Mobile Ad Hoc Network is a self-configuring, autonomous, infrastructure less network of movable nodes which will be automatically connected by wireless connection with no access point. The nodes are free to move anywh...Mobile Ad Hoc Network is a self-configuring, autonomous, infrastructure less network of movable nodes which will be automatically connected by wireless connection with no access point. The nodes are free to move anywhere in the set of connections and the topology of MANET remains unpredictable. The major complexity with the available two hop relay protocols in MANET environment is to achieve the optimized throughput with reduced packet delivery delay. A generalized collection based two-hop relay and redundancy of the packet is used to attain the reduced delay. The complex process of packet delivery in MANET utilizes Routine Response Control and Modified Markov chain model. Tuning carefully the parameters, the transmission range, the packet redundancy and the group size help to achieve the optimized throughput with the reduced packet delivery delay. Transmission power of a node is controlled which helps in achieving the optimized throughput and reduced delay.展开更多
跨数据中心网络是处于不同地区的数据中心网络(Data Center Networks,DCNs)通过广域网(Wide-Area Network,WAN)连接组成的网络,分布式应用通常基于该网络为用户提供高质量的服务。DCNs和WAN的缓冲区大小、往返时延存在显著差异,这导致...跨数据中心网络是处于不同地区的数据中心网络(Data Center Networks,DCNs)通过广域网(Wide-Area Network,WAN)连接组成的网络,分布式应用通常基于该网络为用户提供高质量的服务。DCNs和WAN的缓冲区大小、往返时延存在显著差异,这导致现有的Cubic拥塞控制算法在跨数据中心网络场景下出现降速不准确、丢包率过高以及与其他拥塞控制算法兼容性差等问题。针对以上挑战,提出了一种通过匹配不同发送速率模式的改进Cubic算法Cubic+。具体地,Cubic+整合了网络中的时延、ECN(Explicit Congestion Notification)和丢包信号。当拥塞发生在浅缓冲交换机时,Cubic+会周期性地排空队列;当拥塞发生在深缓冲路由器时,Cubic+会快速减少堆积的数据包。基于大规模NS3仿真实验结果表明,在跨数据中心网络流量模型下,Cubic+与现有流行算法相比,平均流完成时间最多减少了20.77%,第99百分位流完成时间最多减少了15.88%,为跨数据中心网络提供了一种高吞吐的拥塞控制算法。展开更多
A dynamic network Qo S control mechanism was proposed based on traffic prediction. It first predicts network traffic flow and then dynamically distributes network resources, which makes full use of network flow self-s...A dynamic network Qo S control mechanism was proposed based on traffic prediction. It first predicts network traffic flow and then dynamically distributes network resources, which makes full use of network flow self-similarity and chaos. So it can meet changing network needs very well. The simulation results show that the dynamic Qo S control mechanism based on prediction has better network performance than that based on measurement.展开更多
文摘Mobile Ad Hoc Network is a self-configuring, autonomous, infrastructure less network of movable nodes which will be automatically connected by wireless connection with no access point. The nodes are free to move anywhere in the set of connections and the topology of MANET remains unpredictable. The major complexity with the available two hop relay protocols in MANET environment is to achieve the optimized throughput with reduced packet delivery delay. A generalized collection based two-hop relay and redundancy of the packet is used to attain the reduced delay. The complex process of packet delivery in MANET utilizes Routine Response Control and Modified Markov chain model. Tuning carefully the parameters, the transmission range, the packet redundancy and the group size help to achieve the optimized throughput with the reduced packet delivery delay. Transmission power of a node is controlled which helps in achieving the optimized throughput and reduced delay.
文摘作为网络传输控制机制的核心,拥塞控制关注如何在异构网络环境中最优化特定传输性能目标。已有拥塞控制机制忽略了不同应用的性能偏好在吞吐量-时延两个维度上的帕累托最优前沿(Pareto optimal frontier,POF)分布,难以满足差异化应用的性能需求。针对上述问题,本文提出了一种面向应用性能偏好的帕累托最优拥塞控制机制pBBR(ParetooptimalBBR),结合离线网络场景学习和在线控制参数优化的思想,最大程度满足应用的差异化性能偏好。实验结果表明,pBBR能够在一个采集-识别周期内判断出网络场景的切换,从而快速选择当前网络场景的最优控制参数。每个网络场景下,pBBR都能够最大化满足不同的应用性能偏好:针对吞吐量敏感业务,pBBR可以达到Cubic(吞吐优先)的97%,且时延只有Cubic的52%;针对时延敏感业务,pBBR的时延可以达到Sprout(时延优先)的95%,同时吞吐量损失只有1%。此外,多参数优化可进一步提升pBBR性能,例如在高铁长期演进技术(long term evolution,LTE)通信场景下,单参数pBBR的吞吐量、时延分别是Cubic的94%和99%,而三参数pBBR则分别提升到Cubic的101%和93%(优于Cubic)。
基金Funded by the National Natural Science Foundation of China(No.41301084)the Scientific Research Project of Hunan Province Education Department,China(No.13C713)the Natural Science Foundation of Hunan Province,China(No.13JJ6075)
文摘A dynamic network Qo S control mechanism was proposed based on traffic prediction. It first predicts network traffic flow and then dynamically distributes network resources, which makes full use of network flow self-similarity and chaos. So it can meet changing network needs very well. The simulation results show that the dynamic Qo S control mechanism based on prediction has better network performance than that based on measurement.
