【作者】 龚盼芸；

【导师】 谭连生；

【作者基本信息】 华中师范大学 ， 计算机系统结构， 2012， 硕士

【摘要】 伴随着网络用户的急剧增加,网络拥塞控制问题显得越来越重要。然而传统的TCP拥塞控制协议是基于端系统的流量控制的,它们已经无法适应这些变化。微观经济学中效用和价格概念的引入为拥塞控制的研究提供了新的方向。价格工具能够有效的使各种市场达到平衡条件,它们同样可以用来控制复杂的网络系统。网络定价能够更好的来理解网络管理并让用户理性使用网络,而且对网络服务的适当定价能使用户选择合适的服务质量,满足应用服务器的平衡要求,以及正确评估不同级别和质量的服务。效用概念表达了用户对其所获得的网络服务的满意程度。效用最大化(NUM)问题已经得到多方面学者的注意,它结合了计算机网络、数学优化方法、控制理论和经济学等多方面的知识。拥塞控制和资源分配的关系密切,合理的资源分配是拥塞控制的一种手段。效用最大化模型提供了一种新的资源分配研究方法,它结合了资源分配的效率和公平性,能够从这两个方而进行统一的研究,从而更全面的衡量网络资源分配。对于NUM的研究方向主要是改变效用函数或者增加约束条件,这样会产生不同的定价方案,直接影响拥塞控制算法。本文介绍了效用最优化的一般模型和它的求解过程,文中都采用的是梯度投影算法来计算链路价格,分析了它的基本性能。本文综合考虑用户和链路,引入链路重要性结合效用做了两方面的工作：一是基于链路重要性提出了新的最短路算法,将其与效用模型相结合,产生了一种新的拥塞控制方法；二是将其引入到考虑链路成本的效用模型中。另外介绍了一个新的效用模型,它改进了原有的链路价格定价方案,引入了链路利用率变化因子,与原有链路拥塞价格一起来调节源端的发送速率。本文的目的分为两个方面：一是体现使用价格机制来管理网络的重要性和可行性。另一个就是说明将链路重要性和效用模型相结合的可行性,并通过最终的对比分析可以看出它的有效性,反映了链路端反馈的信息除了可以促使源端调整发送速率外,还可以优化路由选择。更多还原

【Abstract】 Accompanying with the rapid increase of Internet users, it will be more important to control congestion. However, the traditional TCP congestion control protocols based on end system have been unable to adapt to these changes. Utility and prices in microeconomics provide a new direction for congestion control. Price instruments are useful in achieving market balance conditions in various markets. They can be also used for control of composite network systems. Internet pricing is essential not only for better possible understand of the network operation but it may also provide means to achieve rational behavior of the network users. Dynamic pricing may be useful, in particular, for balancing demand for access to application servers and for proper valuation of different classes and qualities of service. Utility can measure the user’s satisfaction or happiness about the service they acquire. Network Utility Maximization (NUM) have caught the attention of many scholars, it involves various knowledge of the computer network, the mathematical optimization methods, control theory and economics. Congestion control involves with resource allocation, reasonable allocation of the network resources will control congestion. Utility maximization model optimizes allocation which involves efficiency and fairness, so it measures resource allocation more comprehensively.The main research direction for NUM is to change the utility function or increase the constraint conditions; this will produce a different pricing method, and then influence the congestion control algorithm. This paper introduces the general model of NUM and how to solve it:using gradient algorithm to calculate link prices, and analyses its basic performances. This paper consider utility involved link importance on tow points:First, proposing a new shortest path algorithm based on link importance, which combined with utility model, producing a new congestion method; Second, introducing a new utility model, which improves the original pricing method, introduce a link utilization change factor, together with the original link congestion prices to adjust the source rate.The objective in this paper has two sides:One is to present emerging, opportunities to use price mechanisms for management of computer networks. Another is to present opportunities to involve utility and the link importance, and the final contrast analysis shows the validity, in this method, the information which links feedback can not only make the source adjust rate, but also optimize routing.更多还原