In reading “Understanding Big Data eBook”, I came across a report that Reliance cuts database total cost of ownership in half with IBM DB2 . Reliance Life Insurance in India got the following benefits from migrating from Oracle to DB2:
Increases application uptime from 80 to 95 percent; decreases total cost of ownership by 50 percent; supports four times the number of concurrent users.
I am amazed by the increase in concurrency. My experience would indicate that the opposite is true: moving from DB2 to Oracle greatly increases concurrency.
One of the reasons I changed from being a DB2 DBA to an Oracle DBA is that my life is much more easier under Oracle. I have a richer set of commands to administer the database, and I do not have to contend with locking as much as I did with DB2.
Given my experience, I was intrigued by the report.
This was not a straight migration as:
IBM provided best practices to help the Reliance IT staff redesign its data model for improved performance.
So we have a major problem in evaluating the benefits of the migration. How much of the performance improvement is due to this redesign? What would have been the performance improvement on Oracle with this new data model?
It seems to me that the major cause of the downtime with the Oracle system was the time needed to load data into the data warehouse. With the faster data capture times by DB2, the company was able to increase its availability from 80% to 95%. Again, we have the problem of the improved data model obscuring the real differences between Oracle and DB2 implementations.
And the increased performance allowed for a reduction in CPU and licence costs. This is where the company achieved its operational savings:
This CPU savings along with reduced hardware and software licensing fees has resulted in a 50 percent decrease in total cost of ownership. “The total cost of ownership with DB2 running on IBM systems is almost half the cost of Oracle Database on Sun systems,” says Ms. Daga.
As I said before, this all could be due to the improved data model.