Self-adaptive learning type indexing method for workload in memory database

Abstract

The invention discloses a self-adaptive learning type indexing method for workloads in a memory database. According to the method, a cardinal number tree and a piecewise linear model with a maximum error bound are combined, memory occupation of indexes is reduced by utilizing data distribution through a machine learning model, and meanwhile stable query performance is kept. On this basis, an efficient insertion buffer is used to reduce the cost of data insertion update, and in order to alleviate the influence of data insertion on index performance, two workload adaptive recombination optimization methods are used to optimize hotspot data involved in point query and range query in workloads in a targeted manner. The method has relatively high construction efficiency and relatively low memory occupation, relatively efficient query performance is also ensured, and insertion and updating can be well supported; And meanwhile, recombination optimization is performed in a targeted manner by sensing the query workload, so that the influence of insertion on the index performance is reduced at a relatively low cost.

Description

Technical field [0001] The present invention belongs to the technical field of database systems, and in particular to a high efficiency index method in a memory database. Background technique [0002] The index is used as a data structure for efficient retrieval, and has been widely used in the memory database system. However, due to the explosive growth of the amount of data, the number of memory consumptions in the index affects the performance and scaife of the database system. Reduce the memory of the index, while maintaining the efficient and stable query performance of the index has important practical significance and application value. [0003] The proposal of the learning index provides a new opportunity to solve the above problems, and the size of the index can be significantly reduced by utilizing data distribution. But the biggest problem with learning index is to weaken the insertion and update, which greatly limits the use scenario of the learning index. The existin...

AI Technical Summary

Problems solved by technology

However, the biggest problem with learning indexes is that they have weak support for insert and update, which greatly limits the usage scenarios of learning indexes.

Existing research work on inserting and updating learning-type indexes often requires a large amount of movement of key values ​​or a large-capacity buffer, which will bring large overhead and increase the delay of various operations, while the memory database is essentially for the pursuit of Designed for faster system response, reducing the cost of data insertion and update is very urgent for the application of learning index in memory database

Images