解决FAST TCP缓存溢出相关问题的改进pacing technique算法和α参数调整算法  

Improved Pacing Technique Algorithm and α-adjusting Algorithm to Solve Problems Related to Buffer Overflowing of FAST TCP

作　　者：朱小松[1] 

机构地区：[1]南京邮电大学,南京210003

出　　处：《信息通信技术》2012年第1期60-65,共6页Information and communications Technologies

摘　　要：分析FAST TCP在缓存溢出发生时的性能,发现在缓存溢出场景中,收敛中的FAST TCP流经历严重的报文段丢失。相反,已经收敛了的FAST TCP流维持着高吞吐量和低报文段丢失概率。这种不公平是由FAST TCP缩减其窗口时的零传输率导致的。通过修改FAST TCP pacing算法,可以解决此问题。文中提出的α-adjusting算法,通过动态调整FAST TCP协议中的α参数来避免频繁的缓存溢出。通过分析ns2仿真结果,证明该算法在公平性和稳定性方面可获得令人满意的性能。By analyzing the performance of FAST TCP when buffer overflows occur.We find that,in this case,converging FAST TCP flows suffer from heavy packet losses.In contrast,converged FAST TCP flows maintain high throughput and low packet loss probability.This unfairness results from the zero transmission rate when FAST TCP decreases its congestion window.We tackle this problem by modifying FAST TCP pacing algorithm.Morevoer,we propose an α-adjusting algorithm to avoid frequent buffer overflows by dynamically tuning the α parameter of the FAST TCP protocol.Validated by ns2 simulations,this algorithm demonstrates satisfactory performance in fairness and stability.

关 键 词：FAST TCP 缓存溢出 节拍技术 α参数调整 算法 

分 类 号：TP333[自动化与计算机技术—计算机系统结构] TP301.6[自动化与计算机技术—计算机科学与技术]