高算发集群监控系统中内存数据库的设计与应用

摘要

In high-concurrent cluster monitoring system with large amount of concurrent connections, traditional disk resident database systems are unable to store and process data in real time due to considerable overheads caused by I/O swapping operations.Therefore many realtime systems adopt main memory database systems as data support modules.Proceeding from introducing key techniques of main memory database, we propose a design and implementation of an efficient main memory database module that is used in underpinning high-concurrent cluster monitoring system through importing virtual shadow memory and coarse-grained intent share lock to improve data organisation and concurrency management of main memory database.Its application status in practical system is studied as well.%在具有大量并发连接的高并发集群监控系统中,传统磁盘数据库由于内外存交换开销过大,无法支撑数据的实时存储与处理,因此大量实时系统都选择采用内存数据库作为数据支撑模块.从介绍内存数据库的关键技术点出发,通过引入虚拟影子内存和粗粒度意向锁来分别改进内存数据库的数据组织和并发控制,设计实现了一个用于支撑高并发集群监控系统的高效内存数据库模块,并且研究了其在实际系统中的应用情况.