JNT Visual - Fotolia
Craig Mullins has seen a lot of changes in the duties of database administrators in his more than 30 years in database management, and he shared tips and advice based on his experiences in a book for DBAs that was first published in 2002 and updated in 2012. Database Administration: The Complete Guide to DBA Practices and Procedures is a compendium of database management system (DBMS) know-how and guidance on DBA responsibilities. SearchDataManagement talked with Mullins, who is president and principal consultant of Mullins Consulting Inc. in Sugarland, Texas, about the book's second edition and the many changes currently afoot in the DBA environment. This is part one of a series; also read part two on rules of thumb for DBAs.
How did you approach the second edition of Database Administration?
Craig Mullins: The book is very much about heterogeneous database administration -- knowing what to do as a DBA regardless of the kind of DBMS you're using. It has become increasingly true that a lot of people have to manage both Oracle and DB2, or DB2 and MySQL or SQL Server. They could benefit from a discussion about the tasks they needed to do, rather than an instance-by-instance checklist that had to be accomplished. There [are] a lot of books that show you how to manage Oracle or how to work with SQL Server. But there wasn't really anything that looked at the core responsibilities of a DBA -- the things DBAs had to get their arms around in order to do the job.
What are some of the big changes you've seen in database administration since the first edition came out?
Mullins: DBAs have been forced more into the role of protecting data. That may mean auditing -- who did what to which data when -- or it may mean protecting data more on the front end from any access at all. Security has always been a built-in part of managing a DBMS, but 10 or 15 years ago it wasn't quite as interesting a topic for DBAs. Now most new apps are not monolithic, and a lot are Web-based, and security is one of the first things that DBAs look at. They're being forced from the top down to really make sure they're doing a good job of protecting corporate data from theft. You want to make sure you're doing as much as possible in the back end on the server to make sure the data isn't co-opted there.
You touched on the trend of NoSQL databases in your book, and suggested that DBAs should begin to learn about them. How do you think NoSQL software -- as well as Hadoop -- might impact DBA responsibilities?
Craig Mullinsconsultant and author
Mullins: If you look at the four types of NoSQL database models, it's not as new as people think. But it is more firmly embraced than it has been in the past. The wide column stores have been around for a while. The key-value stores basically say, 'Here's a key and here's a lump of data' -- that's VSAM, which has been around for over 40 years! And the document stores are basically the latest iteration of what used to be called object-oriented database management systems, which were once going to take over the world. The one NoSQL type that is more new is the graph database. It's new in the way people are adopting it to keep track of relationships. You might call it the Kevin Bacon database.
There is some benefit to some of these new data models in cases where you don't have the hardened use case for production around financial data. That's where relational database management systems have been so good. But there is not necessarily coherence of instance from one [data] item to another with NoSQL designs. From a developer's perspective, that's flexibility. From a data integrity perspective, it can become troubling. You have to ask, 'How do you protect data integrity from instance to instance?' With NoSQL, the developer's programs do the work. With a relational database management system, it comes with the package.
Still, the overhead that an RDBMS can use to impose consistency can sometimes be overkill. For that reason, I think you'll eventually see the major relational databases add NoSQL engines to their flagship products. The new stuff is going with the new data models, and in some cases, you're going to need people to administer it. Right now, with both NoSQL and Hadoop, security is an afterthought. But that will change, and at that point, the role of the DBA will come more into play.
Jack Vaughan is SearchDataManagement's news and site editor. Email him at firstname.lastname@example.org, and follow us on Twitter: @sDataManagement.
Learn how the cloud is changing familiar database rules
Find out what big data portends for the data management profession
Discover adjustments afoot as NoSQL challenges the RDBMS's dominance