Limit or constraints of data

When I use XGBoost to train data, is there any limit of data? Specially I am confused about two issues.

  1. What are the recommended ranges of data? C++ int range aka -2147483647 ~ 2147483648?
  2. When different features are in different orders of magnitude. Does XGBoost do normalization internally?

I am not sure what the answer to (1) is. For (2), I am by no means an expert, but I believe normalization is not particularly necessary for decision tree methods, such as xgboost. There is good explanation given here: