Data Plane Connection Fast Recovery Mechanism Based On Data-Driven And Simulation Implementation

In the information age,with the rapid growth of network users and network demands,the data center business has grown massively.Once the network fails,it will cause immeasurable business losses.The longer the interruption time,the more serious the loss.Therefore,the swift recovery technology of data center network failure has become a research hotspot.Existing link failure recovery technologies based on the SDN architecture all require control signals to drive the SDN switch to recover from the failure,and at least two southbound interface communications are required,which makes the system find and restore the link failure time too long.The method of restoring the link failure based on the data signal driving system saves the communication time of the southbound interface.The existing method Blink is data-driven.However,Blink has long recovery time,poor robustness,and is not suitable for data center networks.Therefore,this thesis deeply studies the data-driven data plane connection rapid recovery mechanism and proposes AB(Adaptive Blink),which achieves outstanding time performance,strong robustness,and is suitable for fault detection and recovery systems in data centers.The main research work and innovations of this thesis are as follows:1.This thesis researches the mechanism and method of fast recovery of data plane connection.For data-driven connection recovery time performance issues,first find the key parameter RTO value of the failure recovery mechanism.Through studying the Linux kernel TCP code,find the limiting system time performance parameter TCP?RTO?MIN,and modify the kernel parameter to realize the connection recovery speed up,increase the time performance.Aiming at the robustness of data-driven connection restoration,a dynamic time sliding window scheme is proposed,so that AB can maintain the normal operation of system functions under the condition of network delay jitter.2.This thesis researches the design process of fast recovery of data plane connection based on P4.Based on P4 language,completes the design and implementation of AB,especially the design and implementation of dynamic time sliding window;and designes and implements AB application network,distributed cloud data center network,where AB will play different roles.In-domain AB and network edge AB respectively deploy ECMP load balancing strategies and Vx LAN technology on them,and proposes new ECMP rules based on P4.3.An experimental platform was built to verify the proposed method AB.Through the deployment of a simulation platform and an experimental environment,the time performance,accuracy and robustness of AB were analyzed.The experimental results show that the time performance of AB was very good,and the failure recovery time reached 30 ms.Meet the 50 ms requirement of the data center for failure recovery time.At the same time,AB has good robustness and can maintain the normal operation of system functions when the network is jittered.