IDP平台负载均衡技术的研究与实现

发布时间：2018-04-01 18:49

本文选题：集群　切入点：负载均衡　出处：《北京邮电大学》2010年硕士论文

【摘要】： IDP(Integrated Data-Service Platform)是开放的、模块化的、基于标准的基础平台,它向下提供与短信网关、彩信中心、BOSS (Business Operation Support System,运营支撑系统)、WAP (Wireless Application Protocol,无线应用协议)网关等功能实体的通信,向上提供用户鉴权、短信/彩信编辑发送、用户管理等公共功能。其主要能力和结构建立在满足现有移动数据业务的基础上,包括彩信、短信和WAP等。基于IDP平台可以进行多种业务的开发和整合,包括手机报、手机广告、彩信营销等等。基于IDP平台的多种业务已经成功在国内多个省份商用,随着业务种类和业务用户量的不断增加,各种业务对IDP平台的可用性和可扩展性要求越来越高。为应对越来越严重的平台负载,基于集群实现IDP平台的负载均衡是十分必要的。负载均衡是由多台服务器以对称方式组成一个服务器集群,每台服务器都具有同等地位,均能单独对外提供服务。通过特定的负载均衡技术,将外部请求根据服务器集群中各服务器上的负载状况合理分配到某台服务器上,籍此大幅提高获取数据的速度,提高服务器的整体处理能力,解决海量并发访问问题,并提高可靠性,可用性,可维护性,最终目的是加快服务器的响应速度,从而提高用户的体验度。本文将重点论述基于负载均衡技术对现有平台增加集群的支持,并针对IDP平台的特点,提出适合IDP平台的负载均衡算法。 IDP平台所面临的是如何及时处理大量数据的问题,具体的说就是针对消息下发任务,创建大量的消息包并进行下发处理的工作,这和传统的负载均衡系统存在着较大的差异。我们必须考虑到IDP平台中的负载均衡与传统的负载均衡系统的差异性,进行有针对性的设计。本文在深入研究负载均衡系统基本原理和常用算法的基础上,针对IDP平台的具体特点提出了基于任务分解的自适应负载均衡算法,并在IDP平台中实现了该算法,对实验系统进行了详细测试和性能评估。实验结果表明,所提出的算法性能开销较小,系统性能稳定可靠,能很好的适应IDP平台的特殊环境。基于任务分解的自适应负载均衡算法的基本思想是：基于负反馈的基本原理,根据各负载服务器节点当前可用资源和负载状况,进行均衡决策。在进行负载调度时,会根据各服务器节点不同的当前可用权值将庞大的消息下发任务分解为若干个大小不等的子任务,然后将每个子任务分配到权值大小与子任务大小匹配的某个服务器节点上,从而动态适应负载的变化。采用基于任务分解的自适应负载均衡算法后的IDP平台,既充分利用了各个节点的服务器资源,又保证了对任务的及时响应。
[Abstract]:The IDP(Integrated Data-Service platform is an open, modular, standards-based basic platform that provides down communication with functional entities such as SMS gateways, MMS's boss Business Operation Support system, Operations support Systems (OSS), WAP Wireless Application Protocol (WAP) gateways, etc. Provides user authentication, SMS / MMS editing and sending, user management, etc. Its main capabilities and structures are based on satisfying existing mobile data services, including MMS, SMS and WAP. Based on IDP platform can be used to develop and integrate a variety of services, including mobile newspaper, mobile advertising, MMS marketing and so on. Many kinds of services based on IDP platform have been successfully used in many provinces in China. In order to deal with the more and more serious platform load, it is necessary to realize the load balance of IDP platform based on cluster. Load balancing is a cluster of servers composed of multiple servers in a symmetrical manner. Each server has the same status and can provide services separately. According to the load condition of each server in the server cluster, the external request is allocated to a certain server reasonably, so as to greatly improve the speed of obtaining data, improve the overall processing ability of the server, and solve the problem of massive concurrent access. And improve reliability, availability, maintainability, the ultimate goal is to speed up the response speed of the server, and thus improve the user experience. This paper will focus on the load balancing technology based on the existing platform to increase cluster support, According to the characteristics of IDP platform, a load balancing algorithm suitable for IDP platform is proposed. IDP platform is faced with the problem of how to deal with a large number of data in time. Specifically, it is to create a large number of message packets and send them for the task of message sending. This is quite different from the traditional load balancing system. We must consider the difference between the load balancing in the IDP platform and the traditional load balancing system, and carry on the targeted design. On the basis of deeply studying the basic principle and common algorithms of load balancing system, this paper proposes an adaptive load balancing algorithm based on task decomposition for the specific characteristics of IDP platform, and implements the algorithm in IDP platform. The experimental results show that the proposed algorithm has the advantages of low performance overhead, stable and reliable system performance, and can adapt well to the special environment of IDP platform. The basic idea of the adaptive load balancing algorithm based on task decomposition is: based on the basic principle of negative feedback and according to the current available resources and load status of each load server node, balancing decision is made. Large message sending tasks are decomposed into several subtasks of varying sizes based on the currently available weights of each server node, and then each subtask is assigned to a server node whose weight size matches the size of the child task, The IDP platform based on the adaptive load balancing algorithm based on task decomposition not only makes full use of the server resources of each node, but also ensures the timely response to the task.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2010
【分类号】：TN929.5

【相似文献】