面向移动通信大数据的云存储系统优化

Cloud storage system optimizations for telecom big data

移动通信大数据对底层云存储系统提出了三大苛刻的需求:高性能、高可靠、低成本,但是通用云存储系统无法满足移动通信业务的现实需求。针对三大需求,提出对应的三项优化技术:面对高性能需求,提出一种小文件读写访问性能优化技术;面对高可靠需求,提出一种日志写入与回放优化技术;面对低成本需求,提出一种固态硬盘垃圾回收优化技术。相关测试表明,在典型的移动通信大数据应用场景中,相较于优化前,性能最多提升54%,平均提升27%~37%;可靠性处理效率最多提升67%,平均提升18%~22%;成本最多降低80%,平均降低26%~42%。

Three major requirements on the cloud storage system were put forward by telecom big data： high performance, high reliability and low cost, which universal cloud storage cannot meet. For high performance requirement, a small file reading and writing performance optimization technology was put forward. For high reliability requirement, a journal writing and synchronizing technology was put forward. For the requirement of low cost, a solid-state hard disk garbage collection optimization technique was put forward. Related tests show that in the typical telecom big data application scenarios,compared with the system before optimization, the performance increases up to 54% at an average of 27%-37%, the reliability increases up to 67%, at an average of 18%-22%, the cost decreases down to 80%, at an average of 26%-42%.