In recent years, a Cloud computing system has been popularly used. Among them, a hybrid Cloud is focused, which combined a public Cloud operated by service providers and a private Cloud constructed inside a company. Because a private Cloud is considered to be secure and a public Cloud is scalable, it is possible to build a Cloud computing system that works effectively by combining the both types of Clouds. However, when a big disaster occurrs, huge volume of data is produced by monitoring censors and users, and flowed into such an information infrastructure. In addition, enormous number of people access to the system in such a case, thus a load of the system becomes extremely high in a short time. Therefore, it is important to change an environment dynamically among inter-Cloud and intra-Cloud to deal with the load. In this paper, bursty increasing load is predicted based on Earthquake Early Warning (EEW), and a speculative control is performed in a short time between a time of occurrence of an earthquake and a time when the system is heavily loaded actually. In addition, network tra ffi c is controlled to give a higher priority to important data replication. The system is constructed on a Cloud computing platform using public domain software, OpenStack in this experiment. Network tra ffi c of the system is controlled by Software Defined Network (SDN), which is controlled by the OpenFlow protocol. As a result of an evaluation, our proposed system works fine and achieves good performance.
展开▼
