How would you design a system to process a file that is significantly larger than the available RAM in non-FAANG US (org related to data)?