A resource scheduling method based on FSM control and a frame system thereof

Abstract

The invention provides a resource scheduling method and a frame system based on FSM control, wherein scheduling policy scheduling events are respectively received from the outside in real time, basedon the preset scheduling event of each scheduling policy and the corresponding relationship between each FSM state of the finite state automata, the present FSM state is correspondingly migrated to each corresponding FSM state corresponding to an external scheduling policy scheduling event received by the finite state automata, and based on the correspondence between the output results of each FSMstate of the pre-established finite state automata and each scheduling policy preset in the multi-mechanism scheduler, the multi-mechanism scheduler determines the corresponding target scheduling strategy corresponding to the corresponding FSM state that the finite state automata currently migrates to, and schedules the resources in the resource pool respectively according to the corresponding determined target scheduling strategy. The invention can meet the requirements of various scheduling scenarios of an enterprise in real time, and increases the flexibility of resource scheduling.

Description

technical field [0001] The invention relates to the field of resource scheduling, in particular to a resource scheduling method and framework system based on FSM control, which are mainly used for resource scheduling of big data computing platforms. Background technique [0002] In today's era of explosive big data technology, the management of massive storage and computing resources has become especially important. In order to realize the effective utilization of resources, or to realize the low-latency execution of time-sensitive calculations, a variety of resource scheduling modes have appeared in the prior art, such as the commonly used YARN resource scheduling framework and mesos resource scheduling in the prior art Framework, as well as the kubernetes resource scheduling framework, these scheduling modes provide relatively optimized solutions in most scenarios, but these scheduling modes either have the problem of a single scheduler scheduling mode, or have a small spa...

AI Technical Summary

Problems solved by technology

[0009] ① The preemption mechanism is good for time-sensitive applications, but this mechanism mode is global and can only be started when the system starts, and cannot be switched dynamically;

[0010] ② When there are multiple departments in the enterprise who use resources mixedly, different departments often need different scheduling strategies, and the granularity of YARN scheduling is often very fine, and the flexibility of use is relatively weak;

[0011] ③ Although the plug-in scheduler of the YARN resource scheduling framework increases the flexibility of the YARN framework to a certain extent, the scheduler also needs to be enabled at system startup and cannot be changed dynamically

It can be seen that both the Mesos resource scheduling framework and the Kubernetes resource scheduling framework provide fewer scheduling mechanisms, and basically require a lot of effort for complex scheduling management

Images