Rule-Driven Mobile Intelligent Agents for Real-Time Configuration of IP Networks

Even though intelligent agent has proven itself to be a promising branch of artificial intelligence (AI), its mobility capacity has yet been paid enough attention to match the pervasive trend of networks. This paper proposes to inject intelligence into mobile agent of current literature by introducing ruledriven mobile agent so as to maintain both intelligence and mobility of current agent. Particularly, this methodology is fully exemplified in the context of realtime IP network configuration through intelligent mobile agent based network management architecture, policy specification language and policy information model. A case study for inter-domain IP VPN configuration demonstrates the design and implementation of this management system based on the test-bed developed in the context of European Union IST Project CONTEXT. 1 Background and Rationale After years of recession, Artificial Intelligence (AI) regained it vitality relatively thanks to the inception of Intelligent Agent (IA). Agent was even highlighted as another approach of AI by S. Russell et al. [1]. Intelligent agent usually is a kind of software of autonomous, intelligent and social capability. Intelligent Agent and its related areas have been intensively researched over last decades and enormous achievement covering a wide range of research fields are available in the literature. As computers and networks become more pervasive, the requirement of intelligent agent being more (automatically) moveable is getting more a necessity than an option. As an active branch of agent technology researches, mobile agent paradigm intends to bring an increased performance and flexibility to distributed systems by promoting "autonomous code migration" (mobile code moving between places) instead of traditional RPC (remote procedure call) such as CORBA, COPS (Common Open Policy Service) [2]. It turns out that more attention has been given to the mobility of mobile agent whereas the intelligence of mobile agent is seldom talked about in the mobile agent research community. Mobile agent technology is very successfully used in the network-related applications, especially network management, where its mobility feature is largely explored [3], but these mobile agents are usually lack of intelligence. We believe mobile agent is first of all an agent that has intelligence. This paper aims to explore the potential use of mobile agent to manage IP network in a more intelligent and automated way. For this purpose, mobile agents should Kun Yang, Alex Galis contain certain extent of intelligence to reasonably respond to the possible change in destination elements and perform negotiation. This kind of intelligence should reflect the management strategy of administrator. A straightforward way for network administrator to give network management command or guide is to produce highlevel rules such as if sourceHost is within finance and time is between 9am and 5pm then useSecureTunnel . Then mobile agent can take this rule and enforce it automatically. By using rules to give network management command or strategy, a unique method of managing network can be guaranteed. The use of rule to management network is exactly what so called Policy-based Network Management (PBNM) [4] is about since policies usually appear as rules for network management. Here in this paper, we don’t distinguish the difference between rule-based management and policy-based management and in many cases, the term “policybased” is more likely to be used. In order to put this idea into practice, a specific network management task is selected, i.e., IP VPN (Virtual Private Network) configuration. VPN enables an organization to interconnect its distributed sites over a public network with much lower price than the traditional leased-line private network. VPN is a key and typical network application operating in every big telecom operator as a main revenue source. But the lack of real-time and automated configuration and management capabilities of current IP VPN deployment makes the management of growing networks timeconsuming and error-prone. The integration of mobile agents and policy-based network management (as such making a mobile intelligent agent) claims to be a practical solution to this challenge. This paper first discusses an intelligent mobile agent based IP network management architecture with emphasis on IP VPN; then a detailed explanation with respect to policy specification language and IP VPN policy information model is presented. Finally, before the conclusion, a case study for inter-domain IP VPN configuration is demonstrated, aiming to exemplify the design and implementation of this intelligent MA-based IP network management system. 2 An Intelligent MA-based IP Network Management Architecture 2.1 Architecture Overview An intelligent MA based IP network management system architecture and its main components are depicted in Fig. 1, which is organized based on the PBNM concept as suggested by IETF Policy Working Group [5]. Please note that we use IP VPN configuration as an example, but this architecture is generic enough for any other IP network management tasks provided that the corresponding PDP and its information model are given. The PBNM system mainly includes four components: policy management tool, policy repository, Policy Decision Point (PDP) and Policy Enforcement Point (PEP). Policy management tool serves as a policy creation environment for the administrator to define/edit/view policies in an English-like declarative language. Rule-driven Mobile Intelligent Agents for Real-time Configuration of IP Networks