Switching time delay optimization method based on task fusion and task scheduling

Abstract

The invention provides a switching time delay optimization method based on task fusion and task scheduling, and the method comprises the steps: firstly analyzing the call relation of an application main function, constructing an initial task graph, and measuring the calculation time of each task in the initial task graph and the completion time of the whole application; obtaining an executable sequence of the tasks by utilizing topological sorting, scheduling the tasks according to the executable sequence, and selecting equipment with the shortest completion time for the tasks; calculating task fusion efficiency to determine whether the current task and the precursor task are merged or not, and recording a fusion strategy; obtaining a reconstruction task graph according to the fusion strategy, and calculating the completion time after task scheduling fusion; comparing the application completion time before and after task scheduling fusion, determining whether to reserve fusion operation, and determining a final task graph; and deploying the tasks to corresponding equipment for execution according to the final task graph. The advantages of a task scheduling method and a task fusionmethod are combined, and the task context switching delay is effectively reduced.

Description

technical field [0001] The invention relates to the field of multi-task computing, and more particularly, to a method for optimizing handover delay based on task fusion and task scheduling. Background technique [0002] In mobile edge computing scenarios, the execution of deep learning applications usually requires multiple execution devices to execute collaboratively. Therefore, a multi-tasking and multi-device system is usually deployed on edge servers and cloud servers. When end users need to perform tasks with a large amount of computation, they can choose to migrate them to this task system. As a typical computing-intensive application, a deep learning application will choose to migrate tasks to a multi-task system composed of edge servers and cloud servers during the execution of such applications. Deep learning can divide the entire computing task into fine-grained and smaller sub-task execution by converting it into a computational graph (tensorflow). However, dat...

AI Technical Summary

Problems solved by technology

[0003] Chinese patent CN107015856A published on August 4, 2017 discloses a method and device for generating a task scheduling scheme in a scientific workflow in a cloud environment. The method includes the following steps, step 1, obtaining all tasks to be scheduled; step 2 , describe the tasks to be scheduled through the directed acyclic graph with weights; step 3, in the directed acyclic graph, traverse from the start node to the end node to obtain all paths; step 4, calculate the Time consumption, sort all paths in descending order according to the time consumption from large to small, and determine the priority of the path; step 5, generate a task scheduling plan according to the priority order of the path; the invention uses the method of task assignment to use the processing The idle time of the computer is improved, the utilization rate of the processor is improved, and the total time for task completion is shortened, and no solution is proposed for the task switching delay

Images