While active networking provides tremendous benefits, it also adds to the complexity of a network. The additional complexity makes network and systems management a challenging and interesting problem, because it is a problem in which distributed computing can now more easily and rapidly be brought to bear. Distributed computing algorithms can be more easily implemented and more quickly deployed in an active network.

The goal of this research is to develop active networks that are capable of predicting their own behavior and to use this capability to develop predictive active network management. This research concentrates on the development of distributed computing techniques by means of the design and analysis of an algorithm referred to as "Active Virtual Network Management Prediction, AVNMP." The selected publications are related to this technology. Visit Information Assurance and Fault Tolerant Networking for continuing work in this area. You can also join the Active Virtual Network Management Prediction email list for discussion of Active Network Management and the AVNMP (Active Virtual Network Management Prediction) code.

The Atropos Toolkit, developed from the AVNMP concept and code (which can be downloaded from the button on the left), allows experimentation with predictive capability within a network while the network is operating. This might be best described as 'in vitro' prediction experimentation whose purpose is to address a severe limitation in state-of-the-art network management: current management techniques are reactive. The toolkit is an active application that executes within a network assuming an overlay active network exists (included as part of the Atropos Toolkit).

Active networking provides a framework in which executable code within data packets executes upon intermediate network nodes. The Atropos Toolkit provides an infrastructure that maintains state and enforces event causality easing the development of numerous, small, and predictive network component models. User-defined algorithms, injected into the network, allow system state to be predicted and efficiently propagated throughout the network. These algorithms enable operation in real time simultaneously with a continuously projected and refined future state.

The Atropos Toolkit allows experimentation with prediction algorithm parameters including tradeoffs among prediction accuracy, computational complexity, memory size, bandwidth, and projection window length (sliding window into the future). Thus Atropos facilitates experimentation in a distributed, active, and truly proactive management environment. Currently Atropos is being used in the study of Algorithmic Information Theory, including Kolmogorov-Chaitin Complexity estimation, for determining and optimizing the amount of code versus data within active packets. In recent versions, load and processor usage prediction applications have been experimentally validated using the Atropos Toolkit.

Libraries implementing an active overlay network and management graphing capabilities are included. There is an Atropos mailing list for users to provide feedback and limited support. Users can join the mailing list via the Download link to the left of this text.

Approach

Active Virtual Network Management Prediction

Consider predictions for network load, capacity, and configuration
Utilized in both mobile and fixed environments

Local Predictors at Active Network Nodes

Simple predictors across network, devices for process measurement
Flexibility to deal with heterogeneous traffic

Dynamic Message Priority*

Utilize bandwidth in near-optimal manner
Brokerage system for packet forwarding through active nodes

Features

Advantages/impact

Intermediate network elements predict their own behavior.
Better integration of communications and computation within an active network.
Equipment vendors can provide active logical processes to accomplish prediction for devices.
Better utilization of the network and enhanced quality of service for the users.

Publications

Books

Active Networks and Active Network Management: A Proactive Management Framework by Stephen F. Bush and Amit Kulkarni, Kluwer Academic/Plenum Publishers, New York, Boston, Dordrecht, London, Moscow, 2001, 196 pp. Hardbound, ISBN 0-306-46560-4

Papers

In-Line Network Management Prediction (TEXT format) (HTML format) (XML format) IETF Internet-Draft
Active Virtual Network Management Prediction: Complexity as a Framework for Prediction, Optimization, and Assurance by Stephen F. Bush, Proceedings of the 2002 DARPA Active Networks Conference and Exposition (DANCE 2002), IEEE Computer Society Press, pp. 534-553, ISBN 0-7695-1564-9, May 29-30, 2002, San Francisco, California, USA.
Predicting and Controlling Resource Usage in a Heterogeneous Active Network by Virginie Galtier and Kevin L. Mills (NIST) and Yannick Carlinet and Stephen F. Bush and Amit B. Kulkarni, Proceedings of the Third International Workshop on Active Middleware Services, August 6 2001, pp. 35-44.
Active Virtual Network Management Prediction by Stephen F. Bush, Trans-European Research and Education Networking Association (TERENA) 2001, Antalya, Turkey, May 14-17, 2001
Predicting Resource Demand in Heterogeneous Active Networks by Virginie Galtier and Kevin L. Mills (NIST) and Yannick Carlinet and Stephen F. Bush and Amit Kulkarni, MILCOM 2001, McLean, VA, October 28-31

Detecting Distributed Denial-of-Service Attacks using Kolmogorov Complexity Metrics by Amit B. Kulkarni, Stephen F. Bush, and Scott C. Evans, GE CRD Technical Report 2001CRD176, December 2001.
Predicting and Controlling Resource Usage in an Active Network by Stephen F. Bush and Amit Kulkarni (GE), Virginie Galtier and Yannick Carlinet and Kevin L. Mills (NIST), Livio Ricciulli (Metanetworks), DARPA Active Networks PI Meeting, December 6-9, 2000, Atlanta, GA
Active Jitter Control by Stephen F. Bush and Amit Kulkarni and Scott Evans and Luiz Galup, 7th International IS&N Conference, Intelligence in Services and Networks (ISN) '00, February 23-25, 2000, Kavouri, Athens, Greece, (PDF) Presentation (PPT)
Islands of Near-Perfect Self-Prediction by Stephen F. Bush, Proceedings of the International Conference on Virtual Worlds and Simulation Conference 2000: 2000 SCS Western Multi-Conference, San Diego, SCS (2000), ISBN 1-56555-192-3, Vol. 32, No. 2, pp. 139-144, International Conference on Virtual Worlds and Simulation (VWSIM 2000), January 2000 (PDF) Presentation (PPT) (HTML)
Active Virtual Network Management Prediction by Stephen F. Bush, 13th Workshop on Parallel and Distributed Simulation, May 1-4, 1999, Atlanta, Georgia, USA, PADS '99 (PS)
On Respecting Interdependence Between Queuing Policy and Message Value by John E. Hershey and Stephen F. Bush, Nov 99, CRD Technical Report 99CRD151 (PDF)
Active Virtual Network Management Prediction by Stephen F. Bush, Nov 99 DARPA Next Generation Internet Meeting Presentation (PPT or HTML)
Active Virtual Network Management Prediction by Stephen F. Bush, Oct 98 DARPA Next Generation Internet Meeting Presentation (PDF or PostScript).