TODO

Double host increase rather than only by 50%
Deal with long run times (AWS session time out)
Calculate host increase based on memory needs rather than using naive approach
Add max / min service / host constraints