Influence Control for Dynamic Reconfiguration

Dynamic reconfiguration is a useful technique for software update because it can achieve an architectural change without shutdown of a system. However, so far in the state-of arts, there has not been an approach that can evaluate and control both the functional influence and performance influence of reconfiguration in a unified framework. In this paper, we present an approach that addresses the above drawback. In our approach, we use a reconfiguration algorithm and a reconfiguration scheduler to control these two types of influence. The algorithm reduces the logical performance influence by allowing old and new components coexisting and uses a version management mechanism to avoid functional side effect in the coexisting period. The scheduler controls the physical performance influence through restricting the processor time spent on the reconfiguration procedure. We implement the algorithm and the scheduler in our reconfiguration data flow model.