Research On Acceleration Technology Of IPSec VPN Base On GPU

With the rapid development of Internet, computer network to people’s life and workhas brought great convenience, but the problem of information security is increasinglybecoming the focus of attention. IPSec VPN is based on IPSec standard VPN technologyor network referred to, by encryption and authentication mechanisms to improve the secu-rity of the network. However, encryption and authentication consumes more system re-sources and reduce the system throughput, which affects the performance of the network.Use of GPU’s powerful hardware level multithreaded parallel computing abili-ty,through the CUDA (Compute Unified Device Architecture, Unified computing equip-ment structure) to the IPSec VPN takes longer3DES, AES and RSA encryption and au-thentication algorithms such as parallel processing. And according to the GPU only suita-ble for processing large data, and the actual network smaller message occupies a percen-tage of the situation. In order to guarantee the system real time, this paper proposes aCPU_GPU system model in the system at the same time, the existing CPU algorithm andthe GPU algorithm, according to the length of the packet dynamic selection algorithm im-plementation CPU or GPU algorithm. The algorithm level and system level optimizationto improve the whole the purpose of IPSec VPN system performance.CUDA parallel algorithm for test analysis, show that the optimized algorithm encryp-tion speed relatively common CPU algorithm is high, especially in dealing with large datacan show better performance. Commonly used by application layer protocol HTTP, FTPand SMTP simulation of the actual network flow of optimized IPSec VPN system test,compared to CPU system, GPU system and CPU_GPU system experimental results showthat CPU_GPU system in guarantee high throughput can also keep low delay, give fullplay to the CPU and GPU their advantages, allowing the system to achieve optimal per-formance.