An automatic clustering technique for query plan recommendation
| Autor: | Mehdi Hosseinzadeh, Aso Mohammad Darwesh, Nima Jafari Navimipour, Arash Sharifi, Elham Azhir |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
DBSCAN
Information Systems and Management Distributed database Computer science 05 social sciences 050301 education Dunn index 02 engineering and technology Reuse computer.software_genre Query optimization Computer Science Applications Theoretical Computer Science Query plan Artificial Intelligence Control and Systems Engineering Schema (psychology) 0202 electrical engineering electronic engineering information engineering 020201 artificial intelligence & image processing Data mining Cluster analysis 0503 education computer Software |
| Zdroj: | Information Sciences. 545:620-632 |
| ISSN: | 0020-0255 |
| DOI: | 10.1016/j.ins.2020.09.037 |
| Popis: | The query optimizer is responsible for identifying the most efficient Query Execution Plans (QEP’s). The distributed database relations may be kept in several places. These results in a dramatic increase in the number of alternative query’ plans. The query optimizer cannot exhaustively explore the alternative query plans in a vast search space at reasonable computational costs. Henceforth, reusing the previously generated plans instead of generating new plans for new queries is an efficient technique for query processing. To improve the accuracy of clustering, we’ve rewritten the queries to standardize their structures. Furthermore, TF representation schema has been used to convert the queries into vectors. In this paper, we’ve introduced a multi-objective automatic query plan recommendation method, a combination of incremental DBSCAN and NSGA-II. The quality of the results of incremental DBSCAN has been influenced by Minpts (minimum points) and Eps (epsilon). Two cluster validity indices, Dunn index and Davies–Bouldin index, have simultaneously been optimized to calculate the goodness of an answer. Comparative results have been shown against the incremental DBSCAN and K-means regarding an external cluster validity index, namely, the ARI. By comparing different types of query workloads, we’ve found that the introduced method outperforms the other well-known approaches. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect