Object Store Garbage Collection

该项目调整和扩展了现有的逐块主存垃圾收集方案，称为成熟对象空间，用于收集持久对象存储。该方案将定位并回收包含不再可访问的数据的数据库内存。这将使程序员从繁重且容易出错的任务中解脱出来，即指示哪些内存可以重复使用，从而允许更好的编程风格。该算法是第一个逐块工作的算法(这对于很好地与数据库集成很重要)，并保证找到所有不可访问的内存。它还可以帮助重新组织(重新集群)数据，从而提高存储性能。该算法是增量的，具有粗粒度(一次收集一个块)。这项研究的一个重大贡献是算法本身。也许更重要的将是算法的实施和测量，以及如何纳入重新聚类的调查。由此产生的存储管理技术将适合纳入实际的数据库管理系统，确保存储不会“丢失”，但最终会“找到”和“回收”以供进一步使用。这些技术将在很大程度上独立于存储管理策略，允许数据库实施人员和应用程序构建者选择最适合其需求的策略。

The project adapts and extends an existing block-by-block main-memory garbage collection scheme, called Mature Object Space, for use in collecting persistent object stores. The scheme will locate and reclaim database memory containing no longer accessible data. This will relieve programmers of the burdensome and error-prone task of indicating which memory can be reused, thus allowing better programming styles. The algorithm is the first to work block by block (important for integrating nicely with databases) and to guarantee to find all inaccessible memory. It can also help reorganize (recluster) data and thus improve store performance. The algorithm is incremental, with a coarse granularity (collecting one block at a time). A significant contribution of the research is the algorithm itself. Perhaps more significant will be the implementation and measurement of the algorithm, and investigation of how to incorporate reclustering. The resulting storage management techniques will be suitable for incorporation into practical database management systems, guaranteeing that storage is not `lost` but eventually `found` and `recycled` for further use. The techniques will be substantially independent of storage management policy, allowing policy to be chosen by database implementors and application builders to best suit their needs.