytechnic Inistitute

March 10, 2006
11:00 a.m. - 12:00 p.m.
Fuller Labs 320

This work seeks to better integrate the management of applications and network infrastructure. The approach is to identify the dependencies among computing devices, network elements and applications through passive observation of traffic flow using a tool such as NetFlow.

Using the idea of application dependencies as a starting point, the key idea of this work is that a derived graph of dependencies between hosts and applications can be used as a basis to make assertions about application configuration, performance and security in the network. To illustrate the capabilities of this approach we have taken initial steps to define an assertion language for specifying what is expected and not expected in the network in terms of dependencies between applications and computing devices. We provide a number of illustrative examples of assertions to describe the assertion language and the type of dependency expectations we envision it can capture.

This work is attractive because it takes advantage of existing network flow data for a low-cost, yet powerful approach to verify the correct operation of applications in a network. Built on the identification of dependencies between applications and computing devices, assertions can be made that allow expected traffic in the network to be confirmed and allow unexpected traffic to be flagged.

Craig Wills is an Associate Professor of Computer Science at WPI, and started at WPI in 1990. He has a Ph.D. in Computer Science from Purdue University. His work experience includes AT&T Bell Laboratories in New Jersey. Professor Wills is interested in issues concerning a user's access to a computer system, particularly when the system is a distributed network of machines. His current work primarily involves Web performance and caching.

Host: Michael Gennert

Refreshments will be served.