The Genetic Algorithm For Task Scheduling In Distributed Statistics Systems

The statistical analysis system for all kinds of basic data provides great help for managers to make decisions,exploit user preferences and data consolidation.With the increase of the basic data and the complication of statistical logic,the statistical system gradually changes to the distributed.In a distributed system,how to combine the task characteristics and the characteristics of statistical system to carry on the task scheduling,become the key to solve task scheduling distributed statistical system.The task scheduling for distributed system mainly including business systems of FIFO scheduling,fair scheduler and capacity scheduler and scientific research of Min-min,PSO,ant colony algorithm,fuzzy logic,etc.These methods are relatively simple,unable to adapt to the complex scheduling environment,and cannot combine the characteristics of the system and task,so the result is low efficient.Combining with the master's stage to participate in the project,in view of the distributed task characteristics of patent statistical analysis management system and resource requirements,investigates the characteristics of patent import calibration tasks and performance of computing nodes.formalize task scheduling and build statistics system and transformed into a multi-objective optimization problem,the design of genetic algorithm is suitable for the patent data import task scheduling calibration system.Combining with the characteristics of system and user needs,the goal for scheduling is to quickly respond to user requests and ensure load balancing system.To ensure the effectiveness of the results,improves hybrid operator to find better solution.The details are as follows:First,a detailed analysis of distributed patent business statistical analysis management system of business processes,tasks characteristics and resource requirements is made.Make a formalization description of the goal of task scheduling,process of task scheduling,task characteristics and compute node.This work is the base of algorithm each relevant parameter design.Then,the genetic algorithm applied to the task scheduling of patent business statistical analysis management system is proposed to solve the long system response time and unbalanced problem.Resources-task population coding strategy,the population initialization strategy combined with the chromosome connection system task completion time,the parameters of the load balance to build controllable adaptation function a d retain the best individual as the optimal solution in the process of evolution is introduced in detail,in the improve part of algorithm,with threshold adaptive hybrid probability function is proposed,which in prophase provides assurance to promote population evolution,the late decrease the evolution probability so as to keep good results.Finally,the experimental platform is set up,simulated tasks and compute nodes is made,and the algorithm is applied to the statistical.system task scheduling.Experiment validates the effectiveness of the algorithm,at the same time,comparative experiments prove that algorithm can find a better solution,and cross probability and mutation probability parameters and the biggest mutation step length parameter are optimized.