Archive for July 2011
Previously, in the wake of the Oracle Itanium announcement, I blogged about how IBM DB2 Welcomes Oracle Database/HP Itanium Customers. I also relayed the results of an early survey of the HP-UX community about their view on their options in HP-UX Administrators Consider their Oracle Database Options.
Since then I have had a little time to reflect, and have put together a Webcast titled Options for your Oracle Database Workloads on HP Itanium Servers. If you have Oracle Database on HP-UX, make sure to check out the webcast, and please add your feedback in comments so others can benefit from your experiences/opinions.
If you are an Oracle Database DBA who is going to Oracle OpenWorld in San Francisco this year, you should check out a free training course from IBM. Immediately before the conference, we are offering a free hands-on course that is designed to help people with Oracle Database skills learn DB2. Of course, because DB2 now supports so many Oracle Database-like features, it is much easier for Oracle professionals to work with DB2. Not only does the course discuss the differences between the products, but it also gives you hands-on experience with DB2. This is a great opportunity. It is an easy and affordable way to expand your skills/resume to include DB2. You’re going to San Francisco anyway, so all you have to do is get your employer to spring for a couple of additional hotel nights. There’s even an opportunity to take a free official certification test as part of the course. The course is two days long, and takes place on Friday 30 Sept and Sat 01 Oct in San Francisco. To learn more and register, please go to www.ibm.com/db2workshop.
Anyone who has spent any length of time around IT will have encountered their fair share of shameful vendor-sponsored studies. It is an unfortunate reality of the over-aggressive marketing that sometimes comes to the fore. In my opinion, a recent study from ORC International that was commissioned by Oracle is one of the most worthless vendor-sponsored studies that I have ever seen.
Please understand that I, as someone working with DB2, have a natural bias against Oracle. Please also understand that the product I work on is presented in an unfavorable light in this study. That being said, I think I have reasonable grounds for my position. Please let me know if you agree or disagree. And, by the way, I am open to an equally critical examination of any vendor-sponsored studies that I commission.
The study in question is titled Database Manageability and Productivity Cost Comparison Study: Oracle Database 11g Release 2 vs. IBM DB2 Enterprise 9.7. It claims that database administrators can perform administrative tasks in 43% less time and with 45% fewer steps in Oracle Database 11g when compared with IBM DB2 9.7. This is actually the fifth such study commissioned by Oracle. The original study was written by Rauch Associates, then the Edison Group wrote the next three instances of the study, and now the latest installment comes from ORC International.
The first thing I’d like to point out is that the methodology for this report is overly simplistic, and could be prone to incorrect conclusions. It uses two metrics: the amount of time it takes a DBA to perform a task, and the number of steps involved in performing a task.
Of course, when you think about the time measurements, you’ve got to wonder about the relevant levels of experience of the DBAs used for the tests. This metric is, as you can imagine, subject to wild disparity. In order to put this metric into perspective, we need to know the experience levels of the DBAs involved. Without this information, the time measurements are incomplete. Even with this information, the time measurements are still subject to question, given the inexact nature of human experience levels and human ability.
And it goes without saying that a simple count of the number of steps needed to complete a task is an overly simplistic metric. For instance, when creating a tablespace in Oracle Database, there is a single screen with a number of settings for creating a tablespace. There is little or no on-screen guidance, and no context-sensitive help for the settings. Even though this is a usability nightmare, this would count as a single step. On the other hand, in DB2 you have the option of using a nice 7-screen wizard with lots of great context-sensitive help. Is this one step, or seven steps? You can quickly see that a simple count of the number of steps is simply not reliable.
To compound matters, the logic also appears to be flawed. The study goes to great lengths to construct the framework for the tasks that will be compared. They claim that “DBAs were interviewed about the amount of time they typically spent in each of the four areas of study to determine the weighted workload value for each subject area, and multiple database disciplines were chosen, including 24x7x365 production DBAs as well as application and development DBAs“. The upshot of all those interviews is that 45% of a DBAs time is spent with Performance and Tuning Tasks. I don’t think this is an unreasonable estimation of the amount of time, especially in Oracle Database environments. However, the interesting thing is that there are three tasks in the Performance and Tuning section, and two of those require 0 steps and 0 time for Oracle Database. This area where DBAs report that they are spending 45% of their time requires only 2 steps and 8 seconds across all three tasks in an Oracle Database environment. It doesn’t take a rocket scientist to figure out that something doesn’t add up here. Not only does this supposedly time-consuming set of tasks take almost no time and effort, but this relative disparity is then weighted heavier than everything else in the findings. As if the flawed methodology was not enough, this flawed logic makes the findings even more questionable.
As I mentioned earlier, this study is essentially an updated version of the study from Edison that compared Oracle 11g and DB2 9.5. Both studies essentially use the same methodology. The primary difference is that this latest study uses the current versions of the products. But if you compare the two studies, you will notice something very mysterious. In several cases, DB2 now has a higher complexity score, even though the same interfaces are available. For example, in the previous report from Edison, creating a table took 17 steps in Oracle Database and 5 steps in DB2. Now, mysteriously, it takes 17 steps in Oracle Database and 17 steps in DB2. How can Oracle and ORC International explain this fact? Why does something that had a complexity score of 5, now suddenly have a complexity score of 17?
Is this an Apples-to-Apples Comparison?
It is also interesting to note that this study takes advantage of Oracle’s add-on products that must be separately purchased (namely, the Diagnostic and Tuning Packs). However, they do not do the same for DB2 (in particular, they do not use IBM Optim Performance Manager or DB2 Recovery Expert). Instead, they use the deprecated DB2 Control Center. Why did they choose to perform tasks only with a subset of the tools available to DB2 DBAs, yet they include extra charge products for the Oracle tasks?
And a Suspicious Twist
And the final thing I’d like to draw your attention to is the “Out of the box installation” step. Notice that the DBA performing the tests encountered a bug that elevated the DB2 complexity score from 1 up to a whopping 15. The study states that they encountered JR28800. However, the IBM support site states that this bug was resolved in DB2 8.1 Fixpack 17. How is it possible that they ran into a problem in DB2 9.7 that was resolved in 8.1 Fixpack 17? Did Oracle and ORC International really use the current version of DB2, or did they use a version that is so old that it is no longer even in service?
There is a Better Way
The good news is that there is a better methodology for comparisons like this. It was developed by researchers looking for a general way to determine time and labor costs associated with IT systems. They presented their findings at the 10th IFIP/IEEE International Symposium on Integrated Network Management 2007 in a paper titled Predicting Labor Cost through IT Management Complexity Metrics. It would be very interesting if Oracle were to use these more accurate and objective complexity metrics, rather than their own suspiciously rudimentary measures. If they did, they may find that the results are closer to the finding in the study covered in the Comparing Common DBA Tasks for IBM DB2 and Oracle Database post, where DB2 is shown to be clearly superior.
Let’s Hear from Actual Users of both Products
By now, you’ve probably had your fill of hearing about vendor-sponsored studies. Let’s hear from people who switched from Oracle Database to DB2, and are willing to go on record with their experiences:
“The total cost of ownership with DB2 running on IBM systems is almost half the cost of Oracle Database on Sun systems. With DB2, our current DBAs will be able to support more applications and more users.”
- Anuprita Daga, Chief Manager, IT, Reliance Life Insurance
“Our IT team now has more time to work on other value-added tasks due to DB2 self-tuning properties and the seamless integration with the SAP software.”
- Tom DeJuneas, IT manager at Coca-Cola Bottling Co Consolidated
“DB2 requires less attention from the Basis DBAs than Oracle”
- Andrew Juarez, Lead SAP Basis Administrator at Coca-Cola Bottling Co Consolidated
“We switched from Oracle Database to IBM DB2 and cut our data management costs in half, while improving performance and reliability of business applications.”
- Sandro Reátegui, Systems Architect, Banco de Crédito del Peru
“Data administration costs have also decreased. DB2 automates many tasks, such as memory allocation, storage management and business policy maintenance, freeing the company’s database administrators to focus on new projects.”
- Sergey Golubev, Head of IT Maintenance and Networking Department, JSC Rietumu Banka