NETWORK MANAGEMENT BASED ON PC COMMUNICATION PLATFORM WITH SNMP AND MOBILE AGENTS Jae-Kyu Chun ,Ki-Yong Cho, Seok-Hyung Cho, Young-Woo Lee, Young-Il Kim Access Network Lab., R&D Group, Korea Telecom, Korea E-mail: { jkchun, kycho, shjo, ywlee, yikim}@kt.co.kr Abstract We have constructed Information Communication Processing System(ICPS) and Advanced ICPS(AICPS) as a kind of gateway system using communication service network to provide PC communication service(on-line dial-up service) to users (customers). Users could connect ICPS or AICPS using 01410/01411/01412 connection dial number, especially management of 01412 PC communication service network resource provides resource data through one Local O&M System(LOMS) per each AICPS. Nation-wide integrated network management has been provided centralized integration operating network management service through POWWOW-NMS which manages Network Elements(NEs) in PC communication network. We implemented network management system(NMS) of POWWOW-NMS in PC communication network. We also analyze SNMP pollingbased network management in comparison with agentbased network management. 1. Introduction Korea becomes the one with rapid growth in the Internet community and related business. Internet access through super-high speed network has also exponentially increased about 4.5 million today. It catches the international attention as a unique and viable business model of Internet community. The motivation of ICPS/AICPS’s development is as follows; ICPS/AICPS provide services using PSTN, ISDN with multiple network connection to users, even though Remote Access System(RAS) provides connection between PSTN, ISDN and Internet network. ICPS/AICPS provide connection service of packet network, frame network as well as internet, including Menu Manager service, Inforshop/ Web-Inforshop service and so on. Network management is a critical issue in today’s rapidly changing network environment. POWWOW-NMS effectively operates and manages network traffic and network resources on PC communication network based on ICPS/AICPS. POWWOW-NMS has been implemented that it collects, stores and manipulates necessary Operation & Management(O&M) data for configuration management, fault management, performance management, keeping pace with phased development of ICPS/AICPS since 1995. This paper is organized as follows; The first section provides a brief overview of the trends of PC communication service, especially dial-up service. The section following that briefly outlines the development of ICPS/AICPS and describes Network Management System(NMS) of POWWOW-NMS and Service Management System(SMS), A-GAUSS, providing PC communication network service and also provide POWWOW-NMS’s architecture and operating user interfaces. This paper goes on to compare with traditional network management method and mobile agent model applied to POWWOW-NMS. Last section concludes the paper with future works. 2. Phased PC communication network development The system have been developed to provide PC communication service using 01410/01411/01412 dial-up connection number. We have provided PC communication service through three level’s system upgrade. As the first step, ICPS has been developed and operated with connecting 01410/01411 dial-number in 1995. ICPS consists of Information Service Access Point(ISAP), Mediation Operations, Administration and Maintenance system(MOAM), Central Operations, Administration and Maintenance system (COAM) and Guidance And User management Service System(GAUSS) [1]. As the second step, AICPS had been developed with substantial functions in comparing to ICPS connecting 01412 dial-number. AICPS consists of Network Access Sub-system (xNASs), High Speed Switching Fabric(HSSF), O&M NAS(ONAS), Local O&M System(LOMS) and Advanced GAUSS(A-GAUSS). It totally provides about 80,000 channels. As the final step, AICPS20 had been developed with powerful functions such as increasing integration of xNASs and optimizing system. It totally provides about 110,000 channels. Their general function and specification can be described as follows: xNASs also provides connection interface as the same as AICPS, comprising Telephony NAS(TNAS), PSDN Proceedings of the 22 nd International Conference on Distributed Computing Systems Workshops (ICDCSW’02) 0-7695-1588-6/02 $17.00 © 2002 IEEE Element Management layer(EML) which abstracts all functions provided with NEL and manages each network element can be mapped into MOAM, LOMS and LOMS20. They play a role as managers and ISAP, xNAS are involved with agents. One MOAM manages about 20 ISAPs, one LOMS & LOMS20 manage each one AICPS & AICPS20. Data collection of MOAM, LOMS and LOMS20 performs using socket. Network Management Layer(NEL) is that it can manage the whole networks and system such as network configuration, performance analysis, statistics/report. As the same as EML, COAM and POWWOW-NMS play a part as manager to MOAM, LOMS and LOMS20 as agents. A COAM manages 17 MOAMs, POWWOW-NMS manages 54 LOMSs and 30 LOMS20s and collects data based on MIB value using SNMP. At the same time, POWWOW-NMS uses FTP for traffic analysis data daily. Service Management Layer(SML) manages the services offered to customers. GAUSS and A-GAUSS manage service information such as menu pages window, Information Provider(IP) information and Content Provider(CP) item information for routing INFORSHOP service. Business Management layer(BML) manages the overall business. However it does not support on PC communication network environment. NAS(PNAS), Frame Relay NAS(FNAS) and Web NAS(WNAS) not including ISDN NAS(INAS). ESAS(Ethernet Switch Access Sub-system) provides 32ports with 10/100Mbps Ethernet. It offers switch function between TNAS and other xNASs. LOMS20 is implemented on java machine. It supports web service interface, including AICPS’s ONAS functions. A-GAUSS additionally provides graphics menu management functions as well as text menu services including editing, creating, downloading and verifying menu items. 3. TMN based on hierarchical model Telecommunications Management Network (TMN) is a framework for the management of the networks and the services providing an environment for interfacing a telecommunication network. We are developing techniques based on TMN concepts that enable efficient maintenance and operation of a variety of PC communication systems[2][3][4]. Figure 1 shows the TMN based manager/agent model between network element. The architecture of O&M Model for POWWOW-NMS is as follows; Network Element Layer(NEL) which is comprised of network element could be compared to ISAPs, xNAS of PC communication network. This provides network management data to upper layer. • Modeling - Manager : COAM, POWWOW-NMS - Agent : LOMS,MOAM DCN - KTIS : TCP/IP Protocol : SNMP,FTP BML GAUSS A-GAUSS POWWOW-NMS-toLOMS/LOMS20 Centralized OA&M for 01412 Network COAM-to-MOAM Centralized OA&M for 01410/01411 Network SML COAM POWWOWNMS NML LOMS/LOMS20-to-xNASs MOAM-to-ISAPs Localized Service and OA&M Management EML MOAM MOAM LOMS MOAM LOMS LOMS LOMS20 LOMS20 LOMS20 40ISAPs/1MOAM CMIP SNMP DCOM CORBA JAVA ISAP ISAP ISAP ISAP ISAP ISAP ISAP NEL PNAS TNAS FNAS HSSF INAS ONAS WNAS Figure 1. TMN based on hierarchical model Proceedings of the 22 nd International Conference on Distributed Computing Systems Workshops (ICDCSW’02) 0-7695-1588-6/02 $17.00 © 2002 IEEE Localized Service and OA&M Management 1xNASs/1LOMS PNAS TNAS FNAS CNAS WNAS Service Logic Application Confi. Service Fault Service Perfor. Service Message Service Download Service Fault Service Report Service CORBA Radius Setting NE Access Logic Data Access Logic Application Radius Setter 1.X Service Data Operator Message 2.0 Service Data Alarm Event Service Data Service Data 1X Mediator 20 Mediator OpMesg Sender Event Provider Polling Manager Polling Data Mediation DB Polling control Event Count Collector Trap Monitor FTP Trap Database Access Polled Data Poller Alarm Counter Management Polled Data Processing Logic SNMP LOMS SNMP Agent/FTP Server Figure 2. POWWOW-NMS 4. POWWOW-NMS ARCHITECTURE POWWOW-NMS is separated into three tiers such as user interface logic application with private client, web interface program for public POWWOW-NMS or traffic monitoring and service logic application providing network management application functions and data access logic application collecting network management data based on communicating between POWWOW-NMS and LOMS. It is multi-tire architecture connected to CORBA platform like Figure 2. Functions at each layer are as follows; 4.1 Functions User Interface Logic Application : User interface with circumstances based on windows provides operators with interface environment services. It is also to provide that operators execute operation management of all LOMS and xNASs. Through cooperation of service application and data access application, user interfaces efficiently represent infrastructure network management functions. The kind of user interface is private POWWOW-NMS, web interface, report generator etc.. Service Logic Application : Service logic application interacts with user interface program and data access application, with result of each service xNAS Conf. Table xNAS Event Table AlarmTable Statistics/Report Table Temp Table Architecture Table backup File request/response. This layer is possible to execute stable service requests and flexible network management system. This layer provides 7 services such as configuration service, fault message service, performance service and so on. Data Access Logic Application : Data access logic application servers collect data of fault management, configuration management and performance management necessary to PC Communication network operation management. It is composed of NE Access Logic and Polled Data Processing Logic. NE access Logic function consists of pollers, which collect and transfer management data and polling manager which optimizes the number of poller to effectively manage system. Polled Data Processing Logic is located between database and NE Access Logic, analyzing data of collecting from poller, reprocessing collection data as well as event process to provide service request from service applications. 4.2 Management Services Provided with POWWOWNMS POWWOW-NMS takes advantage of object-oriented software features to archive a simple and readily extensible structure with C++ & Java classes collaborating to perform a particular function such as FCAPS(FM, CM, AM, PM, SM). Private MIB is used to communicate Proceedings of the 22 nd International Conference on Distributed Computing Systems Workshops (ICDCSW’02) 0-7695-1588-6/02 $17.00 © 2002 IEEE between powwow-NMS and LOMS, the number of OID value is about 800. In order to implement management services, POWWOW-NMS provides standard management functions such as; Fault Management : conducting alarm surveillance and fault correction, modifying alarm notification conditions, providing alarm reporting, assigning alarm event criteria functions, managing logs and so on. Configuration management : setting xNAS/LOMS, retrieving configuration data, managing database, installing LOMS, receiving notification of trap for configuration change and so on. Accounting Management : creating, deleting, activating items and so on. Performance Management : monitoring performance, controlling performance management, modifying notification conditions, analyzing performance, managing logs and so on. Security Management : authenticating, auditing trails, managing access logs and so on. 4.3 Method of Effectively Managing Data Collection We have designed and implemented POWWOW-NMS in PC communication network to support network management. However, it has been too much time to manage and collect 84 NE resources. So, we suggest several mobile agent method and implement in next chapter. 5. Mobile agent model To efficiently manage and operate network resources the majority of network management model has been managed by centralized client/server model based on manager and agent architecture. Centralized client/server model consists of managing entity, which plays part of manager, managed Objects and Network management protocol, SNMP(Simple Network Management Protocol) of IETF for data networks and CMIP (Common Management Information Protocol) of OSI for telecommunication networks. It is used to exchange messages in reference to MIB(Management Information Base), which is hierarchical base of information managed by each objects. Mobile agent model reduces the complexity of network management at the managing entity by delegating executable code and data to the managed entity. Providing mobility to an agent in the network management domain is an efficient way to address some of the critical network management problem such as network managed entitles’ scalability, service dynamic extensionability and network overhead etc.. Mobile agents are autonomous, asynchronous, and optionally intelligent software entities which in order to fulfill their tasks, can migrate to and reside in a number of networked nodes [5][6][7][8]. In this paper, in addition to implementing SNMP based POWWOW-NMS in PC communication network, we suggest that we cluster NEs and generate mobile agents simultaneously. In case of many managed NEs, we suggest that it would be possible to manage efficiently applying to clustered mobile agent model rather than snmp based model to provide fault and configuration information in real time to user. 5.1 Performance Measurement by Using Mobile Agents Traditional SNMP-based polling, which is intrinsically centralized, involves a flood of request/response messages. This naturally leads to a significant proportion of available bandwidth being used for management data. On the other hand, mobile agent model generates a single mobile agent responsible for obtaining network management data from each managed entity. In addition, the order in which managed entities are visited is arbitrary. Peer to peer mode suggests that the number of mobile agent is required per managed entities. Roaming mode suggests that a single mobile agent, which is generated at the network management system, roams distributed managed entities for collecting management data. As another approach, the clustering mode suggests that it is to partition the managed entities into logical or physical domains. In this paper, we compare and analyze using four methods based on POWWOW-NMS environment as follows. Figure 3 shows parameters for calculating response time. Figure 4 represents network management method using mobile agents and numerical formula. Processing time of NMS - Processing time of agents - β Processing time of SNMP protocol - r Network delay among agent nodes - t The number of agent nodes - n The number cluster among agent nodes - p Figure 3.ofTime consumption parameters Peer to peer(1:1) Method : Mobile agents launch from NMS to each other NE(Network Element) and returns the results to NMS. Roaming Method : A mobile agent launches from NMS to the first NE, with results a mobile agent moves to the second NE. After all NEs execute, final results return to NMS. Clustering Method : After clustering each NEs, mobile agents move to each clustered group simultaneously. The partitioning criteria in POWWOW-NMS is Proceedings of the 22 nd International Conference on Distributed Computing Systems Workshops (ICDCSW’02) 0-7695-1588-6/02 $17.00 © 2002 IEEE specified by 10 physical distributions of managed entities such as Seoul, Pusan, Chunnam etc.. 1 :1 Method SNMP Method : Polling method based on traditional client/server network management. Roaming Method wv~~v~Tutz wv~~v~Tutz svtz svtz Clustering Method wv~~v~Tutz svtz Figure 4. Network management method using mobile agents The overall response time could be approximately by T = ta + tb + tc , where ta represents the network transmission time of managed data, tb describes processing time of network management system such as processing transmitted data with result of interacting with managed entities and tc represents the time spent by agents or managed entities such as launching or destroying mobile agents, processing the requests of NMS in the managed entities[6]. If we represents network management elapsed(response) time by numerical formula, SNMP method shows 2nt + n + r. Network transmission time between NEs and POWWOW-NMS is double, sending and receiving time. The processing time of POWWOW-NMS is that the number of receiving of data from NEs. Peer to peer method is similar to SNMP based polling except the time of processing agents. Roaming method shows that one agent circulates NEs with n+1 (plus POWWOW-NMS) and the processing time of POWWOW-NMS is just single time because POWWOW-NMS processes one agent. Clustering method represents network transmission time of clustered NEs as (n+1)t/p because POWWOW-NMS simultaneously processes mobile agents. POWWOW-NMS’s processing time is p , sum of processing time with each mobile agent. 5.2 Comparison of response time To implement mobile agents, we applied to POWWOW-NMS operating on PC communication network(1 POWWOW-NMS + 84 NEs(LOMS) – UNIX machine). To implement mobile agent system, script language is widely used. Showing up java language, java makes an excellent environment by offering a portable, distributed, heterogeneous and secure function. So, we selected Java as a test implementation environment because it supports almost all of the required properties. To implement mobile agent, we used Grasshopper1.2 library of IKV++ GmbH[9]. This is a kind of experimental mobile agent systems. Grasshopper is an agent platform that is entirely implemented in Java, running on platforms supporting JDK1.2. To efficiently implement mobile agents, we partly get private OID(Object Identity) values of 50 in POWWOW-NMS’s 1000 polling values. Figure 5 graph shows a comparative analysis of their performance. By using mobile agents, especially clustering method, we could get 20-30% performance improvement rather than SNMP based polling. We determined the performance of SNMP polling and mobile agents by measuring the response time, the time between the launch and the return of a mobile agent from NMS. The response time depends on several factors, such as CPU speed each time and network environment. However, we pay no attention to these factors. By trying to test 20 times, we get average response time. The overall response time consists of network transmission time, NMS processing time and agents’ processing time. The response time has results similar with SNMP method and Peer to peer mobile agent method. As network entities increase, the overall response time drastically increases by using SNMP polling method by which network traffic and processing time have been Proceedings of the 22 nd International Conference on Distributed Computing Systems Workshops (ICDCSW’02) 0-7695-1588-6/02 $17.00 © 2002 IEEE ůũ űŭ Ů ūũ ŭũ Ŭ Ū caused. However, if managed entities are minor, SNMP polling method is better than any other methods. The performance results in a significant improvement in response time by increasing managed entities. As a result, even though most mobile agent based infrastructures exhibit scalability limitations when data intensive management applications are considered, the performance using mobile agent surpass SNMP-based polling performance because mobile agents gather distributed data at a once. MNAS(Mobile Network Access Subsystem) provides wireless, wireline integrated services. which 7. References [1] Young-il Kim, “An Implementation of Centralized O&M System for CPS Network”, APNOMS, September 1998. [2] Dittrich, A, Hoft M., “Integration of a TMN-based management platform into a CORBA-based environment”, Network Operations and Management Symposium, 1996., IEEE , 1996. [3] Agoulmine, N.; De Souza, J.N., “A general implementation architecture for TMN systems”, GLOBECOM '94. Communications: The Global Bridge., IEEE , 1994. [4] Lakshmi Raman, “OSI Systems and Network Management”, IEEE Communications Magazine, Volume: 36 Issue: 3, March 1998. Figure 5. Comparison of response time 6. Conclusion This paper introduced trends of PC communication service and in order to support PC dial-up service, referred to how ICPS/AIPCS had been processed and developed. We presented the design and implementation of powwow-NMS for PC communication network management. We also had experimental test and evaluated mobile agent model in POWWOW-NMS. Mobile agents will have a great potential to influence the design of future network managements, since they represent an alternative to the traditional client/server model. However, agent technology will probably not replace traditional client/server computing environment entirely rather than just regarding “add on” existing platform. Distributed network management is becoming essential due to the growing trend in internetworking and the complexity of network entities. Even though dial-up service’s usage decreases, network management is important to be effectively used in deployed network resources and to be continuously improved in the area of performance, flexibility, scalability and reliability today. In order to use existing ICPS/AICPS system, the research is going on additional functions such as VOIP(Voice over IP) which allows voice information to be transported over the packet network, CNAS(Common Channel Signaling No.7 Network Access Subsystem) which supports No.7 signal instead of R2 signal and [5] Danny B. Lange and Mitsuru Oshima, “Seven Good Reasons for Mobile Agents”, Communications of the ACM, March 1999. [6] Hosoon Ku, et al., “An Intelligent Mobile Agent Framework for Distributed Network Management”, Globecom ’97 Phoenix, AZ. Nov3-8, 1997. [7] Markus Breugst, “On the Usage of Standard Mobile Agent Platforms in Telecommunication Environments”, IS&N, May 1998. [8] Markus Breugst, Thomas Magedanz, “Mobile AgentsEnabling Technology for Active Intelligent Network Implementation”, IEEE, May 1998. [9]IKV++ GmbH, “Grasshopper : A Platform for Mobile Software Agents”, 1999. Proceedings of the 22 nd International Conference on Distributed Computing Systems Workshops (ICDCSW’02) 0-7695-1588-6/02 $17.00 © 2002 IEEE