Autonomous agents for coordinated distributed parameterized heuristic routing in large dynamic communication networks

https://doi.org/10.1016/S0164-1212(00)00100-XGet rights and content

Abstract

Parameterized heuristics offers an elegant and powerful theoretical framework for design and analysis of autonomous adaptive traffic management agents in communication networks. Routing of messages in such networks presents a real-time instance of a multi-criterion optimization problem in a dynamic and uncertain environment. This paper describes the analysis of the properties of heuristic routing agents through a simulation study within a large network with grid topology. A formal analysis of the underlying principles is presented through the incremental design of a set of autonomous agents that realize heuristic decision functions that can be used to guide messages along a near-optimal (e.g., minimum delay) path in a large network. This paper carefully derives the properties of such heuristics under a set of simplifying assumptions about the network topology and load dynamics and identify the conditions under which they are guaranteed to route messages along an optimal path, so as to avoid hotspots in the load landscape of the network. The paper concludes with a discussion of the relevance of the theoretical results to the design of intelligent autonomous adaptive communication networks and an outline of some directions of future research.

Introduction

With the unprecedented growth in size and complexity of modern communication networks, the development of intelligent and adaptive approaches to system management (including such functions as routing, congestion control, traffic/load management, etc.) have assumed considerable theoretical as well as practical significance. Knowledge representation and heuristic techniques (Pearl, 1984) of artificial intelligence, decision-theoretic methods, as well as techniques of adaptive control offer a broad range of powerful tools for the design of intelligent, adaptive, and autonomous communication networks. This paper develops and analyzes heuristic decision functions in support of adaptive routing in large high-speed communication networks.

Routing (Bertsekas and Gallager, 1992) in a communication network refers to the task of propagating a message from its source towards its destination. For each message received, the routing algorithm at each node must select a neighboring node to which the message is to be sent. Such a routing algorithm may be required to meet a diverse set of often conflicting performance requirements (e.g., average message delay, network utilization, etc.), thus making it an instance of a multi-criterion optimization problem.

For a network node to be able to make an optimal routing decision, as dictated by the relevant performance criteria, it requires not only up-to-date and complete knowledge of the state of the entire network but also an accurate prediction of the network dynamics during propagation of the message through the network. This, however, is impossible unless the routing algorithm is capable of adapting to network state changes in almost real time.

Consequently, routing decisions in large communication networks are based on imprecise and uncertain knowledge of the current network state. This imprecision is a function of the network dynamics, the memory available for storage of network state information at each node, the frequency of, and propagation delay associated with, update of such state information. Thus, the routing decisions have to be based on knowledge of network state over a local neighborhood supplemented by a summary of the network state as viewed from a given node. Motivated by these considerations, a class of adaptive heuristic routing algorithms have been developed over the past few years (Mikler et al., 1997). Experiments demonstrate routing by autonomous routing agents that embody such algorithms displays several interesting properties including: automatic load balancing and message delay minimization. The results presented in this paper constitute a step toward the development of a theoretical framework for the design and the analysis of self-managing communication networks that are managed by interacting, proactive, and reactive, autonomous intelligent agents.

The rest of is paper is organized as follows. Section 2 briefly describes the Quo Vadis framework (Mikler et al., 1997, Mikler et al., 1998) for heuristic routing in large communication networks. Section 3 presents some of the results of simulation experiments which motivated the theoretical analysis presented in this paper. Section 4 presents the design and analysis of various routing heuristics with the emphasis on hotspot avoidance. Section 5 concludes with a discussion of the relevance and limitations of the main results and some directions for further research.

Section snippets

A framework for heuristic routing

Any intelligent traffic management mechanism capable of performing in a large communication environment must include an effective knowledge representation (KR) mechanism as well as an efficient knowledge acquisition (KA) engine, that minimizes the overhead that is associated with acquiring and maintaining network state information. In addition, adaptive decision making methods are needed which are designed to optimize the network performance.

The underlying framework for heuristic routing,

Properties of parameterized routing heuristics

A prototype implementation was used to conduct a number of experiments to explore the effects of the various parameters used in the framework. These experiments were conducted in simple regular m×n grid networks. We anticipate that more general network topologies might present several additional specific issues that will have to be investigated. However, our primary objective in this paper is to analyze the behavior of routing mechanisms based on parameterized heuristics within a relatively

Design and analysis of parameterized heuristics

Routing messages in large communication networks so as to optimize some desired set of performance criteria presents an instance of resource-bounded, multi-criteria, real-time, optimization problem. Our proposed solution to this problem involves the use of utility-theoretic heuristics (Mikler et al., 1996). Utility is a measure that quantifies a decision maker's preference for one action over another (relative to some criteria to be maximized) (French, 1986). When the result of an action is

Summary and discussion

Routing and control mechanisms which are based on parameterized heuristics can significantly reduce the resource requirement for storage, acquisition, and use of network state information while achieving the desired performance (as defined by the criteria such as average message delay). Conventional routing and control mechanisms rely on relatively up-to-date information about the state of the entire network. Hence, in large communication networks with thousands of nodes distributed over a wide

Acknowledgements

Honavar was supported in part by the National Science Foundation through grant NSF IRI-9409580.

Armin R. Mikler received his Diploma in Informatik from Fachhochschule Darmstadt, Germany in 1988. He received his M.S. and his Ph.D.(Computer Science) from Iowa State University in 1990 and 1995, respectively. From 1995 to 1997, he worked as a postdoctoral research associate in the Scalable Computing Laboratory at the Ames Laboratory, USDOE. Since 1997, Dr. Mikler has been an assistant professor in the Department of Computer Sciences at the University of North Texas (UNT). His research

References (18)

There are more references available in the full text version of this article.

Cited by (3)

Armin R. Mikler received his Diploma in Informatik from Fachhochschule Darmstadt, Germany in 1988. He received his M.S. and his Ph.D.(Computer Science) from Iowa State University in 1990 and 1995, respectively. From 1995 to 1997, he worked as a postdoctoral research associate in the Scalable Computing Laboratory at the Ames Laboratory, USDOE. Since 1997, Dr. Mikler has been an assistant professor in the Department of Computer Sciences at the University of North Texas (UNT). His research interests include: Intelligent Traffic Management, Coordination of Intelligent Mobile Agents, Decision-Making in Distributed Environments, and Cluster Computing. At UNT, Dr. Mikler founded and currently directs the Network Research Laboratory (NRL) for research on Intelligent Mobile Agents, Autonomous Distributed Systems, Mobile and Wireless Network Management, and Distributed Simulation (see http://www.nrl.csci.unt.edu).

Dr. Mikler is a member of UNT's Institute of Applied Science and has been conducting collaborative research in computational science. Prof. Mikler is an associate editor for Telecommunication Systems and a member of ACM and the IEEE Computer Society.

Dr. Vasant Honavar received a B.E. in Electronics Engg. from Bangalore University, India, an M.S. in Electrical and Computer Engg. From Drexel University, and an M.S. and a Ph.D. in Computer Science from the University of Wisconsin, Madison. He founded and directs the Artifical Intelligence Research Laboratory www.cs.iastate.edu/∼honavar/aigroup.html in the Department of Computer Science at Iowa State University (ISU) where he is currently an associate professor. Honavar is also a member of the Lawrence E. Baker Center for Bioinformatics and Biological Statistics, the Virtual Reality Application Center, and the faculty of Bioinformatics and Computational Biology at ISU. His research and teaching interests include Artifical Intelligence, Machine Learning, Bioinformatics and Computational Biology, Intelligent Agents and Multi-agent systems, Distributed Intelligent Information Networks, Data Mining, Knowledge Discovery and Visualization. He has published over 100 research articles in refereed journals, conferences and books, and has co-edited three books. He is a co- editor-in-chief of the Journal of Cognitive Systems Research published by Elsevier. His research has been partially funded by grants from the National Science Foundation, the National Security Agency, the Defense Advanced Research Projects Agency (DARPA), the US Department of Energy, the John Deere Foundation, the Carver Foundation, Pioneer Hi-Bred Inc., and IBM. Prof. Honavar is a member of ACM, AAAI, IEEE, and the New York Academy of Sciences.

Dr. Johnny Wong is a Full Professor of the Computer Science Department, Iowa State University at Ames, Iowa, USA. Before he came to Iowa State University, he was on the faculty of the Computer Science Department at the University of Sydney. His research interests include Operating Systems, Distributed Systems, Telecommunication Networks, Integrated Services Digital Networks (ISDN), Concurrency Control and Recovery, Object-Oriented Databases, Multilevel Database and Network Security, Multi-Agent Systems for Intrusion and Countermeasures.

He has been an investigator for research contracts with Telecom Australia from 1983 to 1986, studying the performance of network protocols of the ISDN. During this period, he has contributed to the study and evaluation of the communication architecture and protocols of ISDN. From 1989 to 1990, he was the Principal Investigator for a research contract with Microware Systems Corporation at Des Moines, Iowa. This involved the study of Coordinated Multimedia Communication in ISDN. In Summer 1991 and 1992, Dr. Wong was supported by IBM corporation in Rochester. While at IBM, he worked on the Distributed Computing Environment (DCE) for the Application Systems. This involved the integration of communication protocols and distributed database concepts. Dr. Wong is also involved in Coordinated Multimedia System (COMS) in Courseware Matrix Software Project, funded by NSF Synthesis Coalition Project to enhance engineering education. From 1993 to 1996, he was working on a research project on a knowledge-based system for energy conservation education using multimedia project communication technology, funded by the Iowa Energy Center. From 1995 to 1996, he was supported by the Ames Laboratory of the DOE, working in Middleware for Multidatabase systems.

Currently, he is working on several projects, including Intelligent Multi-Agents for Intrusion Detection and Countermeasures funded by the Department of Defense (DoD), Database-generating and X-ray Displaying World Wide Web Applications funded by Mayo Foundation, and CISE Educational Innovation: Integrated Security Curricular Modules funded by National Science Foundation (NSF).

View full text