东篱科研大数据发现系统（DRDS）

位置：成果数据库 > 期刊 > 期刊详情页

一种保证时延的关键流路由调整算法

期刊名称：软件学报（已录用）
时间：0
分类：TP393[自动化与计算机技术—计算机应用技术;自动化与计算机技术—计算机科学与技术]
作者机构：[1]北京邮电大学网络与交换技术国家重点实验室,北京100876
相关基金：Supported by the National Natural Science Foundation of China under Grant Nos.90604019, 60502037 （国家自然科学基金）;the Specialized Research Fund for the Doctoral Program of Higher Education No.200800131019（高等学校搏士学科点专项科研基金）; the New Century Excellent Talents in University No.NECT-07-0109（新世纪优秀人才支持计划）; the National High-Tech Research and Development Plan of China under Grant Nos.2006AA01Z235, 2007AA01Z206 （国家高技术研究发展计划（863））
相关项目：与传统互联网业务友好的P2P流量控制模型和方法的研究

关键词：网络管理, 负载均衡, 服务质量, 路由调整, 最大最小化, network management, load balancing, quality of service, routing adjustment, max-min

中文摘要：

互联网中流量分布的不均衡会导致网络拥塞、网络资源得不到有效利用．而为了负载均衡，现有算法调整路由又会造成新路径过长，服务质量降低．提出了一种路由调整算法LCBA（length-constrained most balanced algorithm），在保证时延的基础上降低网络最大带宽利用率．基于Abilene2网络拓扑和真实流量的实验结果表明，LCBA算法能够有效缓解骨干网拥塞，最多可以降低最大带宽利用率近50％．仿真实验结果显示：与现有算法相比，该算法能够同时满足关键流路径长度和最大带宽利用率两方面的要求．此外，算法复杂度为O（N^2 logN），好于大部分路由调整算法．

英文摘要：

The unevenly distributed traffic of Internet may lead to congestion and underutilization of network resources. Current methods of routing adjustment for load balancing often get a new path with excessively long length that can exacerbate the quality of service. This paper presents a new flow routing adjustment algorithm named LCBA （length-constrained most balanced algorithm）, which can decrease the maximal bandwidth utilization rate of network with delay guaranteed. The experiment based on Abilene2 network topology and traffic data shows that LCBA can mitigate the congestion of backbone network effectively and the maximal bandwidth utilization rate can drease nearly 50% at most. This paper also evaluates the algorithm with an emulation method. Compared with current methods, the new algorithm can satisfy the demand of not only the length of key flow＇s path but also the maximal bandwidth utilization rate. Moreover, the theoretical analysis proves that the computational complexity of LCBA is O（N^2 logN）, which is better than that of a majority of methods nowaday.

同期刊论文项目