There are a few options you could consider here. The first would be to pre-compute all values in the dataset and load the entire set of data int Tableau Server. This would eliminate the need to run Python dynamically and thus to have a separate TabPy Server. This is also something you could do by integrating TabPy into Tableau Prep running on a local machine. It's worth noting that TabPy is a very light application and so requires fairly small hardware allocations to run effectively.