Three-Layer MPI Fault-Tolerance Techniques
Guo Yucheng, Peng Wu, Xiaoyi Tang, Guo Qingping · 2013
To make the MPI adapting to the data-intensive tasks, the MPI must own perfect fault-tolerance capabilities to handle errors. This paper proposes a three-layer fault-tolerance technique that can achieve this purpose excellently. The MPI task dynamic migration technique proposed and implemented by our Lab has rarely seen in literatures.