自相似业务流下的星上交换全局优化缓冲分配算法
|
摘要
针对空间信息网络星上交换节点缓冲资源有限,提出一种适用于星上交换的缓冲优化分配算法。采用了Crossbar交换模型,在此基础上建立了虚通道自相似排队模型,通过计算每个虚通道缓冲溢出概率并采用遗传算法来实现缓冲资源全局优化分配。仿真结果表明,与均匀分配算法和贪婪分配算法相比,新算法具有更好的延时性能,在同等分组平均时延、业务流自相似程度为0.6和0.8的情况下,新算法比均匀分配算法可分别节省24.5%和26.4%的缓冲资源,并且分配效率比贪婪分配算法提高约21.9%。
For the finite buffer resources in satellite switching nodes of space information network(SIN),an optimal buffer allocation algorithm for satellite switching systems is proposed. Crossbar switching model is used,and on this basis,a virtual channel self-similar queuing model is established.Then the genetic algorithm is used to allocate the resources globally and optimally according to buffer overflow probability of every virtual channel.The simulation results show that,the new allocation algorithm has better delay performance than the uniform allocation algorithm and the greedy allocation algorithm.For the same average packet latency,compared with the uniform allocation algorithm,the new allocation algorithm can save 24.5% and26.3% of buffer consumption when the self-similarity of packet traffic is 0.6 and 0.8.And the efficiency of the new allocation algorithm is about 21.9% higher than that of the greedy allocation algorithm.
For the finite buffer resources in satellite switching nodes of space information network(SIN),an optimal buffer allocation algorithm for satellite switching systems is proposed. Crossbar switching model is used,and on this basis,a virtual channel self-similar queuing model is established.Then the genetic algorithm is used to allocate the resources globally and optimally according to buffer overflow probability of every virtual channel.The simulation results show that,the new allocation algorithm has better delay performance than the uniform allocation algorithm and the greedy allocation algorithm.For the same average packet latency,compared with the uniform allocation algorithm,the new allocation algorithm can save 24.5% and26.3% of buffer consumption when the self-similarity of packet traffic is 0.6 and 0.8.And the efficiency of the new allocation algorithm is about 21.9% higher than that of the greedy allocation algorithm.
引文
[1]李德仁,沈欣,龚建雅,等.论我国空间信息网络的构建[J].武汉大学学报(信息科学版),2015,40(6):711-715.
[2]王敬超,于全.基于分布式星群的空间信息网络体系架构与关键技术[J].中兴通讯技术,2016,22(4):9-18.
[3]徐烽,陈鹏.国外卫星移动通信新进展与发展趋势[J].电讯技术,2011,51(6):156-161.
[4] MAHI M J N,ATIK S T,IMRAN M M,et al.An adaptiverouting algorithm for on-chip 2D mesh network with anefficient buffer allocation scheme[C]//Proceedings of2018 International Conference on Computer,Communica-tion,Chemical,Materials and Electronic Engineering. Ra-jshahi:IEEE,2018:1-4.
[5] BAO D,LI X,XIN Y,et al.A virtual channel allocation al-gorithm for No C[C]//Proceedings of 2nd EAI 2017 In-ternational Conference on Machine Learning and Intelli-gent Communications.Weihai:EAI,2017:333-342.
[6] KHAN G N,CHUI S.Congestion aware routing for on-chipcommunication in No C systems[C]//Proceedings of the12th International Conference on Complex,Intelligent,andSoftware Intensive Systems.Torino:AIM,2017:547-556.
[7] LIU F,FU C H,YANG C.A study on performance com-parison of priority-based queueing scheme with twoqueueing buffer allocations in an UMTS core networkgateway[C]//Proceedings of 2008 International Confer-ence on Mobile Technology,Applications,and Systems.Yi-lan:ACM,2008:45-49.
[8] VORUGANTI S,MAHAMUNI A.System and method fordynamically allocating buffers based on priority levels:US8160085[P].2012-04-17.
[9] NI W,LIU Y,ZHANG S,et al.A buffer allocation algo-rithm for network-on-chip with self-similar traffic[C]//Proceedings of 2015 IEEE International Conference on In-formation Science and Technology.Shenzhen:IEEE,2015:230-233.
[10] TSYBAKOV B,GEORGANAS N D.Overflow and lossesin a network queue with a self-similar input[J].QueuingSystems,2000,35(1/4):201-235.
[2]王敬超,于全.基于分布式星群的空间信息网络体系架构与关键技术[J].中兴通讯技术,2016,22(4):9-18.
[3]徐烽,陈鹏.国外卫星移动通信新进展与发展趋势[J].电讯技术,2011,51(6):156-161.
[4] MAHI M J N,ATIK S T,IMRAN M M,et al.An adaptiverouting algorithm for on-chip 2D mesh network with anefficient buffer allocation scheme[C]//Proceedings of2018 International Conference on Computer,Communica-tion,Chemical,Materials and Electronic Engineering. Ra-jshahi:IEEE,2018:1-4.
[5] BAO D,LI X,XIN Y,et al.A virtual channel allocation al-gorithm for No C[C]//Proceedings of 2nd EAI 2017 In-ternational Conference on Machine Learning and Intelli-gent Communications.Weihai:EAI,2017:333-342.
[6] KHAN G N,CHUI S.Congestion aware routing for on-chipcommunication in No C systems[C]//Proceedings of the12th International Conference on Complex,Intelligent,andSoftware Intensive Systems.Torino:AIM,2017:547-556.
[7] LIU F,FU C H,YANG C.A study on performance com-parison of priority-based queueing scheme with twoqueueing buffer allocations in an UMTS core networkgateway[C]//Proceedings of 2008 International Confer-ence on Mobile Technology,Applications,and Systems.Yi-lan:ACM,2008:45-49.
[8] VORUGANTI S,MAHAMUNI A.System and method fordynamically allocating buffers based on priority levels:US8160085[P].2012-04-17.
[9] NI W,LIU Y,ZHANG S,et al.A buffer allocation algo-rithm for network-on-chip with self-similar traffic[C]//Proceedings of 2015 IEEE International Conference on In-formation Science and Technology.Shenzhen:IEEE,2015:230-233.
[10] TSYBAKOV B,GEORGANAS N D.Overflow and lossesin a network queue with a self-similar input[J].QueuingSystems,2000,35(1/4):201-235.