Improving Resource Efficiency for Colocation of Multiple Latency-Critical Applications in Datacenters