论文部分内容阅读
随着云计算系统对块级别云存储服务使用规模的增长,云存储子系统承受的负载也随之增长,严重影响其I/O性能.针对上述问题设计并实现了适用于多节点的块级别网络磁盘缓存系统MN-cache.设计了MN-cache缓存系统的系统架构,包括处理元数据的元数据服务器端以及实现缓存功能的缓存客户端.基于Linux Device Mapper机制实现了MN-cache缓存系统的客户端缓存机制,将局域网内各缓存客户端节点的本地缓存磁盘作为共享缓存,利用局域网及本地缓存磁盘的性能优势提升云存储子系统的性能.根据缓存过程各阶段的耗时情况对MN-cache缓存系统进行了性能分析,并通过实验验证了性能分析结果.实验结果表明:MN-cache缓存系统在网络存储系统I/O性能较差的情况下对存储系统的整体性能提升较大.
With the cloud computing system increasing the use of block-level cloud storage services, the load on the cloud storage subsystem also increases, seriously affecting its I / O performance.According to the above problems, designed and implemented for multi-node block level Network disk cache system MN-cache.A system architecture of MN-cache cache system is designed, which includes the metadata server to process metadata and the cache client to realize cache function.An architecture based on Linux Device Mapper is proposed to implement MN-cache cache system Client caching mechanism improves the performance of cloud storage subsystem by using the local cache disk of each caching client node in the LAN as the shared cache and utilizing the performance advantage of the local area network and local cache disk.According to the time- cache performance analysis of the cache system and verify the performance analysis results through experiments.Experimental results show that: MN-cache cache system in the network storage system I / O performance under the conditions of the overall performance of the storage system to enhance the larger.