Enterprise network management using intelligent mobile agents.

摘要

With the proliferation of new networking technologies and the demand for new services, it will become nearly impossible to manage large-scale networks without an intelligent automated management system. The majority of computer networks today are managed using the Simple Network Management Protocol (SNMP), developed by the IETF to manage TCP/IP based networks. This protocol is based on a client/server model and therefore highly centralized, with very limited security and alarm propagation. The emergence of high-speed telecommunication devices, mobile networks, and broadband transmission systems require a more efficient, distributed, and real-time network management standard. This dissertation presents a novel approach to broadband network management using intelligent mobile agents. First, it details the design and implementation of a highly robust mobile agent architecture, called Discovery, which facilitates the efficient migration and communication of transportable agents. This framework provides an infrastructure for developing secure, platform-independent distributed applications. Second, this dissertation describes a new network management framework that implements fault, configuration, and performance management for Asynchronous Transfer Mode (ATM) networks using intelligent mobile agents. This system is built on the top of the Discovery architecture to provide a device-independent environment for managing heterogeneous network components. Mobile agents are employed to implement an automated scheme for detecting the physical topology of an arbitrary ATM network. The proposed algorithm is fully distributed and therefore significantly more scalable and robust than traditional approaches based on the client/server model. This dissertation also introduces a set of algorithms that facilitate mobile agent based permanent virtual circuit (PVC) provisioning in ATM networks. This includes virtual path and channel establishment, traffic parameter negotiation, virtual circuit reconfiguration, and teardown. The proposed scheme has significant advantages over traditional approaches using SNMP, because it facilitates the automatic provisioning of permanent virtual paths in a highly distributed fashion without requiring constant interaction from the human network manager. Finally, mobile agents were also used to implement a scalable network monitoring scheme for fault and configuration management. Through formal analysis, we provided evidence that in several areas of network management the mobile agent based approach generates a significantly more scalable and efficient solution than previous approaches based on the client/server model.