Similar articles

Abstract: Nowadays, the management of resource contention in shared cloud remains a pending problem. And the evolution and deployment of new application paradigms (e.g., deep learning training and microservices) and custom hardware (e.g., GPU, TPU) have posed new challenges in resource management system design. Current solutions tend to trade cluster efficiency for guaranteed application performance, e.g., resource over-allocation, leaving lots of resources underutilized. To overcome this dilemma is not easy, because different components across the software stack are involved. Nevertheless, massive efforts have been devoted to seeking effective performance isolation and highly-efficient resource scheduling. The goal of this paper is to systematically cover related aspects to deliver the techniques from coordination perspective, and identify the corresponding trends they indicate. Briefly, four topics are involved. Firstly, isolation mechanisms deployed at different levels (micro-architecture, system and virtualization level) are reviewed, including GPU multitasking methods. Second, resource scheduling techniques within individual machine and at cluster level are investigated, respectively. Particularly, GPU scheduling for deep learning applications is described in detail. Third, adaptive resource management including the latest microservice-related researches is thoroughly explored. Finally, future research directions are discussed in the light of advanced work. Hopefully, this paper will help researchers establish a global view of the landscape of resource management techniques in shared cloud, and see technology trends more clearly.