EPON上行链路动态带宽分配算法(DBA)分析

介绍和分析EPON的特点,侧重分析动态带宽分配算法的原理和相应类别的优缺点。主要分析3种DBA算法——间插轮询(IPACT)带宽分配算法算法、BGP算法和基于比例分配的支持服务等级区分带宽分配算法,分析其原理、上行通道的利用率及对不同业务等级之间公平性的影响,得出各自优缺点的结论。     

EPON的性能分析及优化

本文研究了以太无源光网络(EPON)的动态带宽算法,提出了基于轮询-呼叫方式的优化DBA算法.本文以合理地利用和分配带宽,确保系统得到QoS保证为目标,对上行通道剩余带宽的利用提出了相应的优化方法和调度机制,并进行了性能分析和比较.同时在研究了人们往往容易忽略的ONU的退出和启动基础上,进一步对EPON运行过程中ONU的接入和退出时带宽的利用进行了优化,并提出了解决方案.     

应用业务预测的动态带宽分配算法

当前以太无源光网络中动态带宽分配算法主要是基于轮询机制.其中限制带宽分配算法根据各个ONU发送的队列长度能够动态分配带宽,但是这种算法不能实时获得各个ONU的队列长度.因此,利用维纳滤波的方法对自相似业务流量进行预测,预测的结果应用于限制动态带宽分配,得出一种改进的动态带宽分配算法.这种改进算法通过预测ONU的队列长度,更确切地反映了其状态信息,有利于OLT进行准确地带宽分配.对这种算法进行了仿真分析,在业务负载介于0.1～0.5的情况下,平均包时延比较接近0.1ms,达到了改进目的.     

EPON中一种改进的DBA算法

动态带宽分配是EPON系统的关键技术之一。由于多个ONU共享同一上行信道,对上行数据流的控制显得尤为重要。EPON系统使用请求/应答机制对ONU进行带宽分配。在ETRI算法的基础上,提出了一种改进的算法,它支持队列优先级和ONU之间的公平性。通过仿真结果可以看出,改进的算法与ETRI相比,大大减少了平均队列长度。     

基于自相似流的无线Mesh网络信道分配算法

根据业务流的自相似性提出了一种新颖的信道分配算法。该算法首先根据流的自相似性,估计网络中的节点的业务量。然后,根据业务量多少确定占用信道的优先级。业务量大的节点可以优先占用信道。这样,解决了“热点”信道资源紧张的问题。同时,算法对“非热点”的吞吐量影响很小。理论分析证明,这种算法可以有效提高网络吞吐量。仿真结果表明,运用这种算法,网络的吞吐量比最小干扰固定信道分配算法和最小干扰混合信道分配算法可分别提高70％和40％。     

基于带宽借贷的EPON动态带宽分配算法

针对以往以太网无源光网络(EPON)动态带宽算法中存在的带宽分配不公平性问题,提出了基于带宽借贷机制的动态带宽算法LLR和LWR,其基本思想为当某光网络单元(ONU)有空余带宽时,光线路终端(OLT)将此ONU多余带宽借贷给重负载ONU,当债权ONU遇到带宽不足的情况时,OLT将会安排负债ONU偿还其未满足请求部分带宽.通过在正常使用和存在"贪婪"用户的两种情况下的仿真,分析了这两种算法的平均时延和丢包率特性.仿真结果表明这两种算法能为正常使用用户提供有保障的公平性服务,消除了以往算法在若干用户过量使用的情况下所有用户都受影响的问题,使动态带宽分配更加公平.     

EPON上行接入动态带宽分配算法的研究

以太无源光网络的上行方向需要一种多址接入技术来防止以太数据包的碰撞。时分多址技术应用广泛,而波分多址技术极具潜力,是发展的方向。基于时分多址技术,提出了一种改进的动态带宽分配算法,结合了轮询机制和申请的优先级。通过建模仿真,证明了该算法可以取得较小的平均包延时和较高的带宽利用率。     

一种基于节目流行度的自适应流调度方案

提出了一种基于节目流行度的自适应流调度方案。该方案根据节目点播请求的强度自适应地调整分配给节目的信道数和带宽,并给出了在信道数改变过程中如何实现信道平滑过渡的方法。该方案综合了周期广播算法和流合并算法的优点。实验表明,采用此方案,在点播请求很少时能获得同流分并算法（如补丁算法）一样的性能,在点播请求很多时又有和周期广播算法一样少的服务带宽要求。     

一种基于带缓存crossbar的优先级排队机制及其调度算法

基于带缓存crossbar的交换结构提出了一种支持多优先级的调度方案.该方案在带缓存crossbar交换结构的输入端口提出了一种基于端口与优先级标识的层次化排队机制,并结合该排队机制提出了一种加权双轮询(DWRR)的调度算法,在crossbar内部交叉节点处的缓存队列提出了一种易于实现的补偿优先轮询(CPRR)调度算法.这种DWRR-CPRR优先级调度方案不需要在crossbar交叉节点设置多个优先级缓存队列,避免了在带缓存crossbar内部进行复杂的调度,实现简单.时延特性与带宽分配特性的仿真结果表明,该方案可以获得接近于采用加权公平排队(WFQ)调度策略的OQ交换结构的性能.     

无线视频传输中自适应的前向纠错技术

提出了一种改进的无线网络视频传输自适应前向纠错(FEC)算法.该算法首先对丢包原因进行区分,根据不同的丢包原因采用不同的带宽估计方法,使得获得的网络拥塞导致丢包的概率更加准确;然后根据视频的不同部分在解码端重要性的不同,在视频数据和冗余数据之间优化分配所估计的带宽,使得在当前网络条件下视频的失真最小.仿真结果表明,这种算法能够更好地提高网络的吞吐量和视频的服务质量.     

以太无源光网络单级多功能ONU内调度算法

针对以太无源光网络(EPON),为各类业务提供服务质量支持,特别是保障时延敏感型业务的服务质量,提出一种调度算法,应用到EPON的光网络单元(ONU)中.我们设法让这个单级调度算法具备多级调度功能.通过理论分析和仿真测试,证明本算法具有灵活的优先级机制和公平调度功能.     

Dynamic resource allocation for supporting real-time multimedia applications in IEEE 802.15.3 WPANs

The IEEE 802.15.3 medium access control (MAC) protocol is an emerging standard for high-rate wireless personal area networks (WPANs), especially for supporting high-quality real-time multimedia applications. Despite defining quality of service (QoS) signalling mechanisms for interoperability between devices, IEEE 802.15.3 does not specify resource allocation algorithms that are left to manufacturers. To guarantee the QoS of real-time variable bit rate (VBR) videos and utilise the radio resource efficiently, the authors propose a dynamic resource allocation algorithm. The proposed bandwidth allocation algorithm is based on a novel traffic predictor. Recently, the variable step-size normalised least mean square (VSSNLMS) algorithm was employed for on-line traffic prediction of VBR videos. However, the performance of the VSSNLMS algorithm significantly degrades due to the abrupt traffic variation occurring at the scene boundary. To tackle this problem, the authors design a novel traffic predictor based on a simple scene detection algorithm and the VSSNLMS algorithm. Analyses using real-life MPEG video traces indicate that the proposed traffic predictor significantly outperforms the VSSNLMS algorithm with respect to the prediction error. The performance of the proposed bandwidth allocation algorithm is also investigated by comparing several existing algorithms. Simulation results demonstrate that the proposed bandwidth allocation algorithm surpasses other mechanisms in terms of channel utilisation, buffer usage and packet loss rate.     

Study of bandwidth effects in monochromator-based spectral responsivity measurements

The errors caused by monochromator bandwidth in spectral responsivity measurements with a monochromator-based apparatus are discussed. Bandwidth effects are not negligible in high-accuracy cryogenic radiometer-based calibrations. A simple numerical method is used to calculate bandwidth effects for different types of detectors, monochromator slit scattering functions, and monochromator output spectral distributions. The method uses low-order Lagrange polynomials fitted segmentwise to measured spectral responsivity and monochromator spectral distribution data in order to make the calculations. It is shown that the shape of the slit function has only a small influence on the bandwidth errors, whereas the output spectral distribution of the monochromator can strongly affect bandwidth errors. It is also shown that in most cases the magnitude of bandwidth effects will vary as the square of the bandwidth. Bandwidth error calculations are presented for various types of detectors (silicon, silicon trap, germanium, InGaAs), for a V(X) detector, and for a typical filter radiometer. A comparison is made between calculated and measured bandwidth effects to validate the method used. In general, calculations of bandwidth effects will be mostly useful for determining uncertainties associated with monochromator bandwidth in spectral responsivity measurements; however, in certain cases the calculations can be used to apply corrections for such effects.     

Adaptive wavelet-based multifractal model applied to the effective bandwidth estimation of network traffic flows

The authors investigate effective bandwidth estimation and Quality of Service (QoS) aware bandwidth provisioning for multifractal network traffic flows. They develop a novel adaptive wavelet-based multifractal model (AWMM) by using properties of the wavelet coefficients of multifractal cascade processes. The proposed AWMM has real-time updating capability and proves to be efficient in capturing multifractal network traffic characteristics. In addition, the authors derive an analytical expression for the effective bandwidth estimation of AWMM traffic flows, capable of being used to meet desired byte loss probabilities. Finally, they present an online effective bandwidth estimation algorithm that is incorporated into an adaptive bandwidth provisioning scheme and comparatively evaluated against some other bandwidth allocation methods.     

Asymmetric convolution algorithm for blocking probability calculation in full-availability group with bandwidth reservation

The paper proposes a new approximate calculation method of occupancy distribution and blocking probability in the full-availability group (FAG) with multi-rate traffic streams and bandwidth reservation. The algorithm involves calculation of system state probabilities by a convolution operation. In the paper, an effective method of calculating state probabilities in a reservation space of the system is derived. The analytical results of blocking probabilities in the FAG with bandwidth reservation, obtained on the basis of the proposed analytical method, are compared with the data obtained on the basis of other known analytical methods and with the simulation results. The accuracy of the proposed method is evaluated for different multi-rate Bernoulli (Erlang), Poisson (Engset), Pascal traffic streams. The method proposed is characterised by lower complexity than the convolution algorithm for the FAG with bandwidth reservation devised earlier.     

10G EPON的研究

本文探讨了万兆以太网无源光网络（10G EPON）的不同实现方法：上行分别基于时分多址或者波分多址,对上行采用波分多址（WDMA）接入的实现方法进行了更为详细的研究。分析了万兆以太网无源光网络存在的问题,最后讨论了其应用方向。     

Analysis of effect of load-based excess bandwidth reservation on performances of differentiated services in E-PON

Performances of differentiated services in the upstream transmission of Ethernet passive optical network (E-PON) are affected by bandwidth reservation algorithm of an optical network unit, since data packets can be transmitted to an optical line termination only during the reserved transmission window. The effects of time-slot reservation on differentiated services are analysed and an adaptive class-based excess bandwidth reservation algorithm, called service quality pre-engagement (SQP), which reserves a time-slot based on the backlog queue and traffic arrival pattern of class is proposed. SQP applies the prediction-based bandwidth reservation to selected forward reservation classes (FRCs) which are determined by the introduced dynamic FRC selection. The proposed short-term service work-based time-slot reservation and long-term adaptive FRC selection optimise time-slot reservation according to traffic load of class. Analytic and simulation results show that this approach improves the performance of differentiated services, especially in terms of system buffer size, light-load penalty and service fairness, while guaranteeing high link utilisation and throughput     

Mesh模式下基于最小干扰流量的带宽调度算法

研究了干扰链路流量大小对IEEE802．16Mesh网络传输性能的影响,提出了一种通过计算用户站（SS）和基站（BS）的通信路径干扰的总流量来寻找总干扰流量最小的路由算法,并结合路由树给出了一种冲突避免调度的方法。该算法既考虑了干扰链路的数量,又考虑了干扰链路的实际通信强度,在建立路由树的过程中能够更快地使系统达到更佳的负载均衡状态,最大强度地避免调度中可能出现的带宽冲突。由于干扰流量的影响可以得到有效的控制,系统的吞吐量可以明显提高。仿真结果验证了该算法的有效性。