基于容器和微服务的分布式演化计算平台的研究

RESEARCH ON A DISTRIBUTED PLATFORM FOR EVOLUTIONARY COMPUTATION BASED ON CONTAINER AND MICRO-SERVIC

钟慕尧

11849335

硕士

计算机技术

唐珂

2020-05-30

2020-07-08

哈尔滨工业大学

深圳

随着大数据到来，新产生的数据量呈指数增长，人们面临的优化问题的规模也随之增长，通常用于解决此类问题的演化计算方法在大规模的数据下无法令人满意。而分布式计算和云平台在传统的开发模式下，需要对于每一个优化问题开发针对性的计算应用，浪费了大量资源。后来出现的分布式启发算法平台通常有部署复杂、并行化程度不高、扩展困难、用户和开发者界面不友好、算法和问题支持不多等不足之处。本文旨在克服上述计算平台的不足之处，为演化计算提供一个部署便捷、扩展灵活、界面友好、覆盖广泛的分布式计算平台，探索了演化算法的分布式应用可行性，对平台进行了架构和功能需求分析，基于这些需求进行了技术选型并设计了整个平台的架构。本文利用容器技术和微服务架构对平台功能进行了划分，设计并实现了资源管理微服务、任务控制微服务、RESTful风格的用户接口。本文结合典型的分布式演化计算模型设计并实现了算法库，并定义了问题库的开发规范，为开发者提供了便捷的方式进行算法和问题扩充。详细介绍了演化计算平台的整个架构的实现方法，对平台中每个组件都进行了说明。此外，本文还选取三个典型复杂优化问题，利用这些问题解释了组合优化问题在分布式演化计算平台上的实现原理和方法，并测试了平台的设计功能。

With the advent of big data, the amount of newly generated data has grown exponentially, and the scale of optimization problems that people face has also increased. The evolutionary computation methods that are commonly used to solve such problems are not satisfactory under large-scale data. In the traditional development mode of distributed computing and cloud platforms, it is necessary to develop targeted computing applications for each optimization problem, wasting a lot of resources. The distributed heuristic algorithm platform that appeared later usually has deficiencies such as complex deployment, low degree of parallelization, difficulty in scaling, unfriendly user and developer interfaces, and insufficient support for algorithms and problems.This thesis aims to overcome the shortcomings of the above computing platform and provide a distributed computing platform with easy deployment, flexible expansion, friendly interface, and wide coverage for evolutionary computing. This thesis explores the feasibility of distributed applications of evolutionary algorithms, analyzes the architecture and functional requirements of the platform, selects technologies based on these requirements, and designs the architecture of the entire platform. This article uses container technology and microservice architecture to divide the platform functions, and design and implement resource management microservices, task control microservices, and RESTful user interfaces.In this thesis an algorithm library is designed and implemented in conjunction with a typical distributed evolutionary calculation model, and the development specifications of the problem library are defined, which provides developers with a convenient way to expand algorithms and problems. The implementation method of the entire architecture of the evolutionary computing platform is introduced in detail, and each component in the platform is explained. In addition, this paper also selects three typical complex optimization problems, uses these problems to explain the implementation principle and method of the combinatorial optimization problem on the distributed evolutionary computing platform, and tests the design function of the platform.

演化计算平台 分布式计算 容器 微服务 复杂优化

Evolutionary computing platform distributed computing container Micro-Service complex optimization

中文

联合培养

南方科技大学

钟慕尧. 基于容器和微服务的分布式演化计算平台的研究[D]. 深圳. 哈尔滨工业大学,2020.