How Many DBAs?
Figuring out how to staff the DBA group is not a simple matter. There are several non-trivial considerations that must be addressed including the size of the DBA staff and the reporting structure for the DBAs. In my book, Database Administration: The Complete Guide to Practices and Procedures, I discuss many of the issues that impact DBA staffing levels.
Basically, it boils down to complexity. The more complex your environment, the more DBAs you will need to assure availability, ensure data integrity, optimize performance and basically keep your database systems and applications operational. Of course, it is very difficult to combine all of the mitigating factors into a specific formula that will dictate the optimum number of DBAs to employ.
Some industry analysts have tried. A few years ago analysts at META Group published a loose formula for calculating DBA level of effort. Their formula applied weights to six factors: system complexity, application immaturity, end-user sophistication, software functionality, system availability and staff sophistication. By measuring each of these items as much as possible to indicate high or low rates, META showed how you could plug in values to the formula and arrive at a a number that is tranlated into an estimate for the number of DBAs required. Of course, this research is dated and META Group has since been acquired by Gartner Group. And, really, the formula was not comprehensive enough to do an "exact" job on number of DBAs needed.
I've heard of DBA groups who use a simple formula like 1 DBA per 10 database instances. But how useful is that, really? A very experienced DBA could easily handle double that if the activity is low and the systems are well tuned. But if activity is high, the databases are not well designed or SQL is coded inefficiently, that DBA might be stretched to his (or her) limits. And the less experience the DBA has, the fewer instances that can be managed.
Of course, things are different in the mainframe world, too. A single DB2 subsystem would probably be used for multiple databases and applications; setting the same thing up in an Oracle/Unix environment would likely be done with multiple instances. So you have to factor in the platform being used, too.
The bottom line is: you just can't put a simple formula on top of a complex problem.
How does your shop figure out how many DBAs it needs?
