Scientific workflow management systems are become more and more important in e-Science because they provide support for scientists in computer simulations, experiments and calculation. In the process of developing a workflow, the workflow would be modified and rerun many times. Re-executing all the preceding, unchanged steps in the workflow may be redundant and time consuming. In this paper, we propose a smart rerun strategy applies to most scientific workflow management systems. In order to gain better support for PaaS-based scientific workflow platform, we improve the smart rerun strategy by using caching technique.