【作者】 周颖；

【导师】 赵岳松；

【作者基本信息】 武汉理工大学 ， 计算机应用技术， 2003， 硕士

【摘要】 而今随着因特网的迅速发展，无论在企业网还是在广域网如Internet上，网络业务量的发展都超出了过去最乐观的估计，尤其是各个网络的核心部分，其数据流量和计算强度之人，使得单——设备根本无法承担，而如何在完成同样功能的多个网络设备之间实现合理的业务量分配，使之不至于出现一台设备过忙，而别的设备却未充分发挥处理能力的情况，就成了一个问题，对此集群服务发起了挑战，并给出了令人满意的决方案。 集群服务可分为高可用集群服务和高性能集群服务，高可用性主要针对系统在网络上的不间断通信服务，软硬件中的接管机制实现资源可利用性；而高性能则是主要针对系统在服务中能提供更强大的功能，如更大的吞吐量，更低的延迟，良好的负载均衡机制等。本论文对LVS集群服务中的WEB服务的可伸缩性方面进性了深入的研究。 可伸缩集群系统首先是只有可扩展的特性，即具有强大的服务功能具有高性能；然后是具有高可有性，即是在单点和多点失败的情况下能保持不间断的服务的特性。可伸缩性是保证在总体上提高系统性能的情况下，提供高可用的集群服务。 本人的研究探讨成果：Linux下构建可伸缩WEB集群服务。研究主要从：1)分包重写——这个分布式处理请求事务的方法，允许所有主机参与请求过程的重定向；到2)MPLS多协议标签交换技术——用MPLS网络可仅通过L-7web交换实现基于内容路由、客户亲和力、区分服务、负载均衡功能，具有更高的性价比。实现了可伸缩的包分发机制。然后3)用链接聚合和连接接合等方式优化了可伸缩集群服务中的包转发连接的实现机制。更多还原

【Abstract】 With the rapid development of Internet now, no matter in enterprise’s network is or like Internet in WAN , The development of portfolio of the network has gone beyond the most optimistic estimation in the past, especially the key part of each network, The traffic and calculation are big in intensity, make the single machine unable to be born at all, And how realize that rational portfolio is distributed among a lot of network equipments of finishing same function, It is unlikely to make present a set of equipment too busy , but other equipment have not been full utilized, On the issue that have become one, cluster service initiate and challenge , give and publish a satisfactory schemeCluster services can be divided into high availability cluster services and high performance cluster service , High availability is directed against the incessant communication service in the network of the system mainly, The takeover mechanism in the software and hardware realizes the utilizing that resource can; High performance can offer stronger function in serving to the system mainly , For instance larger handling capacity, the lower delayed , good load balancing mechanism ,etc.. Thesis here is about deeply research of scalable WEB cluster serviceThe scalable cluster service system is to have expanded characteristics at first, namely have strong service functions and have high performance; Have high availability is namely can keep incessant characteristic of service while there has failed node. The scalability guaranteeing in case of improving systematic function generally, Offer cluster service of high and available system.The achievement of my research: Structure scalable WEB cluster services under Linux. Study mainly from: 1) Distributed Packet Rewrite-This distributed method to request transaction, allow all host computer participate in redirection; and then reach 2) MPLS ( Multiprotocol Label Switching) technology -Can exchange realizing because of content-based routing, client-server affinity, client-specific service differentiation, server load balancing function only through L-7web in MPLS network, provided at a much improved price-to-performance ratio. Implement a mechanism of scalable distribute packets . Finally 3) Get with chaining converge and connect splicing methods optimize scalable cluster service.更多还原