In recent years, with the rapid development of information technology, network applications in high-concurrency scenarios have become increasingly diversified. Massive network traffic has brought tremendous challenges to the network devices that support business operations and placed higher demands on network traffic processing performance. To address the core issue of improving network traffic processing performance, this paper focuses on two main aspects: improving the network traffic processing capability of a single network device and optimizing the cluster traffic distribution mechanism.
Firstly, to address the problem of the network traffic processing capability of a single device, this paper uses DPDK technology at the software level to enhance it. Furthermore, due to DPDK’s lack of a protocol stack and not supporting OSI model layer-four or above protocol services, this paper designs and implements a user-space protocol stack based on DPDK’s technical characteristics and provides a set of interface functions. Developers can use DPDK technology for TCP and UDP data communication by calling corresponding interface functions.
Secondly, to address the problem of cluster traffic distribution mechanisms, this paper proposes a dynamic weighted feedback load balancing algorithm for performance optimization. This algorithm periodically collects information such as backend network device CPU utilization rate, memory utilization rate, disk I/O and network bandwidth utilization rates and sets load distribution strategies based on current backend device status. Using this algorithm allows for updates in load allocation methods based on current time periods when networks are in different states. The application and porting of Nginx were used for performance testing on both the user-space protocol stack designed in this paper as well as the dynamic weighted feedback algorithm using response time and actual concurrency numbers as performance test indicators. Experimental results show that our proposed user-space protocol stack and dynamic weighted feedback algorithm significantly improve network packet processing performance.
Finally, we applied our user-space protocol stack along with our dynamic weighted feedback load balancing algorithm to a data forwarding module in a signaling detection and parsing project. Results showed that applying our designed user-space protocol stack along with our dynamic weighted feedback load balancing algorithm effectively improved the efficiency of signaling detection and parsing business operations in the project.
This paper focuses on two aspects: improving network traffic processing capabilities of a single device and optimizing the cluster traffic distribution mechanism. We design a user-space protocol stack based on DPDK technology and propose a dynamic weighted feedback load balancing algorithm, both of which were applied to a data forwarding module in a signaling detection and parsing project. The experiments demonstrate that our research method is reasonable and effective, significantly improving network packet processing performance and the efficiency of signaling detection and parsing business operations.
![error: ‘%s’ directive writing up to 63 bytes into a region of size between 33 and 96 [-Werror=format-overflow=]](https://linuxcpp.0voice.com/zb_users/upload/2023/05/202305162216114425773.png)
