The centralized approach in network management has limitations on scalability and efficiency. A new framework is proposed for network management using mobile agents by analyzing the limitations of the existing centralized approach—Simple Network Management Protocol (SNMP). The mobile agents decentralize the processing and control, thus reducing the traffic on management station and distributing the processing load. The proposed work compares the use of mobile agent with the static agent. The mobile agents' use in managing the networks is simulated, and the performance of the mobile agents is compared with the centralized management. The analysis of these two methods shows that performance is improved with the use of mobile agents. The experimental results prove the advantages and the flexibility of the proposed framework.
The current network management technologies follow the two widely adopted approaches namely, ITU-T's Telecommunication Management Network (TMN) and IETF's Simple Network Management Protocol (SNMP). In IP Networks Management (Li and Sandrasegeran, 2005), but they take up different standards and protocols. The current network management systems, i.e., centralized network management environments are based on SNMP protocol and client-server model. The management station acts as a client and it endows a user interface to the network manager and also interacts with the agents. The agents are the servers and they manage remote access to Management Information Base (MIB). According to Subramanian (2003), an MIB is a virtual information store and defines the relationship between managed objects and groups related objects into MIB modules. SNMP helps the network administrators to manage the performance of the network and to locate and resolve the network problems. In SNMP (Subramanian, 2003), the manager protocol is used to establish the communication with the managed system and the agent protocol is used to communicate with the MIB and manager protocol. So, the processing of managed data is on the management station. In the managed nodes, the network management stations intermingle with the SNMP agents, and each SNMP agent is a daemon process that responds to the management station's request. The drawbacks of this approach include—the information bottleneck on the manager, lack of scalability, heavy processing load at the manager, high network bandwidth usage for the management activities, and centralized management (Stallings, 1999; and Pattinson, 2001).
The centralized management model works well while managing simple networks, but as the size of the network system grows, the extraction of the right information becomes increasingly difficult. Networked systems and even networks themselves need constant monitoring for the purpose of management, particularly for fault diagnosis, performance prediction and protection of the system resources from intruders or attackers. In a large enterprise network, network monitoring is very difficult because it consists of hundreds of nodes and resources with varying heterogeneity, in terms of the hardware and software used. |
