Research On Multi-objective Discrete Optimization Oriented Swarm Intelligence Algorithms And Applications In Cloud Computing Scheduling Optimization

With the development and popularization of cloud computing,improving overall resource management and operational efficiency,and optimizing investment have become the key.In the cloud computing application environment,resource scheduling and task scheduling need to consider multiple heterogeneous resources and complex and changeable application requirements,while taking into account various performance requirements,including the overall energy consumption of the data center,resource utilization,economic benefits,and the quality of services,etc.These problems are usually interrelated,promote or inhibit each other,and cannot be solved by simple weight assignment.Therefore,the cloud computing scheduling problem has common characteristics of discrete optimization and multi-objective optimization,and it is very suitable to be solved by optimization algorithms.However,the heterogeneity of resources in the cloud computing environment,the diversity and dynamics of applications,and the requirements of multiple constraints and multiple objective functions put forward higher requirements on the optimization algorithm.It needs to ensure high reliability,stability,and scalability of the optimization algorithm.This thesis focuses on the research of new swarm intelligence algorithms and applies them to solve multi-objective discrete optimization problems in cloud computing environments.The main research work of this thesis includes:1)Research a new kind of invasive tumor growth optimization algorithm ITGO,optimize the design of the basic ITGO algorithm,and expand it to the discretization space so that it can be used to solve discrete problems.ITGO algorithm is a new type of swarm intelligence algorithm proposed by our laboratory based on the growth mechanism of tumor cells.It can solve single-objective problems through the generation and transformation of four types of cells: invasive cells,growing cells,dormant cells,and dead cells,in a nutrient environment.Aiming at improving search efficiency of ITGO algorithm,this thesis optimizes the transformation strategy of cells,gegeration strategy of cells,and step-size optimization strategy to make the algorithm have better performance.Compared with classic optimization algorithms such as the particle swarm optimization PSO,genetic algorithm GA,and differential evolution algorithm DE,the improved algorithm ITGO+ has better convergence.On the basis of ITGO+,this thesis proposes the discretized D-ITGO algorithm,which works on the discretized solution space by mapping the individuals to discrete solution space for processing.Compared with ITGO algorithm,D-ITGO algorithm not only has higher search efficiency and better search performance,but also can be applied to solve cloud computing task scheduling problems.Compared with other task scheduling algorithms,D-ITGO algorithm also take advantages.2)Designed and implemented a vascular invasive tumor growth optimization algorithm VITGO for multi-objective optimization problems.Aiming at the characteristics of general multi-objective optimization and on the basis of ITGO algorithm,this thesis reconstructs the growth model of the entire tumor cell population,utilizes the vascular units to guide the generation of tumor cells by drawing on the vascular growth mechanism,define the growth mechanism of vascular units according to the characteristics of multi-objective optimization problem,and redesign the search methods,transformation,and initialization strategies of various types of cells to match the growth of vascular units,to make the VITGO algorithm solve a series of multi-objective optimization problems.In order to improve the search efficiency of VITGO algorithm,this thesis proposes a more effective boundary determination and detection scheme,endpoint detection and utilization scheme for Pareto front,and the removal of highly similar Pareto solutions to avoid redundant calculations.In most benchmark functions,VITGO work better than current classic and latest multi-objective optimization algorithms in finding Pareto solutions.3)Designed and implemented a hybrid growworm swarm optimization algorithm HGSO to solve the task scheduling problem in cloud computing.Based on the shortcomings of the original glowworm swarm optimization: lower convergence and easy to fall into the local optima,this thesis proposes a hybrid glowworm swarm optimization HGSO,which combines the advantages of swarm intelligence(fast search speed and wide range of utilization)and the advantages of evolutionary algorithm(survival of the fittest/fast convergence),to work for task scheduling problem incloud computiong.The proposed HGSO algorithm includes three targeted improvement strategies: survival of the fittest strategy based on elite individuals,the quantum transition strategy based on neighbor-rule of GSO,and the full random walk strategy which can make the algorithm has a higher convergence speed and jump out of the local optima in time.In solving the cloud computing task scheduling problem,the provided HGSO algorithm has faster convergence,and the final cloud computing task scheduling strategy(optimum solution)has advantages ranging from 12%-35% on Makespan compared with other algorithms.4)Designed and implemented a virtual machine consolidation oriented multi-objective invasive tumor growth optimization algorithm VMITGO for virtual machine consolidation problem in cloud computing.This thesis constructs a basic computing framework MITGO for multi-objective optimization by drawing on the tumor cell's avascular growth model,designs the transformation between different types of tumor cells and an optimal-search rule,and applies it to solve the virtual machine consolidation problem.According to the multi-objective optimization requirements of the virtual machine consolidation problem,this thesis designs an three-objective optimization probem that takes into account energy consumption,virtual machine migration and load balancing to formulate the whole problem,and provides a semi-initialization scheme as well as two sbstitutaion schemes of virtual machine consolidation to reduce migration number and cost.These schemes can also reduce energy consumption of cloud data center and achieve load balancing.Experimental results on the Google Trace Data show that the optimal-search rule and follow-up generation scheme can search with good efficiency.In terms of energy consumption,the number of virtual machine migrations,and load balancing,the VMITGO algorithm performs well,and the overall performance is better than the compared algorithms.The main work of this paper is to study new swarm intelligence optimization algorithms and apply them to solve optimization problems in cloud computing application scenarios.In future,research on swarm intelligence algorithms such as invasive tumor growth optimization algorithm will continue to be in-depth to adapt to more compelx environment in cloud computing,and work for more real-world problems in other fields and scenarios.