Query evaluation techniques for large databases

Reviewer: Margaret H. Dunham

Graefe has written the first comprehensive study of database query processing techniques. It is an extremely well written overview of various query processing problems and solutions. As stated in the introduction, its target is to survey algorithms for complex queries. Included are discussions of sorting and hashing issues as related to query processing; methods to reduce I/O costs; techniques for performing aggregation; techniques to perform queries based on universal quantification (division is usually used for relational databases); scheduling and execution of complex queries, particularly in multiprocessor systems; issues related to parallelism; and techniques for processing queries against nontraditional database systems (nested relations, object-oriented databases, and temporal data). Included throughout the paper are performance results based on actual experiments or analytic formulas, and tables classifying and summarizing the various options or aspects involved in each of the areas covered. The discussion of query execution engines is interesting and provides a unique treatment of the subject. Overall, the paper is complete and the breadth of coverage is impressive. Previous survey papers have examined many related issues, so this paper overlaps other recently published papers. My recent ACM Computing Surveys paper with Priti Mishra examined join processing [1]. Section 5 of Graefe's paper examines joins in detail. The emphasis of this paper, however, is on more complex joins and their performance, while our paper is more of a survey and categorization of different types of joins, with an emphasis on natural joins and equijoins. The introduction cites many other surveys that have covered other topics related to database query processing, including concurrency control, recovery, views, query languages, and compression. While the overlaps have been kept minimal, the fact that some of this survey material is missing means that the target audience is readers very familiar with database concepts. The introduction indicates that the paper assumes “the reader possesses basic textbook knowledge of database query languages.” I find this to be an understatement, however. Experience with database programming and design would be of great benefit in understanding all issues covered in this paper. A major problem is that the author provides no coverage of optimization. Throughout the paper, references are made to I/O cost, but no discussion of optimization is included. The introduction does indicate that this is the case, but due to the other topics covered, this omission is a problem. Another major problem with this paper is its length. The breadth of the paper is astonishing. The size seems to be more appropriate for a book than for a paper published in a journal. I find it difficult to believe that anyone would sit down and read through the entire paper. As a reference tool, it is excellent, however. This excellent paper covers many questions related to the efficient implementation of complex database queries. The length of the paper makes it most appropriate as a reference.