The other day I tweeted that MySQL unions are my new best friend. I just want to take that back. I just spent the past six hours (it's now 5 am) overhauling a MySQL query and removing one of the unions actually sped up performance. Dazah is really my first foray into the world of big data. I've found the hardest problem so far has been writing queries that will work regardless of the subset of the data size they are working against. For example, there are multiple bubbles and audience segments on Dazah, of all sizes and each favoring a different behavior type (some bubbles are more chatty than others), and I've found that every time I optimize a query for one dataset it fubars with a different one. Just a rant.