Soung Chang Liew

TCP Veno: TCP enhancement for transmission over wireless access networks

Wireless access networks in the form of wireless local area networks, home networks, and cellular networks are becoming an integral part of the Internet. Unlike wired networks, random packet loss due to bit errors is not negligible in wireless networks, and this causes significant performance degradation of transmission control protocol (TCP). We propose and study a novel end-to-end congestion control mechanism called TCP Veno that is simple and effective for dealing with random packet loss. A key ingredient of Veno is that it monitors the network congestion level and uses that information to decide whether packet losses are likely to be due to congestion or random bit errors. Specifically: (1) it refines the multiplicative decrease algorithm of TCP Reno-the most widely deployed TCP version in practice-by adjusting the slow-start threshold according to the perceived network congestion level rather than a fixed drop factor and (2) it refines the linear increase algorithm so that the connection can stay longer in an operating region in which the network bandwidth is fully utilized. Based on extensive network testbed experiments and live Internet measurements, we show that Veno can achieve significant throughput improvements without adversely affecting other concurrent TCP connections, including other concurrent Reno connections. In typical wireless access networks with 1% random packet loss rate, throughput improvement of up to 80% can be demonstrated. A salient feature of Veno is that it modifies only the sender-side protocol of Reno without changing the receiver-side protocol stack.

Solutions to performance problems in VoIP over a 802.11 wireless LAN

Voice over Internet Protocol (VoIP) over a wireless local area network (WLAN) is poised to become an important Internet application. However, two major technical problems that stand in the way are: 1) low VoIP capacity in WLAN and 2) unacceptable VoIP performance in the presence of coexisting traffic from other applications. With each VoIP stream typically requiring less than 10 kb/s, an 802.11b WLAN operated at 11 Mb/s could in principle support more than 500 VoIP sessions. In actuality, no more than a few sessions can be supported due to various protocol overheads (for GSM 6.10, it is about 12). This paper proposes and investigates a scheme that can improve the VoIP capacity by close to 100% without changing the standard 802.11 CSMA/CA protocol. In addition, we show that VoIP delay and loss performance in WLAN can be compromised severely in the presence of coexisting transmission-control protocol (TCP) traffic, even when the number of VoIP sessions is limited to half its potential capacity. A touted advantage of VoIP over traditional telephony is that it enables the creation of novel applications that integrate voice with data. The inability of VoIP and TCP traffic to coexist harmoniously over the WLAN poses a severe challenge to this vision. Fortunately, the problem can be largely solved by simple solutions that require only changes to the medium-access control (MAC) protocol at the access point. Specifically, in our proposed solutions, the MAC protocol at the wireless end stations does not need to be modified, making the solutions more readily deployable over the existing network infrastructure.

Channel coding and decoding in a relay system operated with physical-layer network coding

This paper investigates link-by-link channel-coded PNC (physical layer network coding), in which a critical process at the relay is to transform the superimposed channel-coded packets received from the two end nodes (plus noise), Y₃ = X₁+ X₂+W₃, to the network-coded combination of the source packets, S₁ oplus S₂. This is in contrast to the traditional multiple-access problem, in which the goal is to obtain both S₁ and S₂ explicitly at the relay node. Trying to obtain S₁ and S₂ explicitly is an overkill if we are only interested in S1oplusS₂. In this paper, we refer to the transformation Y₃ rarr S₁ oplus S₂ as the channel-decoding- network-coding process (CNC) in that it involves both channel decoding and network coding operations. This paper shows that if we adopt the repeat accumulate (RA) channel code at the two end nodes, then there is a compatible decoder at the relay that can perform the transformation Y₃ rarr S₁oplusS₂ efficiently. Specifically, we redesign the belief propagation decoding algorithm of the RA code for traditional point-to-point channel to suit the need of the PNC multiple-access channel. Simulation results show that our new scheme outperforms the previously proposed schemes significantly in terms of BER without added complexity.

Physical-layer network coding: Tutorial, survey, and beyond ✩

Abstract The concept of physical-layer network coding (PNC) was proposed in 2006 for application in wireless networks. Since then it has developed into a subfield of network coding with wide implications. The basic idea of PNC is to exploit the mixing of signals that occurs naturally when electromagnetic (EM) waves are superimposed on one another. In particular, at a receiver, the simultaneous transmissions by several transmitters result in the reception of a weighted sum of the signals. This weighted sum is a form of network coding operation by itself. Alternatively, the received signal could be transformed and mapped to other forms of network coding. Exploiting these facts turns out to have profound and fundamental ramifications. Subsequent works by various researchers have led to many new results in the domains of (1) wireless communication, (2) information theory, and (3) wireless networking. The purpose of this paper is fourfold. First, we give a brief tutorial on the basic concept of PNC. Second, we survey and discuss recent key results in the three aforementioned areas. Third, we examine a critical issue in PNC: synchronization. It has been a common belief that PNC requires tight synchronization. Recent results suggest, however, that PNC may actually benefit from asynchrony. Fourth, we propose that PNC is not just for wireless networks; it can also be useful in optical networks. We provide an example showing that the throughput of a passive optical network (PON) could potentially be raised by 100% with PNC.

Throughput analysis of IEEE802.11 multi-hop ad hoc networks

In multi-hop ad hoc networks, stations may pump more traffic into the networks than can be supported, resulting in high packet-loss rate, re-routing instability and unfairness problems. This paper shows that controlling the offered load at the sources can eliminate these problems. To verify the simulation results, we set up a real 6-node multi-hop network. The experimental measurements confirm the existence of the optimal offered load. In addition, we provide an analysis to estimate the optimal offered load that maximizes the throughput of a multi-hop traffic flow. We believe this is a first paper in the literature to provide a quantitative analysis (as opposed to simulation) for the impact of hidden nodes and signal capture on sustainable throughput. The analysis is based on the observation that a large-scale 802.11 network with hidden nodes is a network in which the carrier-sensing capability breaks down partially. Its performance is therefore somewhere between that of a carrier-sensing network and that of an Aloha network. Indeed, our analytical closed-form solution has the appearance of the throughput equation of the Aloha network. Our approach allows one to identify whether the performance of an 802.11 network is hidden-node limited or spatial-reuse limited.

Asynchronous Physical-Layer Network Coding

A key issue in physical-layer network coding (PNC) is how to deal with the asynchrony between signals transmitted by multiple transmitters. That is, symbols transmitted by different transmitters could arrive at the receiver with symbol misalignment as well as relative carrier-phase offset. A second important issue is how to integrate channel coding with PNC to achieve reliable communication. This paper investigates these two issues and makes the following contributions: 1) We propose and investigate a general framework for decoding at the receiver based on belief propagation (BP). The framework can effectively deal with symbol and phase asynchronies while incorporating channel coding at the same time. 2) For unchannel-coded PNC, we show that for BPSK and QPSK modulations, our BP method can significantly reduce the asynchrony penalties compared with prior methods. 3) For QPSK unchannel-coded PNC, with a half symbol offset between the transmitters, our BP method can drastically reduce the performance penalty due to phase asynchrony, from more than 6 dB to no more than 1 dB. 4) For channel-coded PNC, with our BP method, both symbol and phase asynchronies actually improve the system performance compared with the perfectly synchronous case. Furthermore, the performance spread due to different combinations of symbol and phase offsets between the transmitters in channel-coded PNC is only around 1 dB. The implication of 3) is that if we could control the symbol arrival times at the receiver, it would be advantageous to deliberately introduce a half symbol offset in unchannel-coded PNC. The implication of 4) is that when channel coding is used, symbol and phase asynchronies are not major performance concerns in PNC.

Physical Layer Network Coding Schemes over Finite and Infinite Fields

Direct application of network coding at the physical layer - physical layer network coding (PNC) - is a promising technique for two-way relay wireless networks. In a two-way relay network, relay nodes are used to relay two-way information flows between pairs of end nodes. This paper proposes a precise definition for PNC. Specifically, in PNC, a relay node does not decode the source information from the two ends separately, but rather directly maps the combined signals received simultaneously to a signal to be relayed. Based on this definition, PNC can be further sub-classed into two categories - PNCF (PNC over finite field) and PNCI (PNC over infinite field) - according to whether the network-code field (or groups, rings) adopted is finite or infinite. For each of PNCF and PNCI, we consider two specific estimation techniques for dealing with noise in the mapping process. The performance of the four schemes is investigated by means of analysis and simulation, assuming symbol-level time synchronization only.

Improving Throughput and Fairness by Reducing Exposed and Hidden Nodes in 802.11 Networks

Two well-known problems that can cause performance degradations in IEEE 802.11 wireless networks are the exposed-node (EN) and hidden-node (HN) problems. Although there have been isolated and incidental studies of EN and HN, a comprehensive treatment has not been attempted. The contributions of this paper are threefold: First, we provide rigorous mathematical definitions for EN and HN in wireless networks (including wireless local area networks (WLANs) with multiple access points (APs) and ad hoc networks). Second, we relate EN to the nonscalability of network throughput and HN to unfair throughput distributions. Third, we provide schemes to eliminate EN and HN, respectively. We show that the standard 802.11 technology is not scalable because, due to EN, more APs do not yield higher total throughput. By removing EN, our schemes make it possible to achieve scalable throughput commensurate with the seminal theoretical results in [1] and [2]. In addition, by removing HN, our schemes solve the performance problems triggered by HN, including throughput unfairness/starvation and rerouting instability.

Proportional fairness in wireless LANs and ad hoc networks

The paper considers scenarios in which fairness and efficiency are two conflicting objectives in wireless networks, and investigates the use of a proportional fairness objective to strike a balance between the two objectives. We explain the physical meaning of proportional fairness in a wireless network and give an analysis showing that proportional fairness is equivalent or close to max-min fairness in terms of air-time usage (as opposed to bandwidth usage). For infrastructure WLANs, two approaches to achieving proportional fairness are discussed. For ad hoc networks, achieving proportional fairness is more complex and requires global information on contention among different traffic flows. We propose and evaluate the use of a distributed max-min air-time allocation algorithm to approximate the proportional fairness objective.

A rate adaptation algorithm for IEEE 802.11 WLANs based on MAC-layer loss differentiation

In a WLAN subject to variable wireless channel conditions, rate adaptation plays an important role to more efficiently utilize the physical link. However, the existing rate adaptation algorithms for IEEE 802.11 WLANs do not take into account the loss of frames due to collisions. In a WLAN with coexistence of multiple stations, two types of frame losses due to (a) link errors and (b) collisions over the wireless link can coexist and severely degrade the performance of the existing rate adaptation algorithms. In this paper, we propose a new automatic rate fallback algorithm that can differentiate the two types of losses and sharpen the accuracy of the rate adaptation process. Numerical results show that the new algorithm can substantially improve the performance of IEEE 802.11 WLANs.