Also Nima ... is the 2.0 your production site? If so, is it a BUSY production site?
If so the entire discrepency might be due to cache contention, and the busy site will win that war over your inactive test site all the time. The notion is that your inactive site will be forced to go to disk to get its data nearly all the time because the busy site will be hogging both the DB shared buffer and the Linux disk block cache.
In fact I diagnosed a similar situation for furfly about six weeks ago ...
If this isn't the production site then there's no point in my continuing, so I'll stop for now until I read your response tomorrow AM. If it is, though, I can suggest an approach for making a diagnosis.
