A framework for building reusable mobile agents for network management

I. Satoh
NOMS 2002. IEEE/IFIP Network Operations and Management Symposium. ' Management Solutions for the New Communications World'(Cat. No.02CH37327)  
Mobile agents can migrate among nodes to perform a set of management tasks at each of the visited nodes. Existing mobile agent-based network management systems often assume that their mobile agents are designed to work in particular networks to raise the efficiency of agent migration among multiple nodes. Unfortunately, such mobile agents cannot be reused in different networks. This paper proposes a framework where a mobile agent for network management is composed of two kinds of software
more » ... ents, a itinerary part and a behavioral logic part. Both components are implemented as mobile agents. The former is a carrier designed for particular networks, and it can efficiently navigate other mobile agents among nodes in its target network. The latter defines management tasks performed at each node independently of any local network. This framework allows a mobile agent for network management to be reused in various networks without being modified. A prototype implementation of this framework and its application were built on a Java-based mobile agent system. new or existing functionalities into a network system are extremely important tasks especially as they potentially allow outdated systems to be updated in an efficient manner. Adopting the mobile agent technology eliminates the need for the administrator to constantly monitor many network management activities, e.g., installation and upgrading of software and periodic auditing of the network. There have been several attempts to apply this technology to network management tasks. However, there have been serious problems associated with the development of mobile agent-based applications, in addition to security problems. Such applications are required to migrate their agents among all specified nodes efficiently to perform their own tasks at each of the visited nodes, because the itinerary of an agent greatly affects its achievement and efficiency. However, it is often difficult to dynamically generate an efficient itinerary among multiple nodes, without having any knowledge of the network. Even if a smart agent can create an efficient itinerary based on its previous processing and the current environment, such an agent is not always be appropriate, because both the cost of discovering such an itinerary and the size of its program tend to be large. Therefore, most existing mobile agent-based applications explicitly and implicitly assume that their mobile agents are statically designed for their target networks for greater efficiency of agent migration over the networks. However, an agent optimized for particular networks cannot be reused in other networks. This results in an inevitable trade-off between the performance and reusability of a mobile agent. Furthermore, this problem becomes more serious when mobile agents are used for network management. This is because network management systems must often handle networks that may have some malfunctions and whose topology may not be exactly unknown. Consequently, it is almost impossible for each mobile agent to efficiently migrate among nodes in such networks. This is one of the reasons why there have not been many attempts to use mobile agent technology in the domain of network management, although the technology can be used in this domain. This paper addresses several problems, including the problem of a trade-off between the performance and reusability of a mobile agent in the development of mobile agent-based network management systems and proposes a new framework for building mobile agent-based network management systems in order to solve these problems. The framework allows us to build efficient and scalable mobile agents for network management without losing their reusability. Like other mobile agent-based network management systems, the framework uses mobile agents to implement network management functionalities, but it allows each mobile agent to be designed independently of any network and dynamically change its itinerary without modifying its application-specific behaviors. That is, when a mobile agent arrives at an unknown sub-network, it can dynamically obtain an itinerary statically designed for the visited sub-network and thus can efficiently migrate among the nodes on the sub-network. The current implementation of the framework is built on a Java-based mobile agent system, called MobileSpaces [13], which is unique among existing systems because it hierarchically organizes multiple mobile agents. This paper is organized as follows: Section 2 discusses the advantages of mobile
doi:10.1109/noms.2002.1015544 dblp:conf/noms/Satoh02 fatcat:o3ydai26tbfgpktfvb5tpw3m2y