Concurrency Optimization Of Embeded TCP/IP Protocol Stack LWIP

Embedded TCP/IP protocol stack LWIP is mainly used in embedded devices with limited resources. Its data structures and algorithms are designed to reduce memory usage and code size, on this basis to meet the communication function and not to large-scale network data transmission. However, it can't meet the real-time and high concurrency requirements of security gateway devices. Hence the data structures and algorithms of LWIP should be optimized for concurrency.The author completes LWIP performance bottleneck analysis, LWIP optimization plan development and implementation, and performance comparative analysis of LWIP after optimization independently. Then the gprof is used to check system performance bottlenecks. The performance bottleneck of LWIP is mainly the TCP session table list structures and SELECT interfaces. For the TCP session table list structure, the most efficient query structure-hash structure is used; For SELECT interface, efficient EPOLL interface is used; And the local port selection algorithm to establish a connection initiatively is optimized; In order to improve performance, the simultaneous multithreading optimization techniques is tried. Then use the software and hardware testing devices to compare performance before and after optimization of the LWIP module. At last we come to the conclusion of the optimization results.The overall performance of LWIP has improved more than 30% after optimizing. The real-time and concurrency has increased dramatically to meet the performance requirements of security gateway products and to achieve the desired optimization goals. This module has been integrated into the current 5.6.5 version engine code of NSFOCUS, which can be stable and has a good run concurrent performance.