Massive working node-oriented preemptive task scheduling method and system

Abstract

The invention discloses a massive working node-oriented preemptive task scheduling method and system. The method comprises the steps of: obtaining a to-be-assigned task from a current to-be-assigned task queue with the highest priority; judging whether an idle working node exists in a current idle working node pool or not; if yes, assigning the to-be-assigned task to the working node selected fromthe idle working node pool; if no, selecting a target busy working node from a busy working node pool, wherein the priority of at least one task in tasks performed by the target busy working node islower than the priority of the to-be-assigned task; recycling a task, which has the priority lower than the priority of the to-be-assigned task, of the busy working node and assigning the to-be-assigned task to the busy working node; and increasing the priority of the recycled task by one level and placing in a to-be-assigned task queue with the corresponding priority. Through the method disclosedby the invention, the scheduling of the mass working nodes and the task priority is realized; the scheduling complexity is reduced; and the work efficiency of the scheduling is improved.

Description

technical field [0001] The invention relates to the field of task scheduling, in particular to a preemptive task scheduling method and system for massive working nodes. Background technique [0002] With the continuous development of the sharing economy and cloud computing, the future task scheduling is no longer task scheduling for limited working nodes in the enterprise, but task scheduling for platform Shanghai (millions and tens of millions) working nodes. Task requirements have also changed from planable or coordinated scheduling tasks to rigid priority requirements driven by bidding. [0003] Existing task scheduling methods mainly include the following two aspects: [0004] Scheduling a large number of working nodes: Currently, the LFU algorithm is generally used for load balancing and preemptive scheduling. [0005] The implementation scheme based on task priority scheduling: generally adopt the completely fair strategy (CFS) or use a priority sorting structure, su...

AI Technical Summary

Problems solved by technology

[0009] 3. Limited working nodes

Although the complexity of this scheduling algorithm has reached the O(Nlog(N)) level, it is still difficult to implement in practical applications with existing computing resources

[0016] At present, no algorithm complexity is low, and a public scheduling scheme that combines fair scheduling of massive working nodes and task priority preemptive scheduling has not been found.

Images