Energy-saving deployment method of reconfigurable fat-tree hybrid data center network

Abstract

The invention belongs to the technical field of maintenance or management of digital switching networks, and discloses an energy-saving deployment method of a reconfigurable Fat-Tree hybrid data center network, and obtains a wireless topology map of a hybrid data center according to different ToR deployment reconfigurable wireless technologies; Combining the wired topology features of Fat-Tree to construct the deployment scheme population and fitness function; using the greedy random adaptation algorithm to realize the joint wired and wireless optimization of the Fat-Tree data center network. The present invention fully considers the relationship between ToRs in the multi-pod layer structure under the Fat-Tree architecture through the fitness function; utilizes the high-efficiency and energy-saving characteristics of wireless links between pods, and considers the uncertainty of traffic distribution in the network to make the hybrid network It has higher energy saving efficiency. The present invention adopts a greedy random adaptation algorithm, fully combines the approximation performance of the greedy algorithm and the global convergence characteristic of random search, and obtains deployment scheme individuals with higher adaptability.

Description

technical field [0001] The invention belongs to the technical field of maintenance or management of digital switching networks, and in particular relates to an energy-saving deployment method of a reconfigurable Fat-Tree hybrid data center network. Background technique [0002] At present, the existing technologies commonly used in the industry are as follows: In recent years, with the rapid development of cloud computing and big data, more and more data centers have been established to provide a large number of online applications and infrastructure services. Today's data centers are designed with hundreds of thousands of servers, which are interconnected through network devices such as switches, routers, and high-speed links to build a data center network (Data Center Network, DCN). With the ever-increasing demand for services and applications, DCNs are required to meet various requirements including high reliability, high cost-effectiveness and high throughput. Although ...

AI Technical Summary

Problems solved by technology

Due to the limitation of reconfigurable wireless directional transmission range, in the hybrid data center network, the utility of establishing wireless is affected by its inherent wired topology, and the Fat-Tree hybrid data center network based on the Fat-Tree wired topology, due to its wired Topological characteristics, establishing wireless links between pod layers is more effective and energy-efficient than establishing wireless links within pods, and usually the connectivity of wireless topologies in hybrid data centers is not uniform in space, while the existing Fat-Tree Hybrid data center networks are built without considering these topological characteristics of wired and wireless

Therefore, the construction of the existing Fat-Tree hybrid data center network has the defects of low wireless utilization efficiency and uneven deployment of wireless links

Through the optimal deployment of the network in the Fat-Tree mixed data, the usage efficiency and energy saving efficiency of the wireless network can be significantly improved; regardless of the wired topology characteristics of the network, it is difficult to efficiently use the wireless link only by simply deploying the wireless link

[0003] To sum up, the problems existing in the existing technology are: the existing reconfigurable wireless deployment of the Fat-Tree hybrid data center network does not comprehensively consider the wired and wireless topological characteristics, resulting in the inability to fully utilize the advantages of wireless, resulting in wireless Waste of resources; due to the unbalanced deployment of wireless networks, the establishment of wireless links cannot effectively relieve congestion and reduce network energy consumption

To solve this problem, it is necessary to consider maximizing the wireless links between the pod layers and evenly deploying the wireless links between the pod layers. As the network scale increases, the size of the wireless topology also increases, resulting in the possibility of obtaining excellent deployment results. more difficult

Images