Research On Key Technologies Of Network Transport Protocols Based On Resource Pooling

Resource within Internet can be divided into data resource and network resource. The first one is understood as the content of a specific service (such as video, voice, web pages, etc.). The second one is used to describe the transmission condition (such as bandwidth, delay, jitter, etc.) which is provided by the network when the user is acquiring service. In the epoch of knowledge explosion, the volume of data resource is getting bigger and bigger. The new-style methodologies of acquiring data resource emerge in an endless stream. And the requirements of users are increasing gradually.Due to the development of hardware technology, the performance of network equipment has been upgraded drastically. Many wireless transmission technologies are providing more choices for user to access the Internet. Each network interface can be assigned an individual address by terminal equipment to achieve multi-homing in IP layer. However, the traditional transport protocols only support single path transmission for the sender and receiver. Therefore the network resource can not be fully utilized.It is easy to find the contradiction between the data resource which is getting complex and the network resource which is increasingly intense. Consequently, resource pooling in a reasonable way is a critical topic for academic research. The work of this article is based on such background. The main contents are:(1) According to the conclusion of current transport protocols, the methods and approaches of resource pooling are given and a modeling scheme for multipath transport throughput is proposed. The problems of expanding single path transport protocols to multiple paths environments are analyzed. The changing principles of drop rate, round trip time and congestion window are also considered. In order to reduce the computational complexity of model, an optimization algorithm is given as well. This part provides theoretical basis for the research of resource pooling.(2) A modeling method for receive buffer blocking (RBB) is proposed after analyzing the reason of RBB. As a result of packet loss, data packets might not be able to reach receiver side in sequence. When using multiple paths for transmission, the difference of performance parameters might also lead to the same situation. And the main reason of RBB is out-of-order packets. This part proposes some potential questions for the research of resource pooling.(3) A method for measuring network one way performance parameters is proposed in multipath environment. Three specific applications are also given:retransmission policy design, bandwidth estimation and shared bottleneck detection. In traditional client and server mode, such applications can be used in relieving RBB, designing packet scheduling scheme and guaranteeing TCP friendly, respectively. This part considers the importance of one way performance parameters in multipath situations and provides a specific solution for resource pooling.(4) Two different methods for comparing one way delay are implemented into a certain multipath transport protocol. A delay-based packet retransmission policy is designed according to the loss type and the state of path. In order to increase the ability of handling path failure, potential failure state is introduced into the policy. Simulation result shows that both of implementations can achieve acceptable accuracy. The retransmission policy is able to relieve RBB in a certain extent.(5) The difference delays of different Internet paths on forward and backward directions are measured. A connection identifier dynamic analytic mapping method is proposed based on the design conception of universal network. When a terminal has multiple access modes, it will be very helpful for designing protocols which are appropriate for resource pooling, if the difference values of one way delay on different paths are available. Such information also provides the basis of measuring other one way performance parameters in practical networks. The main goal of this part is to use current achievements in new generation network design.