An Approach for Self-repair in Distributed System Using Immunity-Based Diagnostic Mobile Agents

In recent years, self-repair has attracted much attention again for fault tolerance in distributed computer network system. In our previous research, abnormal units (host computers and mobile agents) in distributed system can be identified by the immunity-based diagnostic model. The model calculates the credibility of each unit. In this paper, based on the credibility, some units try to self-repair, that is, replace their data with data received from other units. We propose three different repairing conditions: majority vote without credibility, credibility comparison and credibility threshold. Some simulations evaluate the effectiveness of these conditions changing the numbers of initial abnormal hosts.