Achieving high throughput and low delay by accurately regulating link queue length over mobile data network

2015 12th Annual IEEE International Conference on Sensing, Communication, and Networking (SECON)

et al. 2015

Self Cite

-Recent advances in high-speed mobile networks have revealed new bottlenecks in ubiquitous TCP protocol deployed in the Internet. In addition to differentiating non-congestive loss from congestive loss, our experiments revealed two significant performance bottlenecks during the loss recovery phase: flow control bottleneck and application stall, resulting in degradation in QoS performance. To tackle these two problems we firstly develop a novel opportunistic retransmission algorithm to eliminate the flow control bottleneck, which enables TCP sender to transmit new packets even if receiver's receiving window is exhausted. Secondly, application stall can be significantly alleviated by carefully monitoring and tuning the TCP sending buffer growth mechanism. We implemented and modularized the proposed algorithms in the Linux kernel thus they can plug-and-play with the existing TCP loss recovery algorithms easily. Using emulated experiments we showed that, compared to the existing TCP loss recovery algorithms, the proposed optimization algorithms improve the bandwidth efficiency by up to 133% and completely mitigate RTT spikes, i.e., over 50% RTT reduction, over the loss recovery phase.

Section: B Current Sending Buffer Growth Mechanismmentioning

confidence: 99%

Optimizing TCP loss recovery performance over mobile data networks

Zha

2015 12th Annual IEEE International Conference on Sensing, Communication, and Networking (SECON)

et al. 2015

Self Cite

“…TCP-SoD [8] addresses this problem with Sum-of-delay (SoD) queue length estimation, which measures the actual bandwidth from the ACK timings and records the sequence of past measurements, to compensate for link bandwidth variations in estimating the queue length. Chan et al [10] showed that SoD can achieve significantly higher estimation accuracy, even in mobile networks with rapidly changing bandwidth.…”

Section: Queue Length Estimationmentioning

confidence: 99%

“…In our previous research [8], we developed TCP-QLA, which employs the SoD-TS [9] queue length estimation method to address the problem of delay variations. This method directly measures the uplink delay variations experienced by every ACK using the TCP timestamp option [16] to accurately estimate the queuing delay.…”

Section: Queue Length Estimationmentioning

confidence: 99%

“…One method for estimating and regulating the queue length is the queue-length-based congestion control (QCC) algorithm [2,3,8], which uses various methods to estimate the backlog in the bottleneck link and limit the backlog to a desired queue length by adjusting the CWnd accordingly. Other protocols, such as TCP Vegas [2], FAST TCP [3] and TCP-SoD [8], include only the link bandwidth variations in estimating the queue length, and variations in the uplink delay, i.e., from the mobile device to the Internet, are assumed to be constant.…”

Section: Introductionmentioning

confidence: 99%

“…Other protocols, such as TCP Vegas [2], FAST TCP [3] and TCP-SoD [8], include only the link bandwidth variations in estimating the queue length, and variations in the uplink delay, i.e., from the mobile device to the Internet, are assumed to be constant. TCP-QLA [8] uses the Sum-of-delay with Timestamp (SoD-TS) method [9], which exploits the TCP timestamp option, to accurately estimate the queue length in mobile networks with both bandwidth variations and uplink delay variations.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Achieving high throughput and low delay in mobile data networks by accurately predicting queue lengths

Proceedings of the 12th ACM International Conference on Computing Frontiers

Lee

2015

Self Cite

Knowledge of the queue length for a radio link in a mobile data network has a significant effect on the performance of the communication protocol TCP. If the queue length can be accurately estimated and regulated to a target value, then low end-to-end delay and high bandwidth utilization can be achieved. One method for estimating and regulating the queue length is the queue-length-based congestion control (QCC) algorithm. However, this algorithm estimates the queue length over one RTT interval prior to transmission, and the actual queue length after that time can differ significantly, because the bandwidth can vary substantially between the neighboring propagation delays, which could result in a false positive in the queue length adaption, thereby affecting the QoS performance. To address this problem, we propose PQ-TCP, a method that predicts the queue length directly by predicting the bandwidth variations over the ensuing period of time equal to the propagation delay and using post-bandwidth analysis to minimize the prediction error. Trace-driven simulations are used to show that the QoS performance of PQ-TCP is superior to that of current QCC algorithms. PQ-TCP achieves the lowest RTT while maintaining nearly 90% bandwidth utilization for a small target queue length of 5 packets.

Optimizing TCP Loss Recovery Performance Over Mobile Data Networks

IEEE Trans. on Mobile Comput.

Zha

Wan

et al. 2020

Self Cite

Recent advances in high-speed mobile networks have revealed new bottlenecks in ubiquitous TCP protocol deployed in the Internet. In addition to differentiating non-congestive loss from congestive loss, our experiments revealed two significant performance bottlenecks during the loss recovery phase: flow control bottleneck and application stall, resulting in degradation in QoS performance. To tackle these two problems we firstly develop a novel opportunistic retransmission algorithm to eliminate the flow control bottleneck, which enables TCP sender to transmit new packets even if receiver's receiving window is exhausted. Secondly, application stall can be significantly alleviated by carefully monitoring and tuning the TCP sending buffer growth mechanism. We implemented and modularized the proposed algorithms in the Linux kernel thus they can plug-and-play with the existing TCP loss recovery algorithms easily. Using emulated experiments we showed that, compared to the existing TCP loss recovery algorithms, the proposed optimization algorithms improve the bandwidth efficiency by up to 133% and completely mitigate RTT spikes, i.e., over 50% RTT reduction, over the loss recovery phase.