What should be the cluster size for completing VFVS using the REAL lib against one receptor in two weeks?

First, thanks for such a well documented open platform. I would like to run VFVS against a receptor of my interest, using the REAL library. I am thinking of utilizing our institutional research computing cluster. In that case, I will have to request for time allocations in core.hr. The lowest allocation is 50,000 core.h over 3 months OR 1 node(24 cores) for 3 months. The highest allocation is 20 nodes/20*24 cores for 3 months. This seems woefully small to complete the REAL database in reasonable time (in the Nature paper, it is reported to take 5.5 M core.hr (8000 cores for 4wks) ). Am I missing something in my calculations?