Abdullah b Gani

A Brief Statement of Research Topic

Towards Mobile Agents-based for Network Performance Management in Java-based Environment

According to International Data Corp, network grows at 4,000 PCs are connected daily. The rapid growth of network was attributed to the need in sharing information, transferring files between users. The development of networking technology enables more networking services be offered to the users and this has led to some degree of dependency on network availability and reliability. It is the ultimate goal of performance management to ensure availability and reliability of network services. Performance management, as one of the International Organization of Standardization (ISO) Network Management Framework, involves in monitoring and controlling network resources by carrying out measuring performance of network hardware, software and media, analyzing management information and reporting to network manager.

This study investigates the use of mobile agents for the purpose of performance network management in Java-based environment. Java-based environment allows computing resources to be found in the same fashion of ‘Plug-and-Play’ technology. All resources are ‘federated’ in lookup service where potential client could make use all the available resources by performing discovery protocol. Due to highly volatile nature of available resources in the ‘federation’, the issue of availability and reliability are interesting to be resolves. The question of to what extent the mobile agents are not subscribing to network performance degradation need to be addressed. An analysis on performance management parameters will be conducted to find the significance contribution in using mobile agents.

In this research, Grasshopper is chosen as a test-bed for representing Java-based environment. The selection was based on several criterions. Grasshopper is the first mobile agent environment, which is compliant to the Mobile Agent System Interoperability Facility (MASIF) standard. The MASIF standard is built on top of COBRA, thus providing the integration of the traditional client/server paradigm and mobile agent technology. The standardization ensures that agent publications will be open towards other agent environment. Grasshopper allows one to build agent-enabled distributed applications, which take advantage of local high-speed communication and local high-speed data access.

This research requires several tools need to be developed for measuring performance management parameters. In order to facilitate the mobile agent agents mobility, suitable framework need first to be developed. In this framework, consideration of size, number and component of mobile agent, and mobile agent implication on the performance were taken into account in ensuring higher performance can be obtained. As throughput rate is dependent to traffic volume, mobile agents are envisaged not to cause detrimental in performance of the network because of mobile agent does processing at the node and activity of data exchange between Manager and Agent is reduced. This is the main advantage of mobile agent in performance management.

Single mobile agent is chosen because its capability to minimize the probability of packet losses in transmission of data exchange between Manager and Agent albeit limitation of embedded functions. Agent’s functions are compressed (zipped) into loader and stored in AgentLoader. After AgentHeader is invoked by the Manager, it starts to migrate to the determined node and decompress the loader. AgentLoader contains function of read, write, recovery action, and acknowledgement.


Figure 1 - Agent Architecture

Mobile agent requires framework to enable possible migration from one location to another in carrying its specific tasks. Without framework, mobile agents are restricted to move around and incompatible. In this framework, it contains three components - Manager, Agent and Clients. Manager resides in the server authorizes agent to be invoked to start its services by migrating to determined nodes and to collect Management Information Base (MIB) data for the purpose of analyzing the state of network devices. MIB is management information that contains in network devices. Management information of network devices is stored in its database called Management Information Base (MIB). MIB contains information on system description, number of networking interfaces, IP address, count of number of incoming and outgoing datagram, and table of information about active TCP connections (routing). Beside, invoking the agent, it also functions as controller room for all the mobile agents. In functioning as a controller room, it has to record all activities of the mobile agents, changes of device state and to display messages that related to mobile agent activities.

Mobile agent visits the nodes to read MIB and carry out analysis to determine the state of network devices. Intelligent ability helps mobile agent to perform the latter task. Porter as another capability of mobile agent sends acknowledgement to manager for the purpose of recording and displaying. 


Figure 2 - Framework of Mobile Agent for Performance Management

 Mobile agent -based framework minimizes the flow of traffic between server and nodes as processes are done remotely and autonomously. This gives advantages to other processes to be performed by node or network devices with minimum interference. In SNMP framework, agent is deployed to read and write MIB variables, which is similar in mobile agent framework. Agent in SNMP resides in node and performs its task according to request from Manager but mobile agent resides in node performing its tasks intelligently without have to wait request from Manager. Mobile agent decides in migrating to another node autonomously. Less element of request message in communication helps the performance of network always high.

  The size of agents significantly also affects the performance of network in its operation. To lessen the impact of size of agent on network performance, functions of mobile agent are zipped in AgentLoader, which is another advantage compared to agent in SNMP in reducing network load and as result of that it keeps the performance of tansporter always high.

Communication between manager and node in SNMP is dedicated-type fashion where agent interacts directly with manager. In other words, it is point-to-point communication. In mobile agent framework, this only occurs at the end of task completion for sending acknowledgements. Mobile agent in nature migrating from one node to another, and because of short period of stay, utilization of node resources is minimal. This is important due to limitation in node resources of processing power and storage capacity.