Jae Chung

Analysis of active queue management

Active Queue Management (AQM) is intended to achieve high link utilization with a low queuing delay. Recent studies show that RED, one of the most well-known AQMs, is difficult to configure and does not provide significant performance gains given the complexity required for proper configuration. Recent variants of RED, such as Adaptive-RED are designed to provide more robust RED performance under a wider-range of traffic conditions but have not yet been evaluated. This paper presents a router queue behavior model (a queue law) for TCP-dropping and TCP-marking control systems, and uses the queue law to illustrate the impact of TCP traffic on the load and queue behavior of congested routers. Through queue law analysis and simulation, this paper confirms that RED-like AQM techniques that employ packet dropping do not significantly improve performance over that of drop-tail queue management. However, when AQM techniques use Explicit Congestion Notification (ECN) as a method to notify TCP sources of congestion rather than packet drops, the performance gains of AQM in terms of goodput and delay can be significant over that of drop-tail queue management.

Dynamic-CBT and ChIPS—router support for improved multimedia performance on the Internet

The explosive increase in the volume and variety of Internet traffic has placed a growing emphasis on congestion control and fairness in Internet routers. Approaches to the problem of congestion, such as active queue management schemes like Random Early Detection (RED) use congestion avoidance techniques and are successful with TCP flows. Approaches to the problem of fairness, such as Fair Random Early Drop (FRED), keep per-flow state and punish misbehaved, non-TCP flows. Unfortunately, these punishment mechanisms also result in a significant performance drop for multimedia flows that use flow control. We extend Class-Based Threshold (CRT) [12], and propose a new active queue management mechanism as an extension to RED called Dynamic Class-Based Threshold (D- CBT) to improve multimedia performance on the Internet. Also, as an effort to reduce multimedia jitter, we propose a lightweight packet scheduling called Cut-In Packet Scheduling (ChIPS) as an alternative to FIFO packet scheduling. The performance of our proposed mechanisms is measured, analyzed and compared with other mechanisms (RED and CBT) in terms of throughput, fairness and multimedia jitter through simulation using NS. The study shows that D-CBT improves fairness among different classes of flows and ChIPS improves multimedia jitter without degrading fairness.

Application, network and link layer measurements of streaming video over a wireless campus network

The growth of wireless LANs has brought the expectation for high-bitrate streaming video to wireless PCs. However, it remains unknown how to best adapt video to wireless channel characteristics as they degrade. This paper presents results from experiments that stream commercial video over a wireless campus network and analyze performance across application, network and wireless link layers. Some of the key findings include: 1) Wireless LANs make it difficult for streaming video to gracefully degrade as network performance decreases; 2) Video streams with multiple encoding levels can more readily adapt to degraded wireless network conditions than can clips with a single encoding level; 3) Under degraded wireless network conditions, TCP streaming can provide higher video frame rates than can UDP streaming, but TCP streaming will often result in significantly longer playout durations than will UDP streaming; 4) Current techniques used by streaming media systems to determine effective capacity over wireless LAN are inadequate, resulting in streaming target bitrates significantly higher than can be effectively supported by the wireless network.

Rate-based active queue management with priority classes for better video transmission

Video streaming on the Internet often suffers from high frame loss rates due to fragmentation of large frames and inter-frame dependencies needed for high compression. We propose adding lightweight, priority-based queue management to Internet routers to significantly improve performance of video streaming. We extend a Rate-Based RED approach to support three priority classes and apply it to MPEG. The performance of the mechanism on video streams is measured, analyzed and compared with Drop-Tail and RED in terms of transport layer throughput, system fairness, application layer throughput and goodput, and video stream quality. Extensive simulation shows that our approach improves MPEG video stream quality and network system fairness over traditional RED under a variety of workloads.

Empirical evaluation of the congestion responsiveness of RealPlayer video streams

Increasingly popular commercial streaming media applications over the Internet often use UDP as the underlying transmission protocol for performance reasons. Hand-in-hand with the increase in streaming media comes the impending threat of unresponsive UDP traffic, often cited as the major threat to the stability of the Internet. Unfortunately, there are few empirical studies that analyze the responsiveness, or lack of it, of commercial streaming media applications. In this work, we evaluate the responsiveness of RealNetworks’ RealVideo over UDP by measuring the performance of numerous streaming video clips selected from a variety of RealServers on the Internet, analyze the TCP-Friendliness of the UDP streams and correlate the results with network and application layer statistics. We find that most RealVideo UDP streams respond to Internet congestion by reducing the application layer encoding rate, and streams with a minimum encoding rate less than the fair share of the capacity often achieve a TCP-Friendly rate. In addition, our results suggest that a reason streaming applications choose not to use TCP is that the TCP API hides network information, such as loss rate and round-trip time, making it difficult to estimate the available capacity for effective media scaling.

Stochastic Fair Traffic Management for Efficient and Robust IP Networking

As use of non-TCP applications such as streaming media and network games increases, the potential for unfair, misbehaving flows and the threat of congestion collapse also increases. This paper introduces a statistical traffic filtering technique, stochastic fairness guardian (SFG), that effectively regulates misbehaving flows with minimal traffic state information. SFG can be used in conjunction with an active queue management (AQM) mechanism to improve both network protection and efficiency. Simulations are used to evaluated SFG and the integration of SFG with a proportional-integral (PI) controller in comparison with other similar statistical flow management mechanisms including RED-PD, SFB and CHOKe. The SFG-PI combination outperforms other mechanisms in terms of fairness, queuing delay, stability and TCP performance over a wide range of realistic traffic loads and conditions.