一种基于分布式代理的网络管理重构机制

为了提高网络管理系统自身的可靠性，提出一种基于分布式代理的重构容错机制。该机制在集中式管理域内，设置多个具备管理者部分或全部功能的分布式代理实体作为冗余，保证了在域管理者失效时，能通过分布式代理的故障检测模块触发重构，从而选出负载性能最好的分布式代理来替代原管理者，并重构管理域。此外，还对分布式代理实体的结构进行了说明。

Reconstruction Mechanism Based on Distributed Agents for Network Management

To improve the reliability of network management, a reconstruction mechanism based on distributed agents is put forward. In a centralized management domain, domain manager is the most important node. So there are a number of distributed agents that own part or entire function of domain manager as management redundance.Through the redundance, it is assured once the domain manager is in invalidate state, the distributed agents elect the best load performance agent to replace primary manager and reconstructed management domain. To illustrate process of reconstruction, a partial synchronous system model is presented for describing distributed agents. Based on the model, a fault-tolerant reconstruction algorithm is put forward, which is efficient and tolerates crashed agents and lossy links. To meet reconfigure trigger, the heartbeat failure detector is implemented and several crucial parameters are discussed. The structure for distributed agent entities is suggested as well.