A STUDY OF THE BEHAVIOUR OF THE MOBILE AGENT IN THE NETWORK MANAGEMENT SYSTEMS Tarag Fahad, Sufian Yousef & Caroline Strange School of Design and Communication Systems, Anglia Polytechnic University Victoria Road South, Chelmsford, Essex, CM1 1LL, UK. E-mail: {t.fahad, s.yousef, c.m.strange}@apu.ac.uk Abstract This paper investigates the performance management of two different size networks using mobile agents. The performance metrics studied are the response time, traffic and utilisation. Mobile agent performance is analysed through management task simulations using an Optimised Network Engineering Tool (OPNET). Four simulation scenarios were implemented in networks; two with no load applied and two with the following applications - email, ftp application, database application and http. It was found from this study that mobile agents perform efficiently in large networks in terms of reducing response time and traffic. Keywords: Mobile agents, Simulation, OPNET, Network Management. 1. INTRODUCTION The growing complexity of computer networks requires the use of sophisticated management techniques. Current network management systems are typically designed using a centralised network management paradigm that is characterised by a low degree of flexibility. Most of these systems use the Simple Network Management Protocol [1]. Management distribution may be enhanced using mobile agents. These provide a powerful software interaction paradigm that allows code migration between hosts for remote execution [2]. Mobile agents are used in the management of network systems to distribute the management functionality throughout the network, to manage both rapid changes and scalability of complex networks. Mobile Agent technology is a rapidly developing area of research in the fields of network management. As a mobile program it can roam in a network, perform duties on behalf of their creators and returning to the creator after having performed the duties assigned to it [3]. They are self-directed and have the ability to migrate from one node to another performing actions on behalf of the user. In recent years, a number of papers linking mobile agents to network management have been published. These have considered, for example, tasks of network health monitoring [4], fault diagnosis [4, 5], performance management [6] and network configuration [5]. Rubinstein et al [7] investigated mobile agent tradeoffs in network management by performing management task simulations. Their analysis focuses on the performance of two unload strategies in order to reduce the response time for mobile agent. In the first approach, the mobile agent returns to the management station, whereas in the second one the mobile agent sends data to the management station, both after visiting a fixed number of nodes. Rubinstein et al [7] concluded that the use of mobile agents reduces the response time. However, their work failed to focus on the possible advantages of mobile agents in reducing network traffic. In implementing flexible network management strategies Puliafito et al [8] discuss the advantages of mobile agent technology over SNMP agents. They implemented a mobile agent platform (MAP) that uses CORBA as an underlying communication mechanism for the transfer of code and data. This innovative approach to network management focused on security schema for mobile agent and it only identifies a simple traffic model without adequate verification methodologies. Baldi and Picco [9] describe a quantitative model for management related network traffic and a formal tool for determining the best way to perform management operations. Several different mobile code design paradigms are compared with the use of the traditional client-server architecture. Their work provide an extensive quantitative analysis on the model of network traffic for client/server, mobile agent, remote evaluation (REV) and code on demand (COD) approach but lacks in presenting experimental analysis of the deduced equations. Quantitative analysis for latencies involved while managing a network using the above approaches was not evaluated. The rest of this paper is organised as follows. First, an overview of the mobile agent infrastructure is provided. Then, the benefits of mobile agents in network management system are discussed. Next, we explain the mobile agent simulation module and its parameters. Subsequently, we discuss and analyse the simulation results. Finally, The paper is concluded with remarks on future work. ISBN: 1-9025-6009-4 2003 PGNet
2. MOBILE AGENT INFRASTRUCTURE Reduction in network traffic: Code is often smaller than data that it processes. Therefore, In order to perform network management functionality using mobile agent, it is necessary to have an infrastructure that provides a framework for code mobility. Pagurek et al [12] explain the development of Mobile Agent Infrastructure as shown in figure 1. Every network component (NC) contains a Mobile Code Daemon (MCD) running within a Java Virtual Machine (JVM). The MCD provides a number of services that facilitates the execution of mobile agents. Included are: a Mobile Code Manager (MCM) that manages the life cycle of a mobile agent from its arrival an authentication at the network component to its migration or perhaps destruction, a Migration Facilitator to transport mobile agents between NCs, a Communication Facilitator for collaboration between local and remote mobile agents, and an interface called the Virtual Managed Component which provides for mobile agents accessing the NC s managed objects and resources in a controlled and secure way. The VMC is responsible for management of the mobile agents access rights and the allocation of resources to that agents. NC = Network Component MF = Migration Facility MA = Mobile Agent JVM = Java Virtual Machine MCD = Mobile Code Daemon MCM = Mobile Code Manager VMC = Virtual Managed Component Figure 1. Mobile Agent Infrastructure the transfer of mobile agents to the sources of data creates less traffic than transferring the data. Remote objects can help in some cases, however they also involve loads of parameters, which may be large. Space savings: Resource consumption is limited, because a mobile agent resides only on one node at a time. In contrast, static multiple servers require duplication of functionality at every location. Mobile agents carry the functionality with them, so it does not have to be duplicated. Remote objects provide similar benefits, but the costs of the middleware might be high. Efficiency savings: CPU consumption is limited, because a mobile agent execute only on one node at a time. Other nodes do not run an agent until needed. Support for heterogeneous environments: Mobile agents are separated from the hosts by the mobility framework. If the framework is in place, agents can target any system. The costs of running a Java Virtual Machine (JVM) on a device are decreasing. Java chips will probably dominate in the future, but the underlying technology is also evolving in the direction of ever-smaller footprints Robustness and fault tolerance: If a distributed system starts to malfunction, then mobile agents can be used to increase availability of certain services in the concerned areas. For example, the density of fault detecting or repairing agents can be increased. Some kind of meta-level management of agents is required to ensure that the agent-based system fulfils its purpose. Easy software upgrades: A mobile agent can be exchanged virtually at will. In contrast, swapping functionality of servers is complicated; especially, if we want to maintain the appropriate level of quality of service (QoS). Online extensibility of services: Mobile agents can be used to extend capabilities of applications, for example, providing services. This allows for building systems that are extremely flexible. 3. BENEFITS OF MOBILE AGENTS 4. SIMULATION MODULE The use of mobile agents has advantages over other implementations of static agents, which relay heavily on SNMP. The following advantages based on [5] and they include the following: OPNET Modeler 6.0 [11] has been used to develop and simulate the Mobile Agent module. In order to represent and study the behaviour of mobile agent in the network management system we have developed simulation module called Mobile Agent Module. It divided up into four scenarios; two scenarios with
unloaded application (i.e. email, ftp application, database application and http). Figure 2. Mobile Agent 5 Nodes Scenario Contents the manager to the mobile agent, and a Response is the information sent from the mobile agent back to the manager. An initial network utilization of 25% starts after the first 100-second of the three hours simulation time. The Traffic Sent parameter represents OPNET Background traffic type, which refers to analytically modelled traffic that impacts performance of explicit traffic by inducing additional delays. Background traffic is not modelled by individual discrete events. The presence of background traffic results in queue build-up at intermediate devices and causes delays based on the queue length at any given time. Notes that we have assumed that the initial size of the mobile agent is 5000 bytes, which based on [7] and each request or response of a variable is 50 bytes long. As a result, Response Transfer Size attribute will be 5050 bytes. The first one contains five nodes (figure 2) connected to the manager trough switch, and the second one contains ten nodes instead of five. Figure 3 shows how mobile agent behaves in the scenario where it collects information that requested from the manager and migrating from node to other to do so. The other two loaded scenarios are similar to unloaded scenarios except that each of them loaded with the database application, email, ftp and http applications. These were simulated using representative choices from OPNET defined level heavy. The main aim is to see what will happen to the network performance when such applications are added, then comparing the results with the previous two scenarios. 6. SIMULATION RESULTS 6.1 Unloaded Scenarios Figure 4 shows that the response time in the mobile agent ten nodes scenario is higher than response time in mobile agent five nodes scenario. The first one was just above 10 sec, whereas the second was 6 sec. This result was expected, taking into account the number of network managed nodes in each scenario, mobile agent in small network will take short response time compared with large one. Figure 3.Mobile Agent 5 Nodes Scenario behaviour 5. SIMULATION PARAMETERS To study the effect of adopting mobile agents in the network on the overall network performance the response time, traffic sent and utilisation are required from the OPNET simulation. The response time is time taken by the actions required to complete a full task (a complete polling operation or problem investigation). It starts from the time when the manager sends the request to the mobile agent and ends when a response is received. A Request is the information sent from Figure 4. Unloaded scenario response time results This will lead to the fact that for the mobile agent response time increases faster when the number of managed nodes grows up due to the incremental size of the mobile agent when it migrate from node to other. On the other hand, figure 5 shows that the increase in the number of managed nodes in unloaded scenario will lead to decrease in the traffic. The traffic sent in the scenario was decreased by almost 30% compared with scenario, where it was 600 bytes/sec in the scenario and almost 900
bytes/sec in the scenario. This shows the advantages of mobile agent in reducing traffic when the number of network managed nodes is increased. Figure 7. Loaded scenario response time results Figure 5. Unloaded scenario traffic sent results In terms of utilisation, figure 6 shows that the utilisation does not change when the number of network managed nodes was increased, and it remains almost at 25%. Figure 8. Loaded scenario traffic sent results There is no impact on the utilisation by loading application and it remains in both 5 and loaded scenario at almost 25%. Figure 6. Unloaded scenario utilisation results 6.2 Loaded Scenarios It can be seen from figure 7 and 8 that there is no significant change in terms of the response time and traffic sent parameters in both and scenarios compare with the previous two scenarios. Therefore, we can conclude that, mobile agent behaviour does not change when further application added to the managed network nodes. 7. CONCLUSION AND FUTURE WORK This work has focused on investigating the effect of adopting mobile agents in network management system. As discovered during the simulation, mobile agent performed better in terms of reducing traffic when the number of network managed nodes was increased. However, as the number of network managed nodes increase the response time will increase. Based on the above results, we can conclude that mobile agent will perform efficiently in large network management system than in the small one. The results appear to support the intuitive expectations of mobile agents behaviour and indicate their advantages of reducing traffic across the network. The future is likely to bring more use of mobile agent in network
management systems. A number of directions have been identified for future work as follows: To explore the limitations of the mobile agent paradigm, with the intention of finding precisely what the circumstances are under which the use of mobile agent to manage network does create the problems described. To develop a simulation of the mobile agent approach to network management that considers all aspects of network management including fault and security management. To develop a simulation of the SNMP v3 similar to above point in order to identify comparative behaviour and determine the suitability of using each of them in the network management systems. ACKNOWLEDGEMENT The authors sincerely thank Dr. Colin Pattinson and Mr. Joao Ponciano for their helpful comments and support in this work. [7] Rubinstein, M. G., Duarte, O. C. M. B and Pujolle, G. Using Mobile Agent Strategies for Reducing the Response Time in Network Management. 16th IFIP World Computer Congress, ICCT2000, pp. 278-281, Beijing, China, August 2000. [8] Puliafito, A. Tomarchio, O. Using Mobile Agents to implement flexible Network Management strategies. Computer Communication, Vol. 23 (2000) pp. 708-719, April 2000. [9] Baldi, M. Picco, G. Evaluating the tradeoffs of Mobile Code Design paradigms in Network Management Applications. 20 th International Conference of Software Engineering (ICSE 98) pp. 146-155. [10] Pattinson, C. A simulated network management information base, Journal of Network and Computer Applications, Vol. 23 (2000) pp.93-107, April 2000. [11] OPNET Technologies, inc. [Internet] http://www.opnet.com [Accessed 2 March 2002]. [12] Pagurek, B, Y. Wang, and T. White, Integrating of Mobile Agents with SNMP: Why and How, Proceeding of DSOM, 2000. REFERENCES [1] Leinwand, A. Conroy, K. F. 1996, Network Management A Practical Perspective, 2nd ed. Addison Wesley. [2] Subramanian, M. 2000, Network Management Principles and Practice, Addison Wesley. [3] Kona, M. Xu, C A. Framework for Network Management using Mobile Agents, International Parallel and Distributed Processing Symposium: IPDPS 2002 Workshops, April 15-19, 2002, Fort Lauderdale, Florida. [4] Pattinson, C. A Study of the Behaviour of the Simple Network Management Protocol, the 12th International Workshop on Distributed Systems: Operations and Management, DSOM 2001, Nancy, France, October, 2001. [5] El-Darieby, M. Biezczad, A. Intelligent Mobile Agents : Towards Network Fault Management Automation Proc. 6th IFIP/IEEE International Symposium on Integrated Network Management (IM 99) pp. 611-622. [6] Bieszczad, A. White, T. Pagurek, B. Sugar, G. Tran, X. Intelligent Network Modeling using Mobile Agents. (1998), In Proceedings of the IEEE Global Telecommunications Conference GLOBECOM 98, November 8th-12th, 1998, pp. 1082-1087.