If you need to do something like this, then you probably should take a look at the PowerTrack products from Twitter’s Gnip data service. They also allow you to dynamically inject new rules for filtering which do not require you to restart your process or reconnect. The public streaming API is not built for this kind of level of data volume.