The other answers posted so far focus on unit/functional/performance/etc. testing, and they are all reasonable.
However, one the key questions you should ask is, "how effective is my testing?".
This is often answered with test coverage tools, that determine which parts of your application actually get exercised by some set of tests. The ideal test coverage tool lets you test your application by any method you can imagine (including all the standard answers above) and will then report what part and what percentage of your code was exercised. Most importantly, it will tell you what code you did not exercise. You can then inspect that code and decide if more testing is warranted, or if you don't care. If the untested code has to do with "disk full error handling" and you belive that 1TB disks are common, you might decide to ignore that. If the untested code is the input validation logic leading to SQL queries, you might decide that you must test that logic to ensure that no SQL injection attacks can occur.
What test coverage tools let you do it to make a rational decision that you have tested adequately, using data about what parts of your code has been exercised. So regardless of how you test, best practices indicates you should also do test coverage analysis.
Test coverage tools can be obtained from a variety of sources. SD provides a family of test coverage tools that handle C, C++, Java, C#, PHP and COBOL, all of which are used to support web site testing in various ways.