• cost, latency and security of inter-service(cluster) communication • cluster per environment? • Are only “prod” and “dev” really enough? • some services may require “qa”, “sandbox” and PR-based “staging” • Is prod/dev parity really guaranteed? • resource efficiency