MDARTS: A Multiprocessor Database Architecture for Real- Time Systems

先进的实时系统，如用于自动化工厂的下一代工作站/机器控制器(NGC)，需要内置的数据库支持来进行并发数据访问。然而，传统的数据库系统不提供这些实时应用所需的性能水平或响应时间保证。为了满足这些需求，正在设计、实现和评估一个面向对象的软件系统，称为实时系统多处理器数据库体系结构(MDARTS)。MDARTS支持对象粒度级别的实时需求和语义约束的显式规范。与其他方法不同，MDARTS具有内置策略，可以在应用程序初始化期间在运行时检查这些规范，并在每个对象的基础上动态调整其数据管理配置，以提供实时保证。为了在共享内存多处理器上实现最高性能，MDARTS支持并发的、直接的共享内存数据访问。这种独特的设计使事务执行速度比当前的实时数据库系统快两到三个数量级。总而言之，这个项目的贡献是双重的。第一，将引入新的数据库技术，包括提供概率和硬实时保证的方法，以及实施开放的多处理器数据库体系结构，提供极高的性能水平。第二，将MDARTS技术应用于实际制造机床的分布式、开放式体系结构控制器的开发，评估其在满足制造应用程序的高性能实时数据管理需求方面的有效性。

Advanced real-time systems, such as the Next Generation Workstation/Machine Controller (NGC) for automated factories, require built-in database support for concurrent data access. However, conventional database systems do not provide the performance levels or response time guarantees needed by these real-time applications. To address these needs, an object-oriented software system called Multiprocessor Database Architecture for Real-Time Systems (MDARTS) is being designed, implemented, and evaluated. MDARTS supports explicit specification of real-time requirements and semantic constraints at an object-granularity level. Unlike other approaches, MDARTS has built-in strategies to examine these specifications at runtime during application initialization and to dynamically adjust its data management configuration on a per-object basis to provide real-time guarantees. For maximum performance on shared-memory multiprocessors, MDARTS supports concurrent, direct shared-memory data access. This unique design achieves transaction execution times two to three orders of magnitude faster than current real-time d atabase systems. In summary, the contributions of this project are two-fold. One, novel database technologies will be introduced, including methodologies for providing probabilistic and hard real-time guarantees and the implementation of an open multiprocessor database architecture providing extremely high levels of performance. Two, MDARTS technology will be applied to develop distributed, open architecture controllers for actual manufacturing machine tools, evaluating its effectiveness for addressing the high-performance real-time data management needs of manufacturing applications.