Answers are grounded in this paper and cite their sources.
Current shared cloud operating systems (cloud OS) such as Mesos and YARN are based on the "de facto" two-horizontal-layer cloud platform architecture, which decouples cloud application frameworks (e.g., Apache Spark) from the underlying resource management infrastructure. Each layer normally has its own task or resource allocation scheduler, based on either time-sharing or space-sharing. As such, the schedulers in different layers are unavoidably disconnected, --- not aware of each other, which highly likely leads to resource (e.g.,CPU) wastes. Moreover, the tail latency may even be harmed due to the performance interference on shared resources.
No takes yet. Share an insight, caveat, or question.
Wang et al. (2019) studied this question.