Resource scheduling for quality-critical applications on cloud infrastructure

Cloud computing can provide virtualized, elastic, and on-demand computer system resources for supporting complex distributed applications, which has become the ubiquitous and primary computing paradigm for today's business. Meanwhile, the performance of cloud applications is getting more and more attention, particularly along with quality-critical applications which have critical requirements for Quality of Service (QoS) or Quality of Experience (QoE). In this thesis, we investigate how to efficiently schedule resources to satisfy quality-critical requirements of diverse applications on cloud infrastructure. First, we propose a Deadline-aware Deployment System (DDS) for time critical applications in clouds. DDS enables users to automatically deploy time critical applications and provides scheduling mechanisms to guarantee deployment deadlines. Second, we propose an Enhanced Container Scheduler (ECSched) for efficiently scheduling concurrent container requests with multi-resource constraints on heterogeneous clusters. ECSched formulates the container scheduling problem as a minimum-cost flow problem (MCFP) and performs a well-designed algorithm to schedule the concurrent requests in an online manner. Third, we propose new algorithms to optimize the placement of service-based applications in clouds. The objective of the proposed algorithms is to minimize inter-machine traffic while satisfying the multi-resource demands of service-based applications. Finally, we present GoTask, an approach that can learn to well schedule DAG jobs with multi-resource constraints on heterogeneous clusters. GoTask directly learns DAG scheduling policies from experience through deep reinforcement learning.