Researches On The Technology Of Suppressing NACK And Reliable Multicast Based On Reliable Active Node

With the developing of Internet, many new applications have come forth, such as software distribution, video conference, distance learning and shared board. All of the applications have a common feature, i.e. one to many or many to many reliable multicast communication, and the potential receivers have hundreds and thousands. Therefore, the applications require the sender to deliver the data efficiently. However, the technology of Internet faces great challenges because the architecture of the Internet is designed for point to point communication. The technology of deploying the multicast communication over Internet has been proposal to solve the problems. This dissertation focuses on how to solve the following problems: NACK implosion, local recovery the missing data and decreasing the burden of the sender. In short, we obtain a few breakthroughs as follows:Firstly, we propose two methods of suppressing NACK based on the timer. There are several problems in scalable and reliable multicast, such as, feedback implosion and local recovery, etc. Aim at to solve the problem of feedback implosion, we discuss carefully how to set the timer to avoid the NACK implosion in receiver-initiated, integrate FEC system. We propose two methods: divided time equally and based on exponential distribution to set timers, divided time unequally and based on exponential distribution to set timers. By analyzing for up to 10 receivers, both of them can avoid the NACK implosion and feedback delay due to timers is low. The result also shows the second method is better than the first one. Secondly, we propose a reliable multicast protocol based on reliable active node(RANRM). As reliable multicast over Internet has many problems caused by the IP protocol that does not guarantee the reliability. In the layering model, RM layer includes management sub-layer and reliability sub-layer. Our research majors the reliability sub-layer. In RANRM, the active nodes are divided into two classes: the reliable active nodes and the forward active node. The reliable active nodes dealwith some tasks relating to the reliability. In other words, the reliable active nodes implement the reliability sub-layer. The forward active nodes have the same function as the traditional IP nodes. Using the reliable active nodes in the active network can guarantee that the data is reliably delivered from one reliable active node to another. With the flexibility of active network, the application layer should customize the quantity and the location of the reliable active nodes in active network. We suggest that the reliable active nodes should be those which locate at the "edges" of the backbone links. In short, RANRM has many advantages, to avoid feedback implosion because only a few feedback messages are handled by the sender, to reduce the burden of handling loss feedback by using the ability of local loss recovery, to reduce the recovery latency by finding transmission errors in time, and also to efficiently reduce network bandwidth consumption by not transferring the corrupted data. Therefore, the protocol can solve the problems of the reliability and scalability, and can combine the reliability with high transmission efficiency. Theoretical analysis shows that the protocol presented here is, comparing with the traditional network protocol, better for bandwidth and loss recovery delay.Thirdly, CPN has been adopted to describe the RANRM. As the CPN of RANRM is constructed, they are investigated by means of the CPN simulator. All details of the reached markings are inspected. Moreover, we complement simulation with the construction of the state spaces.Fourthly, ANTS has been adopted to simulate RANRM, and the simulation results show the same conclusion of theoretical analysis. The program of simulation can guide how to implement the RANRM protocol.