Cloud　computing　attracts　a　growing　number　of　organizations　to　deploy　their　applications　in　distributed　data　centers　for　low　latency　and　cost-effectiveness.　The　growth　of　arriving　instructions　makes　it　challenging　to　minimize　their　energy　cost　and　improve　Quality　of　Service　(QoS)　of　applications　by　optimizing　resource　provisioning　and　instruction　scheduling.　This　work　formulates　a　bi-objective　constrained　optimization　problem,　and　solves　it　with　a　Simulated-annealing-based　Adaptive　Differential　Evolution　(SADE)　algorithm　to　jointly　minimize　both　energy　cost　and　instruction　response　time.　The　minimal　Manhattan　distance　method　is　adopted　to　obtain　a　knee　for　good　tradeoff　between　energy　cost　minimization　and　QoS　maximization.　Real-life　data-based　experiments　demonstrate　SADE　achieves　lower　instruction　response　time,　and　smaller　energy　cost　than　several　state-of-the-art　peers.　©　2020　IEEE.