The thing i'e found the slowest about o r ms and odiumo, if you're doing document data basis, is the deceralization. In those cases it makes sense to maybe do some kind of projection or something. I don't know how to do i in jango orim, but its a mongo engine. You can say what i man to get back is a initerable set of these objects that ach to the data,. But only actually add, just fill up these two fields. Most of the data just thrown away. And that usually makes it dramatically faster. Yes, we run into a number of bottle necks at the cerialization layer,