Efficient processing of large volume files

The size of the input text file is constantly increasing and can reach 100 GB or more, making it impossible to load it entirely into RAM. How to design processing of such files so that the service remains stable and correctly outputs the top N most frequent words?