10.08.2008

Scientific methods vs Bullet Methods

Of late, there has been a lot of debate, fight, discussion (whatever you call it) in the Oracle community (I would rather say between Oracle experts) on 'which is the best method to tune a database?', a scientific method or Silver Bullets? Many DBAs across the global have expressed their opinions and views, so, I am here to express and share few incidents that I came across. If you decide that I am going to pin point someone or prove which is wrong and which is right, you would be disapointed to know that I am not, simply, becasue  you must be knowing which is the correct method.

I remembered sometime back  a DBA has told me that one of his DBA colleague has rebuild all the indexes in a production database and mailed  to their management and immediate boss stating that this is going (rebuilding indexes) to improve the database performance and also attached an URL which stated that rebuilding indexes will improve the database performance.  I had told my friend to get me the answer for the following questions from his DBA colleague:

Does he know what is the harm or side effects of doing it?
Were there any performance issues in the database, if so, are they related to indexes?
What amount of performance improvements he measured or gained after?

I am yet to get answers for the above questions. And moving on to another incident...

I know a company very personally who follws(still following) so called 'database re-org' method to re-organize their database weekly. When I questioned them why they are doing it so, their answer is that they have gained an immediate performance improvements when they did it first time and they are following this procedure in order to not to drop the performance level. It was really difficult to convience them how dangerious it would to follow this method.

Myself would definitely will not go for 'leaven an arrow in the dark' policy/method. Even if I solve something accidently, I love to spend some time to do R&D to come to a conclusion.

Last but not least, when I visit a doctor, I do expect him to do analysis of my problem and try to cure it permanently rather than giving a pain killer to relief me temporarily from the pain.

I, therefore, invite you all to discuss your wired experiences or incidents that you come across.

Happy reading,

Jaffar









3 comments:

Anonymous said...

Hi Jaffar,
First of all congratulations on getting OCM. My experiences are really wierd. Although not working on Oracle (I miss it really) for the past 9 months. Here in my company we have just upgraded to sql 2005 and following are the best (worst) practices which we follow:
- Don't use transactions!
- Use a caching API created by the architect, which caches all of the data as objects in java.
- Move all the logic out from the database into middle layer.
- Whenever any query performs slow turn that code into caching code instead of tuning the query or rewriting it
- Databases are good for processing large sets of data (millions of records) so use them only for bulk processing and for oltp systems use java cache, use database as just a datastore.
- There is a plan now to create an asynchronous processing of transactions using java cache.

If these guys had read the basics of database...............

Regards,
Ahmed

Anonymous said...

Dear Jaffar,

Hearty congratulations on getting OCM......Can you suggest us on how to go about OCM preparation after finishing our OCP......Can you give details on what all books you read to become an OCM......Apart from Oracle online docs what else did you read......If you feel free to discuss on OCM exam experience at test centre it will be of help to us...

Anonymous said...

Hi Jaffar
I really liked your blog and your dedication to oracle. Well i will be in touch with you. Ok bye