With the widespread application and attention in the field of cloud computing, the cluster container orchestration management platform Kubernetes has been widely used for automated deployment and release of containerized application services, application elastic scaling and rollback updates, fault detection, and self-repairing services. The 5th generation Reduced Instruction Set Computer (RISC-V) has four major technical characteristics and advantages: simplification, modularity, scalability, and open-source, which have attracted widespread attention in academia and industry. Based on the collaborative research points of the Kubernetes ecosystem and the RISC-V ecosystem, this paper provides cloud service task scheduling support for heterogeneous instruction set architectures (ISA) to the Kubernetes scheduler. Through quantitative analysis of various computing task requirements of the RISC-V instruction set architecture in production environments, it was found that the existing cluster container orchestration platform Kubernetes lacks the capability to schedule computing tasks of the RISC-V instruction set architecture, especially that its scheduling algorithm cannot utilize the characteristics of RISC-V user-defined extensible instruction set architecture to provide high-performance reliable services. To solve the above problems, an ISAMatch model for scheduling at creation time is proposed, which comprehensively considers multiple aspects such as instruction set affinity, the number of nodes with the same instruction set architecture, and node resource utilization to achieve optimal task allocation. Based on the existing cluster scheduler, it improves its scheduling needs for tasks of various instruction set architectures. Compared to the default scheduler’s accuracy rate of 62% (scheduling RISC-V basic instruction set tasks), 41% (scheduling RISC-V extended instruction set tasks), and 67% (scheduling RISC-V extended instruction set tasks with “RISC-V” node matching labels), under the condition of not considering resource constraints, the ISAMatch model can achieve a 100% task scheduling accuracy rate.
Scan the QR code to read the full article
