• Home
• About
• Writers
• Landscape
• Wizards Index
• Fraud
• Wanna Be Like Larry?

Comparing Database Costs: Duesenberg Versus Hispano-Suiza

One of the many boutique consulting firms in New York is the Edison Group. The firm’s work is good, certainly better than the azure chip firms whose intellectual antics entertain me each week. I have to say that I am on the other side of the Hudson River when I think about the firm’s findings about database costs. You can read “Edison Group, Comparative Management Costs Study: Oracle Database 11g vs. IBM DB2 Enterprise 9.5”, which consumes 78 pages. The study carries a copyright of 2008 and a first publication date of January 2009. Today is September 18, 2009. My hunch is that the document was generated for a client and has been in circulation for months. I just learned about it, and I made an effort to read it with its nine month old data in mind. I don’t have any quibble with the charts, graphs, and numbers in the report. When one writes about traditional databases, the costs are irrelevant. In fact, in my opinion, it is like two old car buffs arguing about the merits of a Duesenberg and a Hispano-Suiza. Owners of these autos know that if you have to ask how much the vehicles “costs”, the person asking the question cannot afford the vehicles and probably will be unhappy shoveling money into the whirlpool that sucks cash to keep the Duesie and the Suiza humming. Traditional databases are equally voracious money pits.

The Edison Group discloses in its 78 page study, the following items. I have included the page number on which I located each of these points. I have selected only four items because I don’t want to spoil your fun when you read the original.

First, the Edison Group reported:

Benefiting from increased DBA productivity due to lower complexity and higher efficiency cited above, businesses could save up to $35,155 per year per DBA by using Oracle Database 11g rather than IBM DB2 Enterprise 9.5.

This statement from page 5 suggested to me that the study was funded by Oracle, and test data would demonstrate that Oracle is a money saver compared to IBM’s DB2. The precision of the number $35,155 is one of those numeric oddities I enjoy. The savings do not amount to $35,156, nor do I know what the confidence level the numerical method delivered; for example, a confidence level of plus or minus 50 percent gives me one sense of the savings. My hunch is that the difference in costs is probably like the Duesenberg – Hispano-Suiza analysis. For a big company, the cost differences may not be material because of the indirect costs these long-in-the-tooth data management systems impose.

Second, the weighting for the study makes it clear that indirects are not considered nor are capital costs. The table revealing this rather narrow focus on one specific set of administrative costs caused me to chuckle. You can find the table on page 15 of the report. One quick example: day to day admin has a weighting of 34 percent. Now the admin load depends on a number of factors. In these days of petascale data flows, the notion of a 34 percent weighting strikes me as low. Here’s why: the inability of these dinosaur-like database systems requires lots of fiddling around to build datacubes, to update indexes, and to deal with the hands-on fiddling dinosaurs require.

On page 31, the Edison Group reports that the two systems have improved. Here’s the passage I marked:

The differences between the two platforms have slightly increased over the years that Edison has been performing these studies. While IBM has done a good job in addressing many of the criticisms Edison has leveled in the past, most significantly in the areas of general system maintenance, Oracle has also significantly improved its offerings in these areas.

Wow. The idea that these systems have evolved and that Oracle is better than DB2 surprised me. Here’s why:

1. Oracle and DB2 are both old-style data management systems
2. The two companies have similar approaches to selling, cultivating partners, and solving performance problems. In fact, the easiest way to make either Oracle or DB2 perform better is to throw hardware at the problem.
3. Speed up methods so both systems can handle near real time index updates are not included in either system. In fact, an IBM invention that was used with the permission of IBM in the original Speed of Mind speed up “snap in” worked equally well on DB2, Informix, Oracle, and SQL Server. The point is that these old style data management systems suffer similar performance problems related to the engineering in the basement of the data management systems.

In short, Oracle and DB2, like the Duesenberg and the Hispano-Suiza are relics, expensive to own and maintain, and likely to become collector items.

Studies like this one from the Edison Group are useful for organizations who understand that data management has to be a name brand solution from a name brand company. The consultants may not realize that the foundation on which these Codd systems stand is being eroded. The notion that a company like Aster Data or InfoBright are precursors of even more significant disruptions from other firms is foreign.

The data management disruption will have a significant impact on today’s dominant data management companies. In my opinion, when petascale data flows impose sufficiently high costs, customers of Oracle and IBM RDBMS and data management systems will look for a lower cost, speedier, less expensive, more stable, easier to use option. That option will not come from today’s data management leaders. What data management systems does Google use? What happens if these are made available as a key component of Google’s enterprise services? Interesting questions and questions not addressed in the quite interesting analysis of the data management world’s Duesenberg’s and Hispano-Suiza’s of Codd technology. Just my opinion.

Stephen Arnold, September 19, 2009

Comments

Copyright © 2008 Beyond Search · ArnoldIT.com · Disclaimer · Revolution News theme by Brian Gardner · Log in