What are the best SQL Server performance optimization techniques? [closed]

Answer 1

My advice is that "premature optimization is the root of all evil" in this context is absoulte nonsense.

In my view its all about design - you need to think about concurrency, hotspots, indexing, scaling and usage patterns when you are designing your data schema.

If you don't know what indexes you need and how they need to be configured right off the bat without doing profiling you have already failed.

There are millions of ways to optimize query execution that are all well and good but at the end of the day the data lands where you tell it to.

Answer 2

My advice would be to start with techniques applicable to all databases and then try the ones specific to MsSQL.

Optimizing SQL is difficult, and there are no hard and fast rules. There are very few generic guidelines that you can follow, such as:

• 95% of performance improvements will come from the application, not from server or database engine configuration.
• Design for correctness first, tweak for performance later
• Reduce trips to the database
• Try to express things in a way that fits your data model
• Ignore generic advice about performance - yes, at some point you'll find a system or SQL statement where one of those rules does not apply.

But the key point is that you should always apply the 80-20 rule. Which means that in any system you need to tweak 20% (often much less) of your code for the biggest performance gains. That's where the vendor provided tools usually fail, as they cannot usually guess the application/business context of execution.

Answer 3

profiling is key, but when using a profiling set you MUST be sure that it is an accurate test set of data, otherwise the tuning tools will not be able to get you an accurate result what is needed.

Also the management objects with fragmentation an usage reporting in 2005 are very helpful!

Answer 4

You might want to check internal and external framentation of current indexes and either drop and re-create them or re organize them.

Answer 5

profile your queries, not the obvious ones, but the complex that access different tables, views, etc and/or the ones that return many rows from different tables

That will tell you exactly where you should focus

Answer 6

It seems that you're talking about MS SQL.

Start the profiler and record tehe most common queries you run on the database. Then run those queries with the Execution Plan turned on and you will see what (if anything) is slowing your queries down. You could then go and optimize the queries or add more indexes on your fields.

SQL Books will give you a good overview of both profiling and query analysis functionality.