导弹全寿命周期大数据分布式存储系统研究

Research on Large Data Distributed Storage System of Missile Life Cycle

针对Hadoop HDFS存储大数据性能低下的问题,结合导弹全寿命周期大数据存储性能要求,以不同类型的分布式应用为依据,提出一种由大数据分布式存储层、典型缓存策略层、典型数据访问模式层和导弹上层分布式应用层构成的四层次分布式存储系统。试验表明:该系统可实现导弹全寿命周期大数据的稳定快速存储;存储较小数据时平均时间高于WR算法2%,内存占用率降低32%;深度学习框架下的存储效率具有更大的缓存命中率和缓存效用。

Combined the requirement of large data storage for missile life cycle and different types of distributed applications,a large data four-tier distributed storage system which consists of large data distributed storage layer,typical cache strategy layer,typical data access mode layer and missile upper distributed application layer is proposed.The experimental results show that the system can achieve stable and fast storage of large data in missile life cycle;the average time of storing small data is 2%higher than that of WR algorithm,and the memory occupancy rate is reduced by 32%,the storage efficiency under deep learning framework has greater cache hit rate and cache utility.