Pai-Hsiang Hsiao

Load-balancing routing for wireless access networks

The widespread use of wireless devices presents new challenges for network operators, who need to provide service to ever larger numbers of mobile end users, while ensuring quality-of-service guarantees. We describe a new distributed routing algorithm that performs dynamic load-balancing for wireless access networks. The algorithm constructs a load-balanced backbone tree, which simplifies routing and avoids per-destination state for routing and per-flow state for QoS reservations. We evaluate the performance of the algorithm using several metrics including adaptation to mobility, degree of load-balance, bandwidth blocking rate, and convergence speed. We find that the algorithm achieves better network utilization by lowering bandwidth blocking rates than other methods.

WSN07-1: Adjacent Channel Interference in Dual-radio 802.11a Nodes and Its Impact on Multi-hop Networking

We evaluate the performance impact of adjacent channel interference (ACI) in multi-hop wireless networks based on dual-radio 802.11a nodes. Although these nodes use chipsets that satisfy the transmit-mask requirements set by the IEEE 802.11 standard, the multi-hop performance is still significantly affected by ACI. That is, a nodes transmitter can interfere with its own receiver on a different channel; as a result, multi-hop throughput is severely degraded. This degradation is especially pronounced for 802.11a. We use a spectrum analyzer with a signal combiner to quantify ACI under various conditions and propose solutions to mitigate the effect of such interference on multi-hop forwarding. Field experiments with multi-hop relay have validated these findings as well as the effectiveness of our solutions.

Performance Measurement of 802.11a Wireless Links from UAV to Ground Nodes with Various Antenna Orientations

We report measured performance of 802.11a wireless links from an unmanned aerial vehicle (UAV) to ground stations. In a set of field experiments, we record the received signal strength indicator (RSSI) and measure the raw link-layer throughput for various antenna orientations, communication distances and ground-station elevations. By comparing the performance of 32 simultaneous pairs of UAV and ground station configurations, we are able to conclude that, in order to achieve the highest throughput under a typical flyover UAV flight path, both the UAV and the ground station should use omni-directional dipole (as opposed to high-gain, narrow- beam) antennas positioned horizontally, with their respective antenna null pointing to a direction perpendicular to the UAVs flight path. In addition, a moderate amount of elevation of the ground stations can also improve performance significantly.

Video over TCP with receiver-based delay control

Unicasting video streams over TCP connections is a challenging problem because video sources cannot normally adapt to delay and throughput variations of TCP connections. This paper points out a direction on how TCP can be modified such that TCP connections can carry hierarchically-encoded layered video streams well, while being friendly to other competing flows. The method is calledReceiver-based Delay Control(RDC). Under RDC, a TCP connecðtion can slow down its transmission rate to avoid congestion by delaying ACK packet generation at the TCP receiver based on notifications from routers. The paper presents the principle behind RDC, argue that it is TCP-friendly, describe an implementation that uses 1-bit congestion notification from routers, and give our simulation results.

Active delay control for TCP

Active delay control is a novel extension for TCP, where TCP endpoints impose delays on the transmission of packets to improve performance. The amount of delay can be calculated by routers from the level of congestion, or by endpoints from the received congestion signals. In particular, when there are many TCP flows competing for the bandwidth of a link, they can reduce their transmission rates to arbitrary degrees by increasing delays, without experiencing TCP time-outs. Active delay control is therefore useful for those long-lived TCP-based applications that cannot tolerate time-outs. Examples of such applications are video streaming and storage networks. It is also useful for short-lived flows that require short transfer time. Examples of such applications are HTTP transactions. We present the concept and motivation behind active delay control, and evaluate them by simulation.

Transmit Antenna Selection Based on Link-layer Channel Probing

In this paper, we propose transmit antenna selection based on receiver feedback of channel information obtained via link-layer probing. Furthermore, we report the performance gain of the proposed antenna selection scheme in an experimental multi-antenna 802.11 network. We built a low-altitude Unmanned Aerial Vehicle (UAV) testbed using commodity dual-antenna 802.11 hardware and performed field experiments to collect traces of link performance using antennas of various types and orientations. Based on the collected traces, we demonstrate that transmit antenna selection can achieve a significant amount of gain using a link-layer channel probing protocol at a relatively low probing rate. The largest improvement we observed with joint transmit/receive antenna selection in 2x2 systems was 32%, about twice as much as that of receive-only antenna selection in 1x2 systems, which achieved 17%. Moreover, a similar improvement is obtained with probing intervals up to about 200 milliseconds, which is infrequently enough to consume only a small fraction of the available 802.11 channel capacity. Since these results require only a low implementation and operational cost, we conclude that transmit antenna selection is a worthwhile technique to use with the kind of multi-antenna mobile ad-hoc networks we examined.

TCP with sender-based delay control

This paper describes a congestion control method for TCP that adjusts the transmission rate of a TCP connection by changing not only the congestion window size as in normal TCP, but also by delaying the transmission of packets at the sender. We refer to this method as TCP with sender-based delay control, or simply SDC. SDC can keep the window size of a TCP connection above a certain threshold even when its fair share of bandwidth is arbitrarily small. Since TCP fast retransmit and recovery is likely to work when the window size of the connection is sufficiently large, the new method can result in reduced frequency of TCP timeouts for the connection. In particular, SDC allows many TCP flows to share a link without experiencing many timeouts. In addition, SDC can reduce a well-known TCP bias against connections with large RTTs. This paper presents the principle behind SDC, and simulation results demonstrating its properties and advantages.

Wireless open service networks

We describe a networking infrastructure, called wireless open service networks, that allows service providers such as content providers to compete in open marketplaces. Using this open infrastructure, customers are empowered with dynamic choice of a variety of services at various levels, beyond just their traditional choices in network access services. In the meantime, service providers are given the opportunity to bundle services at multiple levels to maximize the performance of service delivery to end users. For the implementation of these wireless open service networks, we present a service address resolution protocol (SARP). Using SARP requests, customers can discover available services over a local wireless network. Using SARP replies, service providers can make their services known to customers. SARP resides at the link-layer and is light-weight, to allow easy and ubiquitous deployment. SARP can be viewed as a natural extension of the widely used address resolution protocol (ARP) which discovers link-layer addresses such as Ethernet MAC addresses. We give application examples to illustrate these concepts. In addition, we discuss the design of SARP and our implementation experience.