Research About Network Traffic Control On Linux

The Quality of Service (Qos) includes the goal to provide more reliable service of networks. To keep the network security access under control and the provision of various quality of services, traffic control is occupied to the routers and servers. Bandwidth allocation ensures the limited network resources used efficient and fairly. For enterprises, bandwidth allocation strategies help to prevent wasting limited network bandwidth and reduce security risks. So far, multiple-network interface cards (NICs) technology is widely used in the enterprises situation, while there is lack of a traffic control algorithm to distinguish flow types with priorties using for the case. On the other hand, there should be a handy and useful bandwidth control method with a friendly interface, for various users like network administrators, developers or general users.Linux is a widely used operating system adopted in network equitment, which offers a large set of Qos methods for classifying and scheduling network traffic. This thesis studies Linux’s traffic control mechanism including principles, functions, algorithm and the two major modules of bandwidth management tools-iptables and tc. Tc manages qdisc, class and filter for traffic control, and iptables can distinguish flow types from each others. Both of them is a user space tool from the kernel modules. In the algorithms part, this thesis deeply studies queuing discipline with or without classes, including First in First out (FIFO), Stochastic Fairness Queueing (SFQ), Token Bucket Filter (TBF), Class Based Queuing (CBQ) etc. In this article, it provides an introduction and overview of this queuing disciplines in Linux.This thesis proposes and implements an improved shared-token bucket filter algorithm working to meet the multiple network devices situation. The original algorithm TBF is one of the most commonly used methods in traffic control of network management to ensure quality of services. The improved algorithm can distinguish network devices by priority assignment, that makes free tokens can be shared between network devices to increase bandwidth utilization. This article demonstrates its principle and purpose, also shows its performance in a real environment. Test results show the shared-token bucket filter algorithm can work effectively in traffic controlling and free bandwidth sharing. Compared with TBF, the average bandwidth utilization of STBF algorithm from73%to97%in the case of only the card with a high priority busy. On the same load condition, STBF shows a faster average delivery time than TBF and HTB.The tc tool has much complexity command formats and parameters, also has limited support for flow filters, In addition, this thesis proposed a new bandwidth control method to these problems--tctables. This new method works for Linux’s traffic control, combines the strengths both of tc and iptables, can implement a efficient traffic control function and mutiple flow filters. This tctables methods include two part:the tctables language and the tctables interpreter. The tctables compiler will translate the sentences writing by the tctables language to a shell scripte with tc and iptables commands.Principles and implementations of each part are present in this article.