Research On The Quality Of 3D Network Service

The integrated circuit industry has developed rapidly, and the integration degree has improved with the Moore’s law. Facing the increasingly complex business, because of the low communication efficiency, poor usability, higher power consumption, traditional bus structure has been replaced by NOC (Network on chip NOC) gradually.The network on chip has become an active research field. Along with the development of 3D IC technology and network performance needs, three-dimensional network on-chip (3D NOC) concept is proposed,3D NOC has advantages both of the 3D IC and network on-chip, effectively overcome the bottleneck problems of bus structure, and improves the performance of the system. Quality of service (of Service Quality, QoS) is the key technology of 3D NoC, along with the development of 3D NoC service, it becomes more and more important to provide the guaranteed quality service for the network communication on the chip. Therefore, it is important to study the quality of service of the network structure on the 3D chip.Aiming at the 3D NoC QoS problems, the 3D NoC QoS is a NP complete problem, because it involves many variables, such as delay, throughput, energy consumption and so on. As an optimization algorithm, genetic algorithm is widely used to solve all kinds of NP complete problems. In this paper, genetic algorithm is used as a routing algorithm to solve this problem. According to the fitness function, the delay and throughput can be synthetically considered, and the QoS parameters such as delay and throughput can be guaranteed, and the optimal or sub optimal path can be selected. In the process of algorithm design, combined with the characteristics of the scale of the network on the three-dimensional chip, the coding method of the visual path ID identification is adopted to avoid the decoding problem. In order to improve the convergence speed of the algorithm, an elite genetic algorithm is used which directly copy the optimal individual to the next generation.The genetic algorithm is integrated into the simulator routing algorithm module, which can be compared with the advantages of genetic algorithm and XYZ routing algorithm in the same environment. Considering the disadvantages of static source routing and three-dimensional chip scale of the network and structure characteristics, modify the source code of the simulator, trealize the dynamic routing selection mechanism by the second calculation, effectively avoid static source routing path which is no longer applicable to the new condition in the network and easliy occuring congestion problems.According to the different size and performance parameters of 3D NoC, different verification schemes are designed.The simulation results show that the delay, throughput and energy consumption of the genetic algorithm are better than the XYZ dimension order routing algorithm in the network structure and communication conditions. Especially, improvement of average time delay is the most obviously. With the increasing of network traffic, genetic algorithm can guarantee the quality of service than the XYZ routing algorithm better.