多处理器环境下降低实时数据库系统更新负载的关键技术研究

项目名称： 多处理器环境下降低实时数据库系统更新负载的关键技术研究

项目编号： No.61300045

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 自动化技术、计算机技术

项目作者： 李剑军

作者单位： 华中科技大学

项目金额： 28万元

中文摘要： 在保证实时数据对象时序一致性的前提下降低更新事务负载对于实时数据库系统（RTDBS）减少资源（包括CPU和能量等）消耗，提高服务质量具有重要意义。单处理器环境下该问题已经得到了广泛研究，出现了大量研究成果，而多（核）处理器环境下针对该问题的研究则几乎还是空白。鉴于目前多处理器的普遍性及其在性能功耗比上的优越性，研究多处理器环境下如何降低RTDBS的更新负载是迫切和必要的。本项目分析多处理器环境下RTDBS中影响更新负载的关键因素及它们的相互关联，研究基于任务划分方式的实时更新事务在各处理器上的分配与调度策略，以及基于全局调度方式的更新事务优先级分派策略、事务周期与截止期计算算法；统筹考虑用户事务与更新事务，建立用户事务质量模型和数据质量模型，研究能够保证一定数据对象"新鲜度"和用户事务价值的混合事务调度策略，以尽可能地降低处理器更新负载，提高系统服务质量，并最终推动RTDBS的更广泛应用。

中文关键词： 实时数据库；多处理器；更新事务；时序一致性；负载

英文摘要： Maintaining temporal consistency of real-time data objects while minimizing the imposed update workload is of great significance for Real-Time DataBase System (RTDBS) to reduce resoure (including CPU and energy, etc.) consumption and improve quality of service. While this problem has been extensivley studied in single-processor environment and a lot of research results have been reported, the research on the same problem under multiprocessor platform is still nearly blank up to date. In view of the current universality of multiprocessor platform and the superiority of multiprocessor in performance per watt, it is urgent and necessary to study how to reduce the update workload in multiprocessor RTDBS. In this project, by taking a thorough analysis of the key factors that impact the update workload in multiprocessor RTDBS and their interrelationship, we study the real-time update transaction assignment problem based on partitioned multiprocessor scheduling, and the priority assignment problem along with the period and deadline derivation problem of update transactions based on global multiprocessor scheduling. By taking both user and update transactions into consideration, we study a hybrid real-time transaction scheduling policy that can guarantee the freshness of real-time data objects and the value of user tran

英文关键词： Real-time database；Multiprocessor；Update transaction；Temporal consistency；Workload