Evaluation and Comparison of Yarn Schedulers in the Science and Technology Services Resource Pool System.

In order to develop the science and technology service industry in Central Plains Urban Agglomeration, as well as to accelerate inter-regional coordination and cooperation, a regional resource pool was built. Science and technology service resources here are numerous, complex and diverse, and contain large amounts of data. Hadoop-2.9.2 could organize and analyze these big data, and achieve high-speed computing and storage. Considering that the concurrent demand of some scarce resources is relatively large, the system needs scheduling capabilities. Yarn is used to provide resource scheduling, multi-tenancy, and linear-scale storage. In experiments, we evaluated all of the provided schedulers, FIFO Scheduler, Capacity Scheduler and Fair Scheduler, and chose one of them as the actual scheduler.