The Buffering Techniques In Parallel Database System PBASE/3

In the early days of parallel database research, most papers focused on the physical organization of parallel database, parallel algorithms, optimization and scheduling strategies. There are few papers working on the execution mechanism. Actually how to manage the system resource efficiently, especially the data buffer is the key to higher performance of parallel database. In this paper, I will focus on the data buffering techniques in parallel database system PBASE/3.According to the data granularities and the sharing level of the buffered data, I categorize the data buffering technique into four types: buffering of base table, buffering of temporary table, buffering of repetitious expensive operator ( cache ) and buffering of multi-query's common operator.In the first part of the dissertation, I analyze the access modes of the database operators in PBASE/3, and present a detailed intra-operator data buffer management strategy according to the different access modes of the operators. After that, I take more factors into consideration, such as query optimization information, inter-operator competition etc... , and present a total solution for base table buffer management.The second part of this dissertation focuses on the intra-query data buffering. I divide this problem into two parts: temporary table buffering and repetitious expensive operator buffering. I analyze the basic access mode of temporary table, distinguish the difference between the temporary table and the base table buffering, and give an unified buffer management strategy.The CACHE technique aims at the repetitious expensive operator. My...