摘要

As the developing of cloud computing, internet of things and mobile internet, the whole society has entered into the big data era, especially the high-performance applications whose dataset size become more and more huge. In this paper, we are focus on the HPC applications and have presented a data storage and analysis framework for big data by utilizing the distributed mixed storage and data mining technologies. We have practiced this framework in the photovoltaic forecasting system to prove its practicalness, feasibility, availability and expandability. By adopting abundant experiments, we shows that the framework can provide higher forecast accuracy- achieve 85% and low latency by deploying the mixed distributed storage architecture.