Design And Implementation Of Load Balancing System Based On OVN Architecture

The data center is an important infrastructure that provides computing and storage resources for many network services.With the development of cloud computing and virtualization technology,the number and scale of data centers are constantly expanding,and traffic aggregated to the data center is increasing exponentially.Load balancing has become a necessary basic network function for cloud data centers.This mechanism can effectively handle concurrent traffic and improve resource utilization,which is crucial to improving the performance of cloud services hosted in data centers.After years of research and practice,the SDN network architecture has made great breakthroughs in both the control plane and the data plane.Many open source and commercial controllers have appeared to manage the data center network and provide various network functions in a centralized control manner.The data plane has evolved from a software switch to a hardware switch with a protocol-independent architecture,gradually combining flexibility and high forwarding performance.The SDN architecture has been widely used in data center networks,and the deployment method of load balancing modules has also undergone new changes.This module can be deployed not only on high-cost proprietary hardware,but also on programmable software or hardware switches.This thesis studies the SDN controller,data plane programmable technology,load balancing algorithm and deployment scheme,designs and implements a network load balancing system based on SDN architecture.The system incorporates the load balancing function into the control layer management.Data layer forwarding acceleration is achieved through smart NIC and P4 switches,meeting the load balancing requirements of enterprise-level data centers.The main work of the thesis is as follows:(1)A complete data center network load balancing system is designed.Based on SDN,a basic network architecture with OVN as the centralized controller,OVS and P4 switches as the data forwarding layer is designed.This architecture enables the load balancing module to be deployed on software or hardware gateways and provides network configuration services.(2)A dynamic feedback load balancing algorithm based on node load is designed.The algorithm adjusts the allocation weight of backend nodes from a global perspective in a timing or thresholdtriggered manner according to service characteristics and feedback information.The experimental results show that after the system uses the feedback equalization algorithm,the response time of computationally intensive requests can be reduced by an average of 33.5%,and the throughput of storage-intensive requests can be increased by an average of 37.56%,which can effectively utilize resources.(3)Implemented a load balancing system and performed functional and performance tests.The control layer uses a feedback load balancing algorithm to implement weighted load balancing decisions.The forwarding layer uses OVS and P4 switches as software and hardware gateways to deploy load balancing modules respectively.The P4 switch is simulated by BMV2 and communicates with the logical network using VXLAN technology.In addition,the cloud management platform based on Open Stack and Web services uses the OVN plugin to implement the network configuration module.The test results show that the network configuration module of the system is effective.The system can use the feedback algorithm to achieve load balancing decisions.The backend node is provided with the function of health check,and the heterogeneous equipment in the network can work in coordination.