Stop hacking your production server (that is likely to introduce functional bugs) and take a step back. Can you reproduce the performance problems in a non-production environment? If not, then do the work to try.
You should try to reproduce the problem as follows:
- Get production-grade hardware in your test environment - web and database servers etc - running the same hardware as production
- Run the same software set as production - this includes the same configuration of ASPNET and all other servicces used.
- Load production-size data (production data if possible) into your databases (remember to firewall your lab from the internet so that it cannot send mail or other things to the internet, or your users might start receiving email notifications from the test system which would be bad!)
- Create simulated traffic to the site to production level - this is potentially quite tricky but there are a lot of tools available
Now you've got a fighting chance to repro the problem in testing, you can try solutions.
Usually database-driven web sites are bottlenecked by the database, so I'd start there. The main tricks are
- Do fewer queries
- Optimise the queries you do do (fetch less data, use appropriate indexes etc)
- Change the database structure so that the queries you have to do are easier for it (clustered indexes etc, denormalise possibly)
But any change you make, try it on your test system, measure the results, and if it doesn't help, ROLL IT BACK.
In general configuration changes are likely to make only minor differences, but you can try those too.
If all this sounds like too much effort, try throwing hardware at the problem - developer time is much more expensive than hardware. In the time it takes you to do the above (could be months depending on the complexity of the app) you could have bought some meaty production boxes. But be sure that it's going to help.
Does your database fit in RAM? Could it possibly fit in ram? If the answers to those questions are no and yes respectively, buy more ram for the database. This is one of the cheapest ways of making your db go faster without code changes.