Parallel and Persistent Object-Oriented Database Programming System for Advanced Database Applications

在实现持久化编程语言和面向对象数据库时，二级存储上的引用格式通常不同于虚拟内存指针。因此，在实现这些系统时通常使用一种称为指针旋转的技术。指针旋转的变体将所有引用替换为虚拟内存指针，潜在地浪费虚拟内存空间。另一方面，在整个程序执行过程中使用长引用会对性能产生不利影响。我们的研究开发了另一种变体，它使用我们所称的代理OID。该方案采用了Paul Wilson的页面错误指针抖动和发现时的指针抖动，这是威斯康星大学在Exodus/E实现中所采用的。我们的研究设计了一种基于C语言的支持永久变量声明的语言，并通过对Gnu C编译器的修改实现了一个编译器。P3L使用代理OID，并在驻留检查方面引入了编译器优化。为了评估使用代理OID在效率方面的优势，我们在Cattel/OO1基准测试中运行了前向遍历基准，并定量验证了该方案比在虚拟内存中使用96位指针的性能提高了30%到40%。

When implementing persistent programming language and object oriented database, the format of references on secondary storage usually differs virtual memory pointer. thus a technique called pointer swizzling is often employed in implementing those systems.A variation of pointer swizzling, which replaces all references with virtual memory pointer, potentially wastes virtual memory space. on the other hand, using long references through whole execution of a program affects performance adversely. Our research developed another variation which uses surrogate OID as we call. The scheme is in the middle of pointer swizzling on page fault by Paul Wilson and pointer swizzling upon discovery which is employed in Exodus/E implementation by Wisconsin university. Our research designed a C language based language which allows persistent variable declaration and implemented a compiler through modification to Gnu C Compiler. P3L uses surrogate OID as well as introducing compiler optimization in terms of residency checking. In order to assess advantage of the use of surrogate OID in terms of efficiency, we run the forward traversal benchmark in Cattel/OO1 bench mark and quantitatively verified that the scheme achieves 30% to 40% improvement over the use of 96bit pointer in virtual memory.