I'm migrating a Flink workload from EMR to K8s and wondering if K8s has a resource allocation concept similar to Dominant Resource Fairness (DRF) that is available in other resource management solutions like Yarn and Mesos? We explicitly specify the requests and limits in K8s manifests but that is limiting the performance of TaskManagers with heavy operators