基于虚拟单元可智能增长的内存池研究

Smart-growth and virtual-unit -based memory pool

针对内存数据库系统对空间利用率和系统健壮性的要求,提出了一种新型的基于虚拟单元可智能增长的内存池(SVMP)。该内存池吸收了传统内存池的优点,改进了内存管理策略,提出了对连续内存区进行逻辑划分以提高空间利用率的虚拟单元和一种以AIMD(Additive Increase Multiplicative Decrease)为核心的智能增长算法,并通过C++的new-handler机制解决了内存池增长中可能会出现的内存不足的问题。理论分析和性能测试表明,该内存池结构具有良好的时间、空间特性和健壮性,能够显著提升内存数据库系统的运行效率。

Main-memory database system requires good space utilization and system robustness. Based on the advantages of the traditional memory pool, a new memory pool structure named Smart-growth ＆amp;Virtual-unit-based（SVMP）is pre-sented. SVMP improves the original memory management strategy according to a new concept of virtual-unit, which in-creases the space utilization rate via logic partitioning in the contiguous memory area, and a smart-growth algorithm with AIMD as its core. It can solve the problem of insufficient memory which might turn up in the process of memory pool growth through the new-handler mechanism of C＋＋. Theoretical analysis and performance testing show that it can signifi-cantly improve operation efficiency of main memory database system.