多路径传输管理技术的研究

发布时间：2018-08-13 18:14

【摘要】：近年来,智能终端逐渐取代桌面个人电脑成为未来网络中最重要的接入点；同时宽带接入技术迅猛发展,泛在网络环境日趋成熟,任一智能终端均可同时使用多个网络。因此,多路径传输作为一种有效提高网络资源利用率的新技术,得到越来越多研究人员的关注。多路径传输技术在一定程度上有效提高传输的可靠性,增强网络的抗毁性,实现吞吐量性能的改进,但同时也加大了资源管理的难度。特别是在智能终端的物理资源受限的条件下,多路径传输的网络资源利用率降低,性能下降。我们的工作基于智能终端访问网络的应用场景,针对多路径传输管理中存在的问题展开,引入最优化的理论方法,以解决多路径传输中网络资源利用率与接收端物理资源受限的矛盾为目的,提出多路径传输中流量与拥塞控制、路径管理、调度管理以及切换管理的方法。我们的具体工作如下： (1)当接收端缓冲区规模小于带宽时延积时,存在带宽利用率较低的问题。原因是发送端对接收端接收窗口的估计变化滞后于接收端窗口的实际变化。而这一点,在很多情况下被人们所忽略。针对以上问题,我们进行了以下工作：首先,我们通过分析可靠传输的行为,提出基于可靠传输的吞吐量分析模型。这里将路径的拥塞窗口和慢启动阈值的组合定义为二元状态空间,分析有限状态空间内的转移概率；再次,我们提出了基于虚拟接收窗口的流量控制与拥塞控制方案。最后,通过理论分析和仿真,验证了我们所提出的方案。 (2)在多路径传输中,存在链接的总体吞吐量随着路径数量的增加而下降的现象。原因是在接收端受限的情况下,路径的差异会导致大量失序分组到达接收端,并消耗接收端的缓冲区,引起接收端阻塞。针对以上问题,我们进行了以下工作：首先我们将基于可靠传输的吞吐量模型扩展到多条路径的场景,并推导稳态下的多路径的吞吐量以及接收端重排序的开销；再次,将路径管理问题抽象为一个以多路径吞吐量为价值和以接收端重排序开销为重量的0-1背包问题,同时给出一个迭代算法求出最优解；最后,通过仿真验证了我们所提出的方案。 (3)简单的轮询调度算法,忽略了路径差异和丢包的存在,导致接收端出现阻塞。针对以上问题,我们进行以下工作：首先,建立分组到达接收端时间的预测模型,并根据预测智能调度分组,尽量实现分组的有序到达；再次,研究两种重传机制,分析丢包引起两种重传发生的概率、重传所需的时间,以及引起的接收端开销。为了增大分组到达的概率,减少接收端开销,引入前向纠错(Forward Error Correction,FEC)冗余机制。但FEC同时也带来一些额外的冗余开销,这里进一步将基于FEC的冗余调度管理抽象为一个受限优化问题,通过求解此问题的最优解进行分组调度；最后通过仿真验证冗余调度管理机制。 (4)传统的切换性能评价仅考虑了切换行为对参与切换路径的影响,而并发多路径传输中,除参与切换的路径外,仍旧有其它路径在切换期间被使用。多路径传输中的切换不仅对参与切换路径造成影响,同时也影响了未参与切换的路径以及链接的整体性能。针对多路径传输切换管理中出现的新问题,本文进行如下工作：首先通过分析路径切换的过程,综合考虑切换的时延大小、切换的信令开销、以及切换的效用,建立多路径传输的切换判决模型；再次我们引入效用最大化原则,提出了以吞吐量效用最大化为目标的切换判决管理；最后通过仿真证明我们所提出的切换方案。
[Abstract]:In recent years, smart terminals have gradually replaced desktop PCs as the most important access points in future networks. At the same time, with the rapid development of broadband access technology, ubiquitous network environment is becoming more and more mature. Any smart terminal can use multiple networks at the same time. To a certain extent, multipath transmission technology can effectively improve the reliability of transmission, enhance network invulnerability and achieve throughput performance improvement, but also increase the difficulty of resource management. Especially, under the condition of limited physical resources of intelligent terminals, multipath transmission network resources benefit. Our work is based on the application scenario of intelligent terminal access network, aiming at the problems existing in multipath transmission management, we introduce optimization theory to solve the contradiction between network resource utilization and physical resource constraints in multipath transmission, and propose traffic in multipath transmission. And congestion control, path management, scheduling management and handover management.
Our specific work is as follows:
(1) When the buffer size of the receiver is smaller than the bandwidth delay product, there is a problem of low bandwidth utilization. The reason is that the estimated change of the receiving window at the sender lags behind the actual change of the receiving window. This is ignored in many cases. To solve the above problems, we have done the following work: First, I By analyzing the behavior of reliable transmission, we propose a throughput analysis model based on reliable transmission. In this paper, the combination of congestion window and slow start threshold is defined as a binary state space, and the transition probability in finite state space is analyzed. Thirdly, we propose a traffic control and congestion control scheme based on virtual receiving window. Finally, theoretical analysis and simulation verify the proposed scheme.
(2) In multipath transmission, there is a phenomenon that the overall throughput of links decreases with the increase of the number of paths. The reason is that the difference of paths will lead to a large number of out-of-order packets arriving at the receiving end when the number of paths is limited at the receiving end, and consume the buffer of the receiving end, resulting in the blocking of the receiving end. Firstly, we extend the throughput model based on reliable transmission to the scenario of multiple paths, and derive the throughput of multipaths in steady state and the cost of rescheduling at the receiver. Thirdly, we abstract the path management problem into a 0-1 knapsack problem with multipath throughput value and rescheduling overhead at the receiver. An iterative algorithm is proposed to get the optimal solution. Finally, the proposed scheme is verified by simulation.
(3) Simple polling scheduling algorithm ignores the existence of path differences and packet dropouts, resulting in blocking at the receiver. To solve the above problems, we do the following work: First, establish a prediction model of packet arrival time at the receiver, and according to the prediction of intelligent scheduling packet, as far as possible to achieve orderly packet arrival; thirdly, study two kinds of retransmission machine. In order to increase the probability of packet arrival and reduce the receiver overhead, a forward Error Correction (FEC) redundancy mechanism is introduced. But FEC also brings some additional redundancy overhead, which will be further based on FEC. Redundant scheduling management is abstracted as a constrained optimization problem, and packet scheduling is performed by solving the optimal solution of the problem. Finally, the redundant scheduling management mechanism is verified by simulation.
(4) Traditional handoff performance evaluation only considers the effect of handoff behavior on the participating handoff paths, while in concurrent multipath transmission, there are other paths used during handoff except the paths participating in handoff. In view of the new problems in the handover management of multipath transmission, the following work is done in this paper: Firstly, the handover decision model of multipath transmission is established by analyzing the process of path switching, considering the size of the handover delay, the signaling overhead and the utility of the handover. Based on the maximization principle, a handoff decision management scheme with the goal of maximizing throughput utility is proposed. Finally, the proposed handoff scheme is proved by simulation.
【学位授予单位】：北京邮电大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TN929.5

【参考文献】