If you still have doubts about tabular model, there is nothing like try it when you are starting a new project. If the model that you’re going to implement has simple business logic and security requirements and has a relatively low data volume, the tabular model can be the ideal solution for you as it wins in performance, especially in detailed queries where multidimensional model lose many points.
However, the migration of an existing multidimensional model to a tabular one should be carefully considered: first, migration is not straightforward, you have to build the model from scratch; and second, because there are some advanced features that are not supported by Tabular models (at least for now).
In some cases, as happened to me in a recent project, it’s advantageous to migrate an existing project to a tabular model because of the way it works and also the ease of modeling. In this particular case, I needed to add to an existing multidimensional model a large number of Distinct Count measures, which would force me to add a new measure group for each new measure. Thus, we decided to migrate part of the model to tabular which makes the model to be simplified (since it is not necessary to create and maintain the multiple measure groups that would be created in multidimensional), and also had gains in performance because of the way it reads the data from memory as well as it's faster with Distinct Counts calculations due to the column-oriented storage.
In conclusion and especially for large models, we still need the Multidimensional model where the Tabular one (Cached mode) is strongly limited by server memory. However, in small/medium solutions in which query performance is a critical requirement, the Tabular model should be our/your first choice.
