提出一种适用于多线程边界网关协议(BGP)的无阻塞路由通告技术,消除了线程间共享队列并行访问冲突,借助提出的快速无阻塞共享队列SCLF实现了无锁并发操作,并改进了邻居播报过程以缓解cache共享数据带来的乒乓效应.通过对有锁方法与无阻塞方法的解析建模分析以及在Intel Xeon双路四核服务器上的实验结果显示,无阻塞方法有效改善了TBGP协议的路由通告耗时,能够更好地支持线程规模与邻居会话规模的可扩展性.
A non-blocking route announcement method is presented to accelerate the route propagation in threaded border gateay protocol (BGP). The method achieves concurrent route announcement among threads by effectively eliminating various contentions and adopting speedy concurrency lock-free FIFO (first in first out) (SCLF) algorithm, and also improves the advertising process with less cache thrashing. The performances of the lock-based scheme and newly-proposed scheme are evaluated by analytical models, and then the experiments on dual quad-core Intel Xeon server show that our scheme greatly reduces the runtime of route announcement with distinct parallelization and also provides better scalability of threads and neighbor sessions in comparison with the lock-based scheme.