Full Link Utilization Research Articles

A Receiver-Driven Transport Protocol With High Link Utilization Using Anti-ECN Marking in Data Center Networks

Existing reactive or proactive congestion control protocols are hard to simultaneously achieve ultra-low latency and high link utilization across all workloads ranging from delay-sensitive flows to bandwidth-hungry ones in datacenter networks. We present an Anti-ECN (Explicit Congestion Notification) Marking Receiver-driven Transport protocol called AMRT, which achieves both near-zero queueing delay and full link utilization by reasonably increasing sending rate in the case of under-utilization. Specifically, switches mark the ECN bit of data packets once detecting spare bandwidth. When receiving the anti-ECN marked packet, the receiver generates the corresponding marked grant to trigger more data packets. The testbed and simulation experiments show that AMRT effectively reduces the average flow completion time (AFCT) by up to 42% and improves the link utilization by up to 38% over the state-of-the-art receiver-driven transmission schemes.

Does fragmentation avoidance improve the performance of dynamic spectrum allocation in elastic optical networks?

Most spectrum allocation algorithms in elastic optical networks apply a greedy approach: A new connection is allocated as long as there are enough spectrum slots to accommodate it. Recently, a different approach was proposed. Named Deadlock–Avoidance (DA), it only establishes a new connection if the portion of spectrum left after allocating it is zero (full-link utilization) or is big enough to accommodate future requests. Otherwise, the connection request is blocked as a way to avoid fragmentation. The performance of DA has been evaluated in a single-link scenario, where its performance is not affected by the slot continuity constraint. In this paper, we evaluate for the first time the blocking performance and fragmentation level of DA in a fully dynamic network scenario with different bitrates and number of slots for a single link, a 4-node bus and a mesh topology. The performance was evaluated by simulation, and a lower bound was also derived using a continuous Markov chain model. Results are obtained for DA and three greedy algorithms: First Fit, Exact Fit and First–Last Fit. Results show that DA significantly decreases fragmentation, and thus, it exhibits a much lower blocking due to fragmentation than the greedy algorithms. However, this decrease is compensated by a new type of blocking due to the selective acceptance of connections. As a result, the extra computational complexity of DA does not compensate a gain in performance.

Congestion Control for Web Real-Time Communication

Applications requiring real-time communication (RTC) between Internet peers are ever increasing. RTC requires not only congestion control but also minimization of queuing delays to provide interactivity. It is known that the well-established transmission control protocol congestion control is not suitable for RTC due to its retransmissions and in-order delivery mechanisms, which induce significant latency. In this paper, we propose a novel congestion control algorithm for RTC, which is based on the main idea of estimating—using a Kalman Filter—the end-to-end one-way delay variation which is experienced by packets traveling from a sender to a destination. This estimate is compared with a dynamic threshold and drives the dynamics of a controller located at the receiver, which aims at maintaining queuing delays low, while a loss-based controller located at the sender acts when losses are detected. The proposed congestion control algorithm has been adopted by Google Chrome. Extensive experimental evaluations have shown that the algorithm contains queuing delays while providing intra and inter protocol fairness along with full link utilization.

Enhanced TCP Congestion Control with Higher Utilization in Under-Buffered Links

TCP Reno is not fully utilized in under-buffered links. We propose a new TCP congestion control algorithm that can utilize the link almost up to 100% except the first congestion avoidance cycle. Our scheme estimates the minimum congestion window size for full link utilization in every congestion avoidance cycle and sends extra packets without touching TCP Reno congestion control. It has the same RTT fairness and the same saw-tooth wave as TCP Reno does. Our scheme does not affect competing TCP Reno flows since it uses only unused link capacity. We provide a simple mathematical modeling as well as ns-2 simulation results which show that the link utilization is improved by up to 19.88% for k=1/8 against TCP Reno when the buffer is k times the optimal buffer size. We claim that our scheme is useful for transmitting large amount of data in under-buffered links.

ARROW-WTCP: A fast transport protocol based on explicit congestion notification over wired/wireless networks

An explicit congestion notification (ECN)-based distributed transport protocol, ARROW-WTCP (AcceleRate tRansmission towards Optimal Window size TCP for Wireless network), was proposed. The ARROW-WTCP enables feasible deployment of ARROW-TCP from wired to wireless networks by providing a joint design of source and router algorithms. The protocol obtains the actual capacity of the wireless channel by calculating the queue variation in base station (BS) and adjusts the congestion window by using the feedback from its bottleneck link. The simulation results show that the ARROW-WTCP achieves strong stability, max-min fairness in dynamic networks, fast convergence to efficiency without introducing much excess traffic, and almost full link utilization in the steady state. It outperforms the XCP-B (eXplicit Control Protocol Blind), the wireless version of XCP, in terms of stability, fairness, convergence and utilization in wireless networks.

Normalized Queueing Delay: Congestion Control Jointly Utilizing Delay and Marking

Depending upon the type of feedback that is primarily used as a congestion measure, congestion control methods can be generally classified into two categories: marking/loss-based or delay-based. While both marking and queueing delay provide information about the congestion state of a network, they have been largely treated with separate control strategies. In this paper, we propose the notion of the normalized queueing delay, which serves as a congestion measure by combining both delay and marking information. Utilizing normalized queueing delay (NQD), we propose an approach to congestion control that allows a source to scale its rate dynamically to prevailing network conditions through the use of a time-variant set-point. In ns-2 simulation studies, an NQD-enabled FAST TCP demonstrates a significant link utilization improvement over FAST TCP under certain conditions. In addition, we propose another NQD-based controller D + M TCP (Delay+Marking TCP) that achieves quick convergence to fair and stable rates with nearly full link utilization. Therefore, NQD is a suitable candidate as a congestion measure for practical congestion control.

Bandwidth Allocation for QoS Using Adaptive Modulation and Coding in IEEE 802.16 Networks

In recent years, there have been many studies on integrating a number of heterogeneous wireless networks into one network by establishing standards like IEEE 802.16. For this purpose, the base station (BS) should allocate the appropriate bandwidth to each connection with a network scheduler. In wireless networks, the signal to noise ratio (SNR) changes with time due to many factors such as fading. Hence, we estimate the SNR based on the error rate reflecting wireless network condition. Using the estimated SNR, we propose a new time slot allocation algorithm so that the proposed algorithm guarantees the delay requirement and full link utilization.

Study on TCP Flow-Competing Congestion and Buffer Requirement of the Congested Links

This paper first presents an analysis model named flow-competing congestion model(FCCM)for this type of congestion.Based on FCCM,the paper derives the distribution of competing flows at the congested link, and analyzes the conditions under which the flow-competing congestion would happen.This paper also explores how much buffers a congested link requires to keep full link utilization when the flow-competing congestion occurs. This paper proves that when sizing buffers for a congested internet link with the aim of keeping full link utilization, the buffer requirement of the flow-competing congestion is not bigger than the minimum buffer requirement of the famous BSCL(buffer sizing for congested internet links)scheme.

On applying fair queuing discipline to schedule requests at access gateway for downlink differential QoS

Scheduling packets is a usual solution to allocate the bandwidth on a bottleneck link. However, this solution cannot be used to manage the downlink bandwidth at the user-side access gateway, since the traffic is queued at the ISP-side gateway but not the user-side gateway. An idea is scheduling the requests at the user-side gateway to control the amount of the responses queued in the ISP-side gateway. This work first investigates the possibility of applying the class-based fair queuing discipline, which was widely and maturely used in scheduling packets, to schedule requests. However, we found that simply applying this discipline to schedule requests would encounter the timing and ordering problems at releasing requests and may not satisfy high-class users. Thus, we propose a minimum-service first request scheduling (MSF-RS) scheme. MSF-RS always selects the next request from the class receiving the minimum service to provide user-based weighted fairness, which ensures more bandwidth for high-class users. Next, MSF-RS uses a window-based rate control on releasing requests to maintain full link utilization and reduce the user-perceived latency. The results of analysis, simulation and field trial demonstrate that MSF-RS provides fairness while reducing 23-30% of user-perceived latency on average. Besides, a MSF-RS gateway can save 25% of CPU loading.

Achieving 100% Throughput in TCP/AQM Under Aggressive Packet Marking With Small Buffer

We consider a TCP/AQM system with large link capacity (NC) shared by many flows. The traditional rule-of-thumb suggests that the buffer size be chosen in proportion to the number of flows (N) for full link utilization, while recent research outcomes show that O(radic(N)) buffer sizing is sufficient for high utilization and O(1) buffer sizing makes the system stable at the cost of reduced link utilization. In this paper, we consider a system where the active queue management (AQM) is scaled as O(N <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">alpha</sup> ) with a buffer of size O(N <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">beta</sup> ) (0 < alpha < beta < 0.5). By capturing randomness both in packet arrivals and in packet markings, we develop a doubly-stochastic model for a TCP/AQM system with many flows. We prove that, under such a scale, the system always performs well in the sense that the link utilization goes to 100% and the loss ratio decreases to zero as the system size JV increases. Our results assert that the system enjoys benefit of largeness with no tradeoff between full link utilization, zero packet loss, and small buffer size, at least asymptotically. This is in stark contrast to existing results showing that there always exists a tradeoff between full link utilization and the required buffer size. Extensive ns-2 simulation results under various configurations also confirm our theoretical findings. Our study illustrates that blind application of fluid modeling may result in strange results and exemplifies the importance of choosing a right modeling approach for different scaling regimes.

Performance analysis of TCP/AQM with generalized AIMD under intermediate buffer sizes

For TCP/AQM systems, the issue of buffer sizing has recently received much attention. The classical rule-of-thumb suggests O(N) buffer size to ensure full link utilization when N TCP flows share a bottleneck link of capacity O(N), while recent empirical study shows the buffer of size O(N) is enough to yield high utilization (say, 95%) for large N. However, these results are all limited to the drop-tail scheme and there has been no systematic modeling framework for any buffer sizing between O(N) and O(N). In this paper, we study the limiting behavior of a TCP/AQM system for an intermediate buffer sizing of O(N^@c) (0.5=<@c<1). We develop a stochastic model in a discrete-time setting to characterize the system dynamics and then show that we can have 100% link utilization and zero packet loss probability for a large number of flows when the buffer size is chosen anywhere between O(N) and O(N). Our model is general enough to cover any queue-based AQM scheme with ECN marking (including the drop-tail) and various generalized AIMD (additive-increase-multiplicative-decrease) algorithms for each TCP flow. We also provide arguments showing that the discrete-time based modeling can effectively capture all the essential system dynamics under our choice of scaling (0.5=<@c<1) for buffer size as well as AQM parameters.

On window control algorithm for TCP efficiency over wireless networks with high bandwidth-delay product

The existing congestion control algorithms such as TCP Reno are not suitable for efficient utilisation of networks with a high bandwidth-delay product because it takes a very long time to achieve the full link utilisation. Furthermore, it is difficult to guarantee the fairness among the TCP connections with different round-trip times. To overcome these problems, a new window control algorithm using the buffer state and variable gains is proposed. The simulation results show that our algorithm improves the performance of TCP in the wireless networks.

Hop-by-hop congestion control for ad hoc wireless-LAN: a model-based control approach

This paper presents a congestion control protocol for ad-hoc Wireless-LAN (WLAN) with Bandwidth-on Demand (BoD) access. The novelty of this paper is in the extensive use of model-based control methodologies to simultaneously compute the capacity requests necessary to access the network (BoD) and the capacity allocations required to regulate the rates of the traffic flows (congestion control). The proposed scheme allows one to compute upper-bounds of the queue lengths in all the network buffers (thus allowing proper buffer dimensioning and, therefore, overflow prevention), avoids that the assigned capacity is left unused (thus entailing full link utilization) and guarantees the recovery of a satisfactory traffic behaviour as soon as congestion situations terminate (congestion recovery). The high-speed WLAN considered in the paper has been developed within the European Union (EU) project Wireless Indoor Flexible High Bitrate Modern Architecture (WINDFLEX). Extensive simulations prove the effectiveness of the proposed scheme.

A Nonlinear Flow Controller for ATM Networks

In this paper congestion control in Asynchronous Transfer Mode networks is considered. A new nonlinear flow control strategy for a single connection is proposed. The strategy combines the relay type controller with Smith predictor used for the delay compensation. The control scheme proposed in the paper is particularly suitable for loss sensitive applications. It guarantees full link utilisation and no cell loss in the controlled connection. Furthermore, transmission rates generated by the controller are limited according to the maximum available bandwidth of the connection. Finally, the paper shows that a special case of a sliding mode is induced by the proposed strategy in the controlled system.

A global stability result in network flow control

This paper concerns recently proposed algorithms for network flow control, where source transmission rates are chosen based on an individual utility function and a network price signal propagated from the links. Link prices are, in turn, dynamically updated based on overall link congestion. In particular, we consider here the second-order update law of Athuraliya et al. [IEEE Network 15 (3) (2001) 48–53], which has attractive features in equilibrium: full link utilization and no backlog. Using a non-linear, continuous-time model which enforces non-negativity constraints in prices and backlogs, we prove the global asymptotic stability of the equilibrium, by means of a Lyapunov argument.