Welcome to the Q&A Forum

With that kind of complexity, there's rarely just one answer, and I'd hesitate to use rel=canonical for every problem. For search pagination, for example, rel=prev/next is probably a better choice. I've got a huge post on the topic here that shows just how complex it is: http://www.seomoz.org/blog/duplicate-content-in-a-post-panda-world A combination of META NOINDEX and rel=canonical can be a big help for the rest, but it depends a lot on the situation. Do you have a sense of what the worst culprits are? If you can start with the ones causing 50K+ indexed pages each, you don't have to necessarily solve everything at once. Don't implement multiple solutions for any given problem - it can do more harm than good. Also, be aware that deindexation takes a while - sometimes a long while. I know that's frustrating, but just because 500K pages don't drop out of the index tomorrow doesn't mean what you did isn't working. It can take 1-2 weeks just to see them start to fall. Don't make a change and then reverse it in 48 hours, or you can cause even more long-term damage (believe me, I've seen it).