TCP传输中往返时延偏移智能响应机制研究

为了提高高丢包率环境下的TCP传输性能,提出一种往返时延偏移智能响应机制。对往返时延偏移值进行标准化处理得到标准延迟因子,用这个因子对拥塞窗口增长和减小量进行修正,实现拥塞窗口增长速度随往返时延偏移自适应调整,能够区分随机丢包和网络拥塞。开发Linux内核模块实现了往返时延偏移智能响应机制,可快速部署到所有基于AIMD策略的拥塞控制机制。仿真结果表明,使用往返时延偏移智能响应机制,平均吞吐量超过cubic算法57%,能够有效提升高丢包率环境的带宽利用率。     

Comparative analysis of the performance of selective and group repeat transmission modes in a transport protocol

We propose a model of a virtual connection controlled by a transport protocol in the selective and group failure modes as a Markov chain with discrete time that accounts for the influence of protocol parameters of window size and timeout duration for waiting for acknowledgements, probabilities of distorting segments in individual links of the data transmission path on the throughput of a transport connection. We have analyzed how the throughput of the control procedure depends on protocol parameters, level of errors in communication channels, and round-trip delay. We have proposed a method for choosing protocol parameters.     

基于QoS的分布式认知无线电网络多信道MAC协议

在分布式认知无线电网络中,动态资源利用不足和中心控制单元的缺失使其MAC层协议的设计面临很多挑战。针对认知无线电网络的特点,提出了一种新的MAC协议,该协议通过在信道预约阶段优先考虑对时延要求较高的应用,来保证网络对这类应用的QoS;同时还解决了认知无线电网络中频谱的利用率低和隐藏终端两个热点问题。为分析研究该协议的性能,首先提出了一种新的分析模型。然后将这种新的MAC协议与两种典型的MAC协议进行了仿真对比,结果表明该协议提高了网络的吞吐量。最后通过数值分析和仿真证实了本协议设计简单高效,具有较高的频谱利用率,不但满足了时延敏感性应用的QoS需求,而且还能有效地解决多信道隐藏终端的问题。     

区分服务网络中双窗口TCP协议的性能分析与研究

提出一种分析模型，用于计算区分服务网络中双窗口TCP的性能，并清楚揭示了TCP流预订带宽、数据分组的丢失率以及TCP连接的往返时延(RTT)对确保服务性能的影响．仿真实验验证了该数学模型的有效性．同时仿真结果也表明。双窗口TCP机制能在较大程度上解决带宽保证和剩余带宽的公平分配问题，并且性能不受网络环境变化的影响．     

Performance analysis of an adaptive,energy-efficient MAC protocol for wireless sensor networks

We propose an adaptive and energy-efficient TDMA-based MAC protocol that significantly reduces energy consumption in the network, while efficiently handling network traffic load variations and optimizing channel utilization through a timeslot stealing mechanism and a timeslot reassignment procedure. We have analytically derived the average delay performance of our MAC protocol, with and without the timeslot stealing mechanism. Our delay model, validated via simulations, shows that the timeslot stealing mechanism can substantially improve the protocol throughput in scenarios with varying and asymmetric traffic patterns. Evaluation results show that the timeslot reassignment procedure is efficient in handling the longer timescale changes in the traffic load, while the timeslot stealing mechanism is better in handling the shorter timescale changes in the traffic patterns.     

On the properties of an additive increase rate accelerator

We propose AIRA, an Additive Increase Rate Accelerator. AIRA extends AIMD functionality towards adaptive increase rates, depending on the level of network contention and bandwidth availability. In this context, acceleration grows when resource availability is detected by goodput/throughput measurements and slows down when increased throughput does not translate into increased goodput as well. Thus, the gap between throughput and goodput determines the behavior of the rate accelerator.We study the properties of the extended model and propose, based on analysis and simulation, appropriate rate decrease and increase rules. Furthermore, we study conditional rules to guarantee operational success even in the presence of symptomatic, extra-ordinary events. We show that analytical rules can be derived for accelerating, either positively or negatively, the increase rate of AIMD in accordance with network dynamics. Indeed, we find that the “blind”, fixed Additive Increase rule can become an obstacle for the performance of TCP, especially when contention increases. Instead, sophisticated, contention-aware additive increase rates may preserve system stability and reduce retransmission effort, without reducing the goodput performance of TCP.     

An access protocol for efficiency optimization in WDM networks: A propagation delay and collisions avoidance analysis

The main investigation of this study is the performance optimization of a WDMA protocol suitable for a network architecture of passive star topology that uses the Multi-channel Control Architecture (MCA) and adopts asymmetric access rights over it. The access asymmetry defines that the MCA uses a separate control channel properly assigned for the free stations transmission, while the remaining control channels are used by the backlogged stations. The proposed asymmetric access protocol prevents from the data channels collisions, while it takes under consideration the loss due to the receiver conflicts. This is achieved by considering the propagation delay latency parameter as appropriate time interval in order to coordinate collisions-free data packets transmission. The performance parameters are derived through exhaustive analytical study based on a Markovian model of finite population, while the throughput optimization conditions are analytically investigated. Comparative results prove that the proposed asymmetric access protocol provides almost 11% throughput enhancement and significant delay deterioration as compared to a relative protocol with symmetric access rights over the MCA. The proposed protocol study is accomplished by examining its performance for various number of control channels and stations population.     

多跳移动Ad Hoc无线网络中UPMA协议的仿真

基于有效竞争预约接人、无冲突轮询传输的思想,该文改进了支持节点移动性和多跳网络结构的依据用户妥善安排的多址接人（UPMA）协议,然后利用网络仿真工具OPNET构建了UPMA协议的仿真框架,并仿真了由自组织分群算法所获得群的群内无线传输情况。该仿真框架为UPMA协议详细配置了物理层和数据链路层模型,可以设置不同的信道特性以及应用场景。在此基础上,该文分析了UPMA协议的信道吞吐量、平均消息时延和平均消息丢弃率,并将其与带冲突避免的载波侦听多址接人协议（CSMA／CA）、轮询协议进行了性能比较。仿真结果表明,UPMA协议可以提供较高的吞吐量、较低的平均消息时延和较小的平均消息丢弃率。     

基于改进BBR的数据报拥塞控制协议

数据报拥塞控制协议(Datagram Congestion Control Protocol,DCCP)是提供拥塞控制和不可靠传输特点的实时多媒体基础协议,DCCP中的CCID2算法仍然采用AIMD的控制机制,这种传统的Loss-Base拥塞控制模型已经不适用于目前高BDP的网络环境,容易引起缓冲区膨胀现象,导致网络延迟增加和抖动等问题.与Loss-Base的算法相比,BBR算法可以有效地控制网络延时,最大限度避免网络排队的情况,在丢包率较高的情况下仍可以保持一定的带宽利用率和较低的链路延时,因此适合于DCCP实时流媒体的应用的协议.本文在DCCP中引入了BBR算法并做相应的改进,增加了丢包率检测模型,使用延时与带宽积模型的拥塞控制算法对上述问题进行改进.通过模拟实验证明,本方法在高负载情况下连接的平均延迟相比CCID2降低了20%,在丢包率较高的环境下也能保持良好的吞吐量.     

Enhanced frameless slotted ALOHA protocol with Markov chains analysis

In this paper, we propose a novel approach to enhance the performance of frameless slotted ALOHA (SA) protocol. We employ signature codes to help the receiver identify the packets contained in collisions, and use successive interference cancellation (SIC) for packet recovery. We model the proposed scheme as a two-state Markov model represented by a uni-partite graph. We evaluate the throughput, expected delay and average memory size of the proposed scheme, and optimize the proposed scheme to maximize the throughput. We show that the theoretical analysis matches well with simulation results. The throughput and expected delay of the proposed protocol outperform the conventional slotted ALOHA protocol significantly.     

Scalability of routing methods in ad hoc networks

In an ad hoc network each host (node) participates in routing packets. Ad hoc networks based on 802.11 WLAN technology have been the focus of several prior studies. These investigations were mainly based on simulations of scenarios involving up to 100 nodes (usually 50 nodes) and relaxed (too unrealistic) data traffic conditions. Many routing protocols in such setting offer the same performance, and many potential problems stay undetected. At the same time, an ad hoc network may not want (or be able) to limit the number of hosts involved in the network. As more nodes join an ad hoc network or the data traffic grows, the potential for collisions and contention increases, and protocols face the challenging task to route data packets without creating high administrative load. The investigation of protocol behavior in large scenarios exposes many hidden problems. The understanding of these problems helps not only in improving protocol scalability to large scenarios but also in increasing the throughput and other QoS metrics in small ones. This paper studies on the example of AODV and DSR protocols the influence of the network size (up to 550 nodes), nodes mobility, nodes density, suggested data traffic on protocols performance. In this paper we identify and analyze the reasons for poor absolute performance that both protocols demonstrate in the majority of studied scenarios. We also propose and evaluate restructured protocol stack that helps to improve the performance and scalability of any routing protocol in wireless ad hoc networks.     

Improving the efficiency of multipath traffic via opportunistic traffic scheduling

Multipath routing, as defined by OSPF extensions and other protocols, enables a network’s traffic to be split among two, or more, possibly disjoint paths. Advantages of multipath vs. unipath routing include load balancing, reduced latency, and improved throughput. However, once the control plane establishes multiple routes, a policy is needed for efficiently splitting traffic among the selected paths. In this paper, we introduce opportunistic multipath scheduling (OMS), a technique for exploiting short-term variations in path quality to minimize delay, while simultaneously ensuring that the splitting rules dictated by the routing protocol are satisfied. We develop a performance model of OMS and derive an asymptotic lower bound on the performance of OMS as a function of path conditions (mean, variance, and Hurst parameter) for self-similar traffic. Finally, we use an extensive simulation-based performance study to evaluate the accuracy of the analytical model, explore the impact of OMS on TCP throughput and performance of real-time traffic, and study the impact of factors such as delayed measurements.     

一种支持实时性业务的多信道MAC协议

为满足军事航空通信中实时性业务对时延的要求,提出一种基于Turbo编码的多信道MAC协议——Turbo_MAC。Turbo编码技术和多信道机制的使用,使得协议在不采用反馈确认机制的情况下仍能确保分组的高可靠性和系统的高吞吐量,同时达到降低分组端到端时延的目的。运用排队模型、组合理论和离散Laplace变换得到分组成功传输概率、系统吞吐量和分组端到端时延均值性能指标。仿真结果表明,理论分析与仿真结果一致,且Turbo_MAC协议的时延性能优于TDMA和IEEE 802.11b多信道MAC协议。     

Mitigating collisions through power-hopping to improve 802.11 performance

In this article, we introduce a power-hopping technique (PH-MAC) that, by alternating between different transmission power levels, aims to deliberately cause packet capture and thereby reduce the impact of collisions in 802.11 WLANs. We first devise an analytical model of the 802.11 protocol with heterogeneous capture probabilities, and show that, depending on the network load, the capture effect can enhance the throughput performance of all nodes. We base the design of PH-MAC on the findings following from this analysis and demonstrate that important performance improvements can be achieved by exploiting the interactions between the MAC and PHY layers to mitigate collisions. Finally, to understand the feasibility of this technique in practical deployments, we present a prototype implementation of PH-MAC which relies on commodity hardware and open-source drivers. We evaluate the performance of this implementation in an indoor testbed under different network conditions in terms of link qualities, network loads and traffic types. The experimental results obtained show that our scheme can provide significant gains over the default 802.11 mechanism in terms of throughput, fairness and delay.     

智能交通系统中CMIMO-SM-ARQ协议性能建模分析

为提高智能交通系统(ITS)中无线传感器网络(WSN)的吞吐量和能效性能,提出了一种具备空间调制的协同多输入多输出自动重传请求(CMIMO-SM-ARQ)协议.考虑了双发送节点和双接收节点的四节点WSN,其接收节点可以互为中继辅助发送端传输数据,以降低系统中断概率.建立了具有3L+1个状态的CMIMO-SM-ARQ协议离散时间马尔可夫链(DTMC)模型,并给出状态空间.根据WSN的中断概率和一步状态转移概率矩阵推导CMIMO-SM-ARQ协议的吞吐量、能效和时延性能解析表达式.数值模拟结果表明,所提协议的中断概率和能效性能优于单输入单输出自动重传请求(SISO-ARQ)协议,在中远距离(大于67 m)传输信息时,CMIMO-SM-ARQ具有较高的吞吐量和较低的时延.     

Design and performance of speculative flow control for high-radix datacenter interconnect switches

High-radix switches are desirable building blocks for large computer interconnection networks, because they are more suitable to convert chip I/O bandwidth into low latency and low cost than low-radix switches [J. Kim, W.J. Dally, B. Towles, A.K. Gupta, Microarchitecture of a high-radix router, in: Proc. ISCA 2005, Madison, WI, 2005]. Unfortunately, most existing switch architectures do not scale well to a large number of ports, for example, the complexity of the buffered crossbar architecture scales quadratically with the number of ports. Compounded with support for long round-trip times and many virtual channels, the overall buffer requirements limit the feasibility of such switches to modest port counts. Compromising on the buffer sizing leads to a drastic increase in latency and reduction in throughput, as long as traditional credit flow control is employed at the link level. We propose a novel link-level flow control protocol that enables high-performance scalable switches that are based on the increasingly popular buffered crossbar architecture, to scale to higher port counts without sacrificing performance. By combining credited and speculative transmission, this scheme achieves reliable delivery, low latency, and high throughput, even with crosspoint buffers that are significantly smaller than the round-trip time. The proposed scheme substantially reduces message latency and improves throughput of partially buffered crossbar switches loaded with synthetic uniform and non-uniform bursty traffic. Moreover, simulations replaying traces of several typical MPI applications demonstrate communication speedup factors of 2 to 10 times.     

CSMA/CA在CSMA/TDMA混合网络中的性能分析

CSMA/CA机制与TDMA在同频段共存时会存在相互干扰,为评估混合网络中CSMA/CA的性能表现,将TDMA视为周期性的干扰提出改进的二维Markov分析模型,能够计算不同条件下的CSMA/CA饱和吞吐量,并通过仿真实验证明了模型的有效性。同时讨论了TDMA时隙分配方式对系统性能产生的影响,仿真结果表明时隙均匀分配方式能保证较低的时延和时延抖动,而连续分配方式能使CSMA/CA获得较高的吞吐量。提出将基于最低信道需求时间的阈值计算方法与吞吐量分析模型相结合,用于在不同应用场景下进行时隙分配方式的选择。     

A case for exploiting self-similarity of network traffic in TCP congestion control

Analytical and empirical studies have shown that self-similar traffic can have detrimental impact on network performance including amplified queuing delay and packet loss ratio. On the flip side, the ubiquity of scale-invariant burstiness observed across diverse networking contexts can be exploited to better design resource control algorithms. In this paper, we explore the issue of exploiting the self-similar characteristics of network traffic in TCP congestion control. We show that the correlation structure present in long-range dependent traffic can be detected on-line and used to predict the future traffic. We then devise a novel scheme, called TCP with traffic prediction (TCP-TP), that exploits the prediction result to infer, in the context of AIMD steady-state dynamics, the optimal operational point at which a TCP connection should operate. Through analytical reasoning, we show that the impact of prediction errors on fairness is minimal. We also conduct ns-2 simulation and FreeBSD 4.1-based implementation studies to validate the design and to demonstrate the performance improvement in terms of packet loss ratio and throughput attained by connections.     

Markov Chain-based performance analysis of multihop IEEE 802.15.4 wireless networks

In this paper, we propose a Markov chain-based analytical framework for modeling the behavior of the medium access control (MAC) protocol in IEEE 802.15.4 wireless networks. Two scenarios are of interest. First, we consider networks where the (sensor) nodes communicate directly to the network coordinator (the final sink). Then, we consider cluster-tree (CT) scenarios where the sources communicate to the coordinator through a series of intermediate relay, which forward the received packets and do not generate traffic on their own. In both scenarios, no acknowledgment messages are used to confirm successful data packet deliveries and communications are beaconed (i.e., they rely on synchronization packets denoted as “beacons”). In all cases, our focus is on networks where the sources and the relays have finite queues (denoted as buffers) to store data packets. The network performance is evaluated in terms of aggregate network throughput and packet delivery delay. The performance predicted by the proposed analytical framework is in very good agreement with realistic ns-2 simulation results.     

面向股票交易分析场景的流式大数据系统测试框架

分布式集群环境使得数据实时计算更为复杂,流式大数据处理系统的正确性难以保障.现有的大数据基准测试框架可以测试流式大数据处理系统的性能表现,但是普遍存在应用场景设计简单、评价指标不充分等不足.针对这一挑战,本文构造了一个面向股票交易场景的流式大数据基准测试框架,通过生成股票高频交易数据,测试系统在高流速场景下的延迟、吞吐量、GC时间、CPU资源等的性能表现.同时,通过横向测试验证流式大数据系统的扩展性.本文以Apache Spark Streaming为待测系统进行测试,实验结果表明,高流速场景下出现延迟增加、GC时间提高等性能下降问题,原因是系统输入速率的提高及并行度的增加.