Data center network based on prismoid structure

Abstract

A data center network based on a prismoid structure is characterized in that eight switches form a triangular prismoid unit network structure, and the triangular prismoid unit network structure comprises six five-interface switches and two four-interface switches. Firstly, the top surface and the bottom surface are respectively an equilateral triangle with the side length of L, and each equilateral triangle is formed by three five-interface switches which are connected in pairs; a four-interface switch is added to the center of each of the two equilateral triangles and connected with the otherthree switches pairwise, the switches corresponding to the upper bottom face and the lower bottom face are connected pairwise, and then the three switches at the vertexes of the bottom face trianglesare sequentially connected with the switches at the upper right corner to form a basic network topology unit. Except the switch in the center of the triangle, each switch is connected with the server. Extension is carried out according to the mode, a vertically connected data center network with the bottom side length being NL and h unit layers can be formed, and h and N are integers larger thanor equal to 1.

Description

technical field [0001] The invention relates to a data center network topology structure, in particular to a data center network based on a prism structure. Background technique [0002] With the rapid development of cloud computing and computer-intensive network technology, the world has entered an era of rapid information development. According to the data released by the International Data Corporation: As of 2018, the number of servers in Tencent's data center has reached 50 Microsoft has 2 million, and Amazon's data center has nearly 3 million servers. When such a large number of servers work simultaneously, a large amount of data will be generated, and the data center network as its infrastructure has become a research hotspot again. How to connect these servers through connecting devices and high-speed links to form a data center network with high performance and provide high-efficiency and high-quality services for the upper-level computer system is definitely a majo...

AI Technical Summary

Problems solved by technology

Although these structures solve the single point of failure problem of the traditional tree structure to a certain extent, they still have disadvantages such as expensive core switches and core routers, poor fault tolerance and poor scalability.

Images