A data-driven and scalable intelligent general task scheduling system

Abstract

The invention discloses a data-driven and scalable intelligent general task scheduling system, including an application information collection system, an information transmission system, an information storage system, a data intelligent processing system, and a scheduling system; wherein the data intelligent processing system is mainly composed of data loading, generating Constructing parameters of neural network, constructing neural network, generating neural network and decision information. The present invention introduces neural network learning into the task scheduling algorithm design of the operating system, can automatically train and learn network parameters from the accumulated task running status, resource requirements and other data, and generate efficient scheduling decision information represented by neural network The decision function realizes the intelligence and automation of task scheduling algorithm design, and solves the problems of poor versatility, repeated development, and manual design rules in the development of task-aware task scheduling algorithms in different scenarios.

Description

technical field [0001] The invention relates to a data-driven and scalable intelligent general task scheduling system (note: "task" is also called "process", "job" and "thread"), which belongs to the technical field of operating system software. Background technique [0002] Desktop operating systems, server operating systems, and smart terminal operating systems (typically such as Android smartphone operating systems) all belong to stand-alone systems (or single-node systems). In stand-alone systems, task scheduling is usually also called process scheduling or thread scheduling. The general concept of task scheduling in a stand-alone system is the same. They all allocate CPU time, GPU, memory, SSD, disk, network card, LCD and other equipment resources for tasks reasonably, in order to meet the basic requirements of each task according to a certain goal. When needed, system resources can be used efficiently. [0003] In recent years, related research at home and abroad, aim...

AI Technical Summary

Problems solved by technology

[0003] In recent years, related research at home and abroad, aiming at the differences in scheduling objectives in different application scenarios, has derived a large number of different scheduling objectives such as energy consumption perception, task correlation perception, trustworthiness perception, resource perception, temperature perception, and process perception on the basis of general concepts. Scheduling algorithm design, collectively referred to as task-aware scheduling, the shortcomings and deficiencies of these task-aware scheduling algorithm design can be summed up as "lack of versatility", "repeated development", "manual design rules", specifically in: for different needs , the design of scheduling algorithms for different goals has emerged. Each algorithm needs to manually write new scheduling rules, manually analyze the scheduling scenario from scratch, manually design the scheduling rules for specific scenarios, and rewrite and compile the specific scheduling algorithm.

Images