The Design And Implementation Of Dynamic Load Balancing Scheduling System For Cloud Gateway

In large data centers,cloud gateway devices play a decisive role as the entrance and exit of data center traffic.With the rapid growth of data traffic in the Internet era,cloud gateways need to provide higher and higher performance to carry the rapidly increasing traffic throughput.In the face of this rapidly increasing demand for data packet processing and the complex Internet traffic environment,the current cloud gateways in data centers not only need to meet high-performance packet processing capabilities,but also need to guarantee the tenant's quality of service level agreement,which is often not possible.Good balance between service quality and equipment resource utilization.On the one hand,in order to ensure service quality,manufacturers must provide sufficient resources for the expected peak load at the expense of equipment utilization.Because of the burst of Internet traffic,the average load and peak load of the traffic are tens of times different,and over-allocated resources cannot be fully utilized,resulting in low equipment utilization.On the other hand,if in order to improve equipment utilization,reduce resource allocation and reduce operating costs,when the load fluctuates,it will face the problem of local resource shortage and cannot effectively guarantee the quality of service.The essence of improving equipment utilization is a load balancing problem.On multi-core processors,the processing power of each CPU core is fully utilized by evenly distributing the load inside the server to achieve high equipment utilization.However,load balancing and scheduling in the cloud gateway mass data forwarding scenario is very challenging.In many traditional scheduling methods,some negligible overheads will be amplified by massive data,which will have a serious impact on the quality of service,and cannot guarantee the utilization of equipment and the quality of service of tenants at the same time.The utilization of equipment and the quality of service of tenants.Therefore,how to solve the contradiction between equipment resource utilization and tenant service quality in the cloud gateway scenario is the main content of this paper.In response to the above problems,this article has conducted an in-depth study on the device load balancing problem in the cloud gateway mass data forwarding scenario,and designed a cloud gateway-oriented load balancing scheduling system based on the data plane development kit(DPDK)to achieve high resource utilization and highquality cloud gateway data processing.First of all,the system is integrated into DPDK in the form of a library,without interfering with the processing of application logic,allowing the original application logic to maintain high-performance data packet processing capabilities.Secondly,this article uses the traffic scheduling technology provided by some existing commercial network cards,combined with the scheduling algorithm on the software,in a more optimized way to effectively and quickly adjust the load input difference between the CPUs in the server,and make full use of the multi-core CPU processing advantages to improve the overall utilization of the equipment.Finally,aiming at the instantaneous burst problem of traffic,this paper realizes the traffic burst detection mechanism based on the hardware queue of the network card by studying the network card receiving mechanism,which can detect the burst traffic before the data packet overflows the receiving queue,and schedule it in time to reduce the data packet occupancy rate of the traffic burst queue to avoid core overload.The experimental results show that,compared with some existing work,the system design in this paper has high-performance packet forwarding capabilities,and can effectively schedule the internal load of the cloud gateway server,allowing the cloud gateway device to make full use of the advantages of multi-core processor.In the same scenario,the device can have higher processing performance.