An Implementation Scheme of Relational Tables Based on History Offset Method and Its Evaluation

本年度在去年开发的历史偏移法实现的关系数据库内核霍特的基础上，设计、实现和评价了关系数据库系统的高层子系统方案。这些计划包括：1.范围键查询的处理方案; 2.并行处理方案，3. SQL处理器的设计与实现; 4.复杂对象的设计。1.在我们提出的方案中，由于记录可以被映射到一个多维可扩展数组的元素和元素的位置被映射到线性地址空间，维度依赖性问题将导致在范围键检索。针对霍特数据结构的特点，设计了一种解决方案。利用实现的原型系统，对范围键查询的性能进行了测试和评价.可扩展数组是霍特的底层数据结构，适合于并行处理。针对霍特的这一优良特性，我们设计并实现了一个负载均衡方案和检索结果收集同步方案。利用实现的原型系统，对并行检索方案的性能进行了测试.我们现在正在建设的SQL处理器纳入范围关键字查询处理方案在1。特别是设计并实现了一种高效的表连接操作处理方案.复杂对象具有另一个对象的对象ID作为其引用属性。我们提出了一个有效的处理方案，复杂的对象集，通过引入记录ID的概念到原来的霍特数据结构。

In this year, based on the relational database kernel implemented by history offset method called HORT developed in the last year, schemes of higher level subsystems of relational database system have been designed, implemented and evaluated. These schemes include: 1. Processing scheme of range key queries, 2. Parallel processing scheme, 3. design and implementation of SQL processors, 4. Design of complex objects.1. In our proposed scheme, since a record can be mapped to an element of a multidimensional extendible array and the element position is mapped to linear address space, dimension dependency problem would be caused in a range key retrieval. This problem has been alleviated by designing a scheme considering the characteristics of HORT data structure. Using the implemented prototype system, the performance of range key queries has been measured and evaluated.2. An extendible array, which is the underlying data structure in HORT, is suited to parallel processing. Focusing on this good property of HORT, we have designed and implemented a load balancing scheme and synchronizing scheme of collecting retrieval results. Using the implemented prototype system, the performance of parallel retrieval scheme has been measured.3. We are now under construction of SQL processor incorporating the range key query processing scheme in 1. Especially an efficient scheme of processing table join operation is being designed and implemented.4. A complex object has an object ID of another object as its reference attribute. We have proposed an efficient handling scheme of complex objects set by introducing the notion of record ID into the original HORT data structure.