Readings in Database Systems

The book is an outstanding compilation of influential database papers in the database world. Most of the papers are readable for a senior undergraduate audience, or a graduate audience, but probably not for a first course in database systems, simply because students would need a decent understanding of database principles to fully appreciate the contents of this book. In particular, I like the introductory commentary that Hellerstein and Stonebraker provide for each of the chapters. These are very nice summaries, and they also provide a teaser for the papers that follow in the chapter, and they justify why those papers were included. Both of the editors have the background to speak authoritatively on the subject of database systems, including OS, network, and architectural issues. The editors are not afraid to make prophetic comments about the direction of database research, and the challenges ahead, especially with respect to the Internet and the explosion of data. I love the two introductory papers by Stonebraker and Hellerstein: "What Goes Around Comes Around" (about 40 pages), and "Anatomy of a Database System" (about 50 pages). The first is a summary of the major models of a DBMS: hierarchical, network, and relational; followed by other models of a DBMS, database, or data: entity-relationship, extended relational, semantic, object-oriented, object-relational, and semi-structured (XML). I intend to give this paper as a reading assignment to my senior undergrads (part of a 2nd database course). The Anatomy paper is a wonderful overview of a DBMS from a systems perspective (e.g., buffering, I/Os, threads, dispatching, bottlenecks, hardware architectures, parallelism, query rewriting, optimization, etc.) I'll have to see how this paper fits into the scheduling of the course topics before assigning it, since many of its topics require an understanding of both systems and database principles for full appreciation. This book is definitely a keeper on my shelf. I'll revisit it many times.

I am somewhat biased in that I'm taking a course which is co-taught by Hellerstein, however if you're looking to complete understand how database or transaction management systems: this will be an invaluable source of information. Crammed into this book are a number of papers detailing the history and state of database design as it has evolved over the last 40years, including most of the semainal papers in the field. Don't expect to read this sucker cover-to-cover, take your time, but it's a good compilation.

This book is well-known in academic database circles as "The Red Book" (in fact, the editors maintain a useful web site at http://redbook.cs.berkeley.edu/) and is a canonical resource in those circles. It's primarily intended as a reader/textbook for a graduate course in database systems and has a heavy emphasis on implementation issues. It contains a fair number of classic papers that should be read by anybody who actually works on database engines as well as a number of more recent papers that should be read by anyone who does research in database systems. The usefulness for end-users of databases (i.e., application writers) is unclear.The 3rd edition, in my opinion, improves upon the 2nd edition considerably. Of course, it freshens the paper selection in some areas. More importantly, it prunes the number of subject areas considerably, resulting in a more manageable collection (in more ways than one!). For example, a great deal of work was performed in the late 1980s and early 1990s in areas such as extensibility and active database management. By the late 1990s, the SQL3/SQL1999 train had already left the station - work still goes on in these areas, but at a greatly reduced rate. Conversely, data mining and decision analysis have become hugely important areas, and the new Red Book has a section on it.If there's a place where this book "missed the boat," it would probably be in terms of applications. The editors cut the section on user interfaces and programming models and have always ignored unstructured/semistructured data models. In these days of the Web, this choice is questionable; on the other hand, a lot of the most reasonable work in these areas has in fact appeared since 1998, so it's a bit hard to criticize with any degree of fairness!