Latency is increasingly becoming a performance bottleneck for Internet Protocol (IP) networks, but historically, networks have been designed with aims of maximizing throughput and utilization. This paper offers a broad survey of techniques aimed at tackling latency in the literature up to August 2014, as well as their merits. A goal of this work is to be able to quantify and compare the merits of the different Internet latency reducing techniques, contrasting their gains in delay reduction versus the pain required to implement and deploy them. We found that classifying techniques according to the sources of delay they alleviate provided the best insight into the following issues: 1) The structural arrangement of a network, such as placement of servers and suboptimal routes, can contribute significantly to latency; 2) each interaction between communicating endpoints adds a Round Trip Time (RTT) to latency, particularly significant for short flows; 3) in addition to base propagation delay, several sources of delay accumulate along transmission paths, today intermittently dominated by queuing delays; 4) it takes time to sense and use available capacity, with overuse inflicting latency on other flows sharing the capacity; and 5) within end systems, delay sources include operating system buffering, head-of-line blocking, and hardware interaction. No single source of delay dominates in all cases, and many of these sources are spasmodic and highly variable. Solutions addressing these sources often both reduce the overall latency and make it more predictable.

/pdf/reducing-internet-latency-a-survey-of-techniques-and-their-ed3oogdolv.pdf

Reducing Internet Latency: A Survey of Techniques and Their Merits

Since the inception of IEEE 802.11 wireless local area networks (WLANs) in 1997, wireless networking technologies have tremendously grown in the last few decades. The fundamental IEEE 802.11 physical (PHY) and medium access control (MAC) protocols have continuously been enriched with new technologies to provide the last mile wireless broadband connectivity to end users. Consequently, several new amendments of the basic IEEE 802.11 gradually came up in the forms of IEEE 802.11a, IEEE 802.11b, and IEEE 802.11g. More recently, IEEE 802.11n, IEEE 802.11ac, and IEEE 802.11ad are introduced with enhanced PHY and MAC layers that boost up physical data rates to the order of Gigabit per second. So, these amendments are generally known as high throughput WLANs (HT-WLANs). In HT-WLANs, PHY layer is enhanced with multiple-input multiple-output antenna technologies, channel bonding, short guard intervals, enhanced modulation and coding schemes. The MAC sublayer overhead is reduced by introducing frame aggregation and block acknowledgement technologies. However, several existing studies reveal that, many a time, the aforesaid PHY and MAC enhancements yield negative impact on various upper layer protocols, that is end-to-end transport and application layer protocols. As a consequence, a large number of researchers have focused on improving the coordination among PHY/MAC and upper layer protocols. In this survey, we discuss impact of HT-WLAN PHY and MAC layer enhancements on various transport and application layer protocols. This paper also summarizes several research works that use aforesaid enhancements effectively to boost up data rate of end-to-end protocols. We also point out limitations of the existing researches and list down different open challenges that can be meaningfully explored for the development of the next generation HT-WLAN technologies.

Impact of IEEE 802.11n/ac PHY/MAC High Throughput Enhancements on Transport and Application Protocols—A Survey

With the proliferation of mobile devices, and their increasingly powerful embedded processors and storage, vast resources increasingly surround users. We have been investigating the concept of on-demand ad hoc forming of groups of nearby mobile devices in the midst of crowds to cooperatively perform computationally intensive tasks as a service to local mobile users, or what we call mobile crowd computing. As devices can vary in processing power and some can leave a group unexpectedly or new devices join in, there is a need for algorithms that can distribute work in a flexible manner and still work with different arrangements of devices that can arise in an ad hoc fashion. In this article, we first argue for the feasibility of such use of crowd-embedded computations using theoretical justifications and reporting on our experiments on Bluetooth-based proximity sensing. We then present a multilayered work-stealing style algorithm for distributing work efficiently among mobile devices and compare speedups attainable for different topologies of devices networked with Bluetooth, justifying a topology-flexible opportunistic approach. While our experiments are with Bluetooth and mobile devices, the approach is applicable to ecosystems of various embedded devices with powerful processors, networking technologies, and storage that will increasingly surround users.

Mobile Computations with Surrounding Devices: Proximity Sensing and MultiLayered Work Stealing

Analyzing the effective throughput in multi-hop IEEE 802.11n networks

According to the amendment 5 of the IEEE 802.11 standard, 802.11n still uses the distributed coordination function (DCF) access method as mandatory function in access points and wireless stations (essentially to assure compatibility with previous 802.11 versions). This article provides an accurate two dimensional Markov chain model to investigate the throughput performance of IEEE 802.11n networks when frame aggregation and block acknowledgements (Block-ACK) schemes are adopted. Our proposed model considered packet loss either from collisions or channel errors. Further, it took anomalous slots and the freezing of backoff counter into account. The contribution of this work was the analysis of the DCF performance under error-prone channels considering both 802.11n MAC schemes and the anomalous slot in the backoff process. To validate the accuracy of our proposed model, we compared its mathematical simulation results with those obtained using the 802.11n DCF in the network simulator (NS-2) and with other analytical models investigating the performance of 802.11n DCF. Simulation results proved the accuracy of our model.

An accurate two dimensional Markov chain model for IEEE 802.11n DCF

In this paper, we present a measurement study of the multi-hop behavior of the new IEEE 802.11n standard in an indoor mesh testbed. As main contribution, we quantitatively describe characteristics of IEEE 802.11n on multi-hop paths like throughput, aggregate size and utilized MIMO features. Furthermore, we show how to build an IEEE 802.11n indoor testbed capable of multi-hop communication. In an 8-hop chain topology, we observe that channel bonding nearly doubles the throughput for any fixed path length. The mean aggregate size in number of frames at each node is also doubled by channel bonding. The aggregate size decreases with increasing path length. Limiting the aggregate size severely impacts throughput both on single and multi-hop paths. The throughput degrades as the path length is increased as in legacy IEEE 802.11a/b/g. We believe that these findings will enhance the understanding of the performance of IEEE 802.11n in multi-hop networks.

Understanding IEEE 802.11n multi-hop communication in wireless networks

In this paper, we study the effects of frame aggregation on video streaming performance in a real-world IEEE 802.11n testbed. We quantitatively characterize delay, video quality, and mean aggregate size in two multi-hop scenarios. We observe that streaming applications naturally take advantage of frame aggregation, both in single- and multi-stream environments, and that a Full-HD video can be transmitted with excellent quality and delay, even on a 6-hop chain. Furthermore, we discover that limiting frame aggregation severely impacts both the delay and video quality. We believe these insights on frame aggregation help for developing an effective cross-layer design for video streaming over IEEE 802.11n multi-hop networks.

/pdf/evaluating-the-impact-of-frame-aggregation-on-video-1z6mxlhueo.pdf

Evaluating the impact of frame aggregation on video-streaming over IEEE 802.11n multihop networks

In this paper we characterize the effective throughput for multi-hop paths in IEEE 802.1 In based wireless mesh networks. We derive an analytical model capturing the effects of frame aggregation and block acknowledgements, features found in the new IEEE 802.1 In standard. We describe the throughput at MAC layer as a function of bit error rate, aggregation level and path length. Using the results of our model as an upper bound, we show that current TCP implementations do not harness the bandwidth provided by the IEEE 802.1 In MAC layer. Subsequently we introduce collision induced rate control which uses cross layer feedback to effectively estimate the available bandwidth based on our analytical model. We conduct a comprehensive performance study, showing that results from our analytical model are closely matched by simulation results. Furthermore we show that the proposed protocol achieves up to 50% more goodput and up to 70% less frame collisions compared to TCP-SACK and TCP-NewReno.

/pdf/analyzing-the-effective-throughput-in-multi-hop-ieee-802-1-1yphodf8lp.pdf

Analyzing the effective throughput in multi-hop IEEE 802.1 in networks

We present an analytically tractable mathematical approach for accurately modeling the distribution of inter-contact times between mobile devices carried by users. The contribution of this paper is two-fold: (1) we show how to employ a Markov-modulated Poisson process (MMPP) for characterizing long-term dependencies in the mobility behavior, and (2) we propose to employ a graph-based clustering approach for taking into account different user groups with inhomogeneous mobility patterns. We illustrate the effectiveness of the proposed approach by considering two comprehensive real-world trace data sets. The presented quantitative results show that the proposed modeling approach closely approximates the dichotomy of the distribution of human inter-contact times into an exponential and power-law distribution observed in recent studies of real-world trace data. As the presented modeling approach for inter-contact times is both analytically tractable and captures long-term dependencies in the mobility behavior, it possesses clear advantages over methods previously introduced for analyzing the performance of opportunistic networking protocols.

Sascha Gubner

Papers

Analyzing the effective throughput in multi-hop IEEE 802.11n networks

Understanding IEEE 802.11n multi-hop communication in wireless networks

Evaluating the impact of frame aggregation on video-streaming over IEEE 802.11n multihop networks

Analyzing the effective throughput in multi-hop IEEE 802.1 in networks

An accurate and analytically tractable model for human inter-contact times