Network management agents supported by a Java environment

Abstract

In this paper we investigate how a network management architecture might take advantage of the Java environment. Network management is by definition a distributed application. It should therefore be no surprise that a new language environment like Java that is purposely developed for the network environment should find a natural match in this application. Problems such as portability across platforms or independence from underlying hardware and software, security aspects such as authentication, enforcement of access restrictions, and preservation of the integrity of the communication process can be addressed easily in a Java environment. The navigator/browser architecture offers an ideal framework for the network management system applications. A Java-based network management system contains management applications based on the browser model and agents in the network elements running in respective local Java environments as agent processes. The manager/browsers monitor and control the network elements. The agent process at a network element performs the network management agent functions, including interaction with the manager. The communication between the manager and the agent is at present carried out by Java classes transported across a network in bytecode format. This approach avoids the currentuse of ASN.l Basic Encoding Rules in the SNMP environment. The Management Information Base-IT (MIB) standard is supported at the agent. The security features inherent in the Java architecture provide an additional layer of security in heterogeneous environments. The manager browser uses a Graphical User Interface, which contains different panels for various management functions. The user selects an agent and corresponding Mill variables for monitoring from the list of agents discovered by the manager browser. The manager encodes this information into a Java class and transmits it to the agent. The result of the query is displayed upon receiving the response from the agent. The main function of an agent is to respond to queries from a manager, parse the MIB tree and report an event to the manager by a trap. The agent is executed as a background process with a main thread listening to incoming requests and creating respective worker threads. The agent interacts with the MIB using the existing ASN.l standard. A system for experimentation has been created to demonstrate feasibility and carry out performance measurements. The performance of a Java agent and an SNMP agent based on a system from CMU have been compared, specifically the response times of the SNMP get command for system and ip groups. This response time depends primarily on the hardware platform on which the agent is running. Measurements were performed for a • manager running on a Sun Microsystems SparcStation 5 under Solaris and Java agents and SNMP agents. As is to be expected, compared to the "plain" SNMP implementation, the Java-based implementations show increases in response time of 132% (Solaris agent) and 241% (Windows95 agent) when doing system group...

Raciocinio baseado em modelo aplicado ao diagnostico de falha e performance em redes locais

“…Especialistas Baseados em Regras [13,22], Raciocínio Baseado em Casos [36. 2], Agentes Inteligentes [31] e Raciocínio Baseado em Modelos para correlação de alarmes [28].…”

“…Outras aplicações se estenderam para o diagnóstico automatizado de componentes eletrônicos/mecânicos [14,37]. Também o diagnóstico de falhas em redes de computadores recebeu atenção de técnicas de IA, um campo bem explorado [31,29,48,30], principalmente devido a demanda por ferramentas que automatizassem este processo, acelerando a resolução de um problema que um bom especialista levaria horas para fazê-lo. Algumas das ferramentas desenvolvidas como o DRUM-II [20] e IMPACT [28] são exemplos de aplicação de técnicas de Inteligência Artificial na correlação de alarmes , uma das etapas no processo de diagnóstico em redes de computadores.…”

“…Outras abordagens podem ser utilizadas no mesmo problema, um bom exemplo é a técnica de Raciocínio Baseado em Casos [2, 36], Agentes Inteligentes [31] ou os Sistemas Baseados em Regras [I 0].…”