Antelope: A Framework for Dynamic Selection of Congestion Control Algorithms

Most congestion control algorithms (CCAs) are designed for specific network environments. As such, there is no known algorithm that achieves uniformly good performance in all scenarios for all flows. Rather than devising a one-size-fits-all algorithm (which is a likely impossible task), we propose a system to dynamically switch between the most suitable CCAs for specific flows in specific environments. This raises a number of challenges, which we address through the design and implementation of Antelope, a system that can dynamically reconfigure the stack to use the most suitable CCA for individual flows. We build a machine learning model to learn which algorithm works best for individual conditions and implement kernel-level support for dynamically switching between CCAs. The framework also takes application requirements of performance into consideration to fine-tune the selection based on application-layer needs. Moreover, to reduce the overhead introduced by machine learning on individual front-end servers, we (optionally) implement the CCA selection process in the cloud, which allows the share of models and the selection among front-end servers. We have implemented Antelope in Linux, and evaluated it in both emulated and production networks. The results demonstrate the effectiveness of Antelope via dynamic adjusting the CCAs for individual flows. Specifically, Antelope achieves an average 16% improvement in throughput compared with BBR, and an average 19% improvement in throughput and 10% reduction in delay compared with CUBIC.

In this chapter, layer-2 hop-by-hop congestion control mechanisms for wireless mesh networks are studied. First, background on development of congestion control as a research area is presented. Then, congestion is defined and the causes of congestion in a mesh network are identified. The steps and challenges for development of a congestion control protocol are explored by dividing congestion control mechanism into three main functions: congestion detection, signaling, and resolution. Stability and fairness in the context of rate control are studied, and global stability in the presence of delay is identified as the main challenge faced by hop-by-bop congestion control algorithms in multihop networks. The congestion control framework in the IEEE 802.11 draft standard on mesh networking is then presented. The chapter concludes with thoughts for practitioners and areas for future research in this field.

Congestion control enables network nodes to adopt certain approaches to avoid network congestion as much as possible, and respond when network congestion occurs. BBR is a transmission rate-based congestion control algorithm proposed by Google for Chromium QUIC and Linux kernel deployment. The new transmission protocol QUIC proposed by Google has a better congestion control mechanism than TCP. Based on the existing BBR congestion control algorithm, this paper proposes the IBCCA congestion control algorithm and evaluates the performance of the IBCCA, CUBIC, and BBR algorithms under the QUIC protocol in wireless LAN. When the network traffic burst flow is small or the burst duration is short, IBCCA achieves higher network goodput and lower transmission round-trip time. In the experiment of multi-stream transmission, the IBCCA algorithm achieves higher fairness while the good-put and RTT are at the same level. The BBR algorithm achieves a smaller RTT, and the goodput of the IBCCA algorithm is not much different from the CUBIC algorithm.

The main objective of congestion control is to best exploit the available network resources while preventing sustained overloads of network nodes and links. Appropriate congestion control mechanisms are essential to provide effcient operation of a network. Ensuring congestion control within vehicular ad hoc networks faces special challenges, due to the specificities of such environment (High mobility of nodes, high rate of topology changes, high variability in nodes density and neighborhood configuration, broadcast/geocast communication nature, etc.). In this context, we present in this paper a cooperative and fully distributed congestion control approach, based on dynamic scheduling and transmission of priority-based messages, to ensure reliable and safe communication architecture within VANET. Messages priorities are dynamically evaluated according to their types, the network context, and the neighboring nodes configuration. Considering the context of high reliability and real-time response required for intervehicular communications (including emergency breaking notification for example), we propose a complete validation method of our congestion control algorithms, taking into account reliability, temporal, and operational aspects.

Congestion control is a fundamental mechanism for TCP protocol, which has been extensively studied in the past three decades. However, our experimental evaluations show that the state-of-art congestion control algorithms such as Cubic and BBR are far from optimal: they have unresolved issues such as insufficient usage of available bandwidth, inadaptable to dynamic bandwidth variants, and compromising on one or more performance dimensions. To address these challenges, we propose a novel congestion control mechanism called NeuRoc that coordinately uses online changepoint detection and deep reinforcement learning (DRL) technique to generate the optimal congestion control policy, which allows TCP operating at Kleinrock ’s optimal operation point to achieve fully bandwidth usage and low latency. To address the practical issues of deploying the deep learning based congestion control mechanism, we propose a cold-started training and deployment framework to reduce the cost of bootstrap. We implement NeuRoc on an emulation platform which connects to the Linux network protocol stack through virtual network interfaces. Extensive experiments show that NeuRoc achieves the best throughput-latency tradeoff compared with the state-of-the-arts in a variety of scenarios.

Maintaining the performance of reliable transport protocols, such as transmission control protocol (TCP), over wireless mesh networks (WMNs) is a challenging problem due to the unique characteristics of data transmission over WMNs. The unique characteristics include multi-hop communication over lossy and non-deterministic wireless mediums, data transmission in the absence of a base station, similar traffic patterns over neighboring mesh nodes, etc. One of the reasons for the poor performance of conventional TCP variants over WMNs is that the congestion control mechanisms in conventional TCP variants do not explicitly account for these unique characteristics. To address this problem, this paper proposes a novel artificial intelligence based congestion control technique for reliable data transfer over WMNs. The synergy with artificial intelligence is established by exploiting a carefully designed neural network (NN) in the congestion control mechanism. We analyze the proposed NN based congestion control technique in detail and incorporate it into TCP to create a new variant that we name as intelligent TCP or iTCP. We evaluate the performance of iTCP using both ns-2 simulations and real testbed experiments. Our evaluation results demonstrate that our proposed congestion control technique exhibits a significant improvement in total network throughput and average energy consumption per transmitted bit compared to the congestion control techniques used in other TCP variants.

The Internet’s excellent scalability and robustness result in part from the end-to-end nature of Internet congestion control. End-to-end congestion control algorithms alone however are unable to prevent the congestion collapse and unfairness created by applications that are unresponsive to network congestion. To address these maladies, we propose a modified Queue Management algorithm and Stream Control Transmission Protocol (SCTP) congestion control mechanism. SCTP mechanisms are based upon TCP congestion control principles. It allows network operators to simultaneously achieve high throughput and low average delay. One of the flaws of SCTP is-fast retransmission procedure is vulnerable to being mistakenly triggered multiple times leading to under utilization of the network during recovery and duplicate retransmissions of the lost packet. To solve these problems, we change in determining the congestion window while leaving the basic idea intact and then evaluate its performance using simulation. Simulation results show that our proposed method can effectively eliminates congestion collapse and give better performance than conventional procedure. Based on extensive simulation, we believe that our technique is sufficiently robust for deployment in routers.

Classical Transmission Control Protocol (TCP) designs have never considered the identity of the competing transport protocol as useful information to TCP sources in congestion control mechanisms. When competing against a TCP flow on a bottleneck link, a User Datagram Protocol (UDP) flow can unfairly occupy the entire link bandwidth and suffocate all TCP flows on the link. If it were possible for a TCP source to know the type of transport protocol that deprives it of link access, perhaps it would be better for the TCP source to react in a way which prevents total starvation. In this paper, we use coefficient of variation and power spectral density of throughput traces to identify the presence of UDP transport protocols that compete against TCP flows on bottleneck links. Our results show clear traits that differentiate the presence of competing UDP flows from TCP flows independent of round-trip times variations. Signatures that we identified include an increase in coefficient of variation whenever a competing UDP flow joins the bottleneck link for the first time, noisy spectral density representation of a TCP flow when competing against a UDP flow in the bottleneck link, and a dominant frequency with outstanding power in the presence of TCP competition only. In addition, the results show that signatures for congestion caused by competing UDP flows are different from signatures due to congestion caused by competing TCP flows regardless of their round-trip times. The results in this paper present the first steps towards development of more ’intelligent’ congestion control algorithms with added capability of knowing the identity of aggressor protocols against TCP, and subsequently using this additional information for rate control.

Information-centric networking (ICN) is expected to be a candidate for future internet architecture, and it supports features such as multicast that improves bandwidth utilization and transmission efficiency. However, multicast itself does not provide congestion control. When multiple multicast groups coexist, multicast traffic may exhaust all network resources, and cause network congestion and packet loss. Additionally, traditional IP multicast congestion control mechanisms cannot be directly applied to ICN architecture. Therefore, it is necessary to consider an effective congestion control mechanism for ICN multicast. This paper proposes a single-rate multicast congestion control mechanism, called SRMCC. It supports router-assisted awareness of the network congestion state and congestion control message aggregation. Moreover, the fair shared rate estimation method is innovatively proposed to achieve protocol fairness. Most importantly, it adjusts the rate according to different congestion states indicated by the queue occupancy ratio. By introducing a rate selection factor, it can achieve a balance between packet loss rate and throughput. Experimental results show that our proposal outperforms other mechanisms in throughput, packet loss rate, total bandwidth utilization, and overhead, and achieves protocol fairness and better TCP friendliness.

Space-ground integrated networks (SGINs) have been recently regarded as a promising way to provide resilient, dependable as well as efficient data transmission in the high-speed railway (HSR) scenario. Applying multipath transmission control protocol (MPTCP) to SGIN can realize data transmission simultaneously via terrestrial and satellite networks. However, since the existing congestion control (CC) mechanisms of MPTCP fail to distinguish between adverse influences (such as packet loss and/or round-trip time increase) caused by congestion and those caused by handovers, it suffers severe performance degradation in the HSR scenario where handover frequently occurs. In this article, we first present the SGIN oriented HSR (SGIN-HSR) with MPTCP. Then leveraging cross-layer information (i.e., reference signal received power), we design a novel cross-layer aided MPTCP CC mechanism targeted at SGIN-HSR based on deep reinforcement learning, which is referred to as HSR-CC, to alleviate performance degradation problems induced by handover. The experimental results show that HSR-CC significantly enhances the goodput and outperforms state-of-the-art MPTCP CC algorithms in SGIN-HSR environments where handover frequently occurs.

Simulation of ATM switches and networks is a computationally demanding problem as compared to simulation of conventional packet-based networks, owing to the large number of cell events that need to be simulated in the former. To address this problem, we are developing a flexible hardware simulation testbed in the High-Speed Networks Laboratory at the University of California, Santa Cruz. The testbed, called FAST (FPGA-based ATM simulation testbed), uses high-density field-programmable gate arrays (FPGAs) to allow implementation of the key simulation components such as traffic generators, switching fabric, buffer management, traffic scheduling, congestion control mechanisms, and statistics recording in hardware. In the first version of the testbed (FAST-1), each board consists of 13 Altera FLEX devices (including 4 multichip modules), providing a total of 336,000 usable gates. Each board can be used to simulate an ATM switch. Multiple boards may be interconnected to simulate large ATM networks. Software tools are being developed for specifying the components of the underlying simulation model, such as the switch structure, traffic arrival process, traffic scheduling algorithm, and congestion control algorithm; synthesizing the specifications into the individual FPGAs; controlling and monitoring the simulation; and collecting and reporting statistics. The paper provides an overview of the architecture of the FAST-1 board, describes its key components, and discusses an example simulation of a traffic scheduling algorithm using the board.

Congestion control is a fundamental mechanism for the stability of the Internet and is a central mechanism for TCP. However, this congestion control mechanism focuses mainly on the core network state and is blind to the characteristics of wireless and mobile access networks. Moreover, TCP window based congestion control ignores totally application layer QoS needs and entails throughput variations which are not compliant with application layer QoS constraints such as bandwidth, delay and jitter. The TCP-Friendly Rate Control protocol (TFRC) was originally designed in the context of wired networks. This protocol is one of the most convincing attempt to provide a congestion control mechanism adapted to multimedia flows, although limited in its capacity to fully address these issues. After an identification and evaluation of the subtle counterproductive interactions between the WLANs MAC layer and the transport layer, this paper shows a new approach towards congestion control for WLANs. This paper also introduces a specialization of TFRC (MTFRC: Mobile TFRC), which is adapted to wireless access networks. This TFRC specialization requires only slight changes to the standard TFRC protocol. Simulation results show substantial improvements for applications over TFRC in scenarios where the bottleneck situates on the MAC layer of the mobile nodes.

As the Internet becomes increasingly heterogeneous, the issue of congestion control becomes ever more important. And the queue length and end-to-end (congestion) delays are some of the important things in term of congestion avoidance and control mechanisms. In this research we continued to study the performances of the New-Additive Increase Multiplicative Decrease (AIMD) algorithm as one of the core protocols for TCP congestion avoidance and control mechanism, we want now to evaluate the effect of using the New-AIMD algorithm to measure the queue length and end-to-end delays and we will use the NCTUns simulator to get the results after make the modification of the mechanism. And we will use the Drop tail mechanism as Active Queue Management (AQM) in the bottleneck router. After implementation of our new approach with different number of flows, we will measure the delay for two types of delays (queuing delay and end-to-end delay), we expect the delay will be less with using our mechanism comparing with the mechanism in the previous study. Now and after got this results as low delay for bottleneck link case, we know the New-AIMD mechanism work as well under the network condition in the experiments.

In 2016, Google proposed the bottleneck bandwidth round-trip propagation time (BBR) as a new congestion control algorithm for improving network speed. As BBR has different behavioral characteristics from loss-based congestion control, it must ensure the coexistence between congestion control algorithms. However, throughput imbalance has been reported in a competition between BBR and loss-based congestion control. In this paper, we propose an opponent congestion control identification model of BBR (OI-BBR) with machine learning to improving inter-protocol fairness. In the emulation experiments, we confirmed that OI-BBR based on the proposed model improved the inter-protocol fairness by operating differently according to the opponent congestion control algorithms.

In the past few decades, many Transmission Control Protocol (TCP) congestion control algorithms have been investigated to meet the growing network demands and to enhance the performance of TCP in lossy or high-bandwidth-delay-product (high-BDP) networks. However, it is still challenging to implement a dynamic congestion control algorithm for wide range of diverse mobile users, network conditions and applications. This paper explores avenues for enhancement of TCP congestion control algorithm for next generation mobile networks by dynamically learning the available bandwidth and deriving the Congestion Control Factor N. N is used to Adaptive Increase/Adaptive Decrease (AIAD) the Congestion Window (CWND) dynamically instead of using the traditional approach that is Additive Increase/ Multiplicative Decrease (AIMD) paradigm. Once there is congestion, our proposed algorithm will not allow the CWND to decrease multiplicatively or steeply. After dropping to a certain level (lower than legacy), we try to take the CWND to previous state adaptively with the help of calculated bandwidth (based on learning). This in turn helps to efficiently control the CWND for better network utilization especially in case of lossy and high-BDP conditions. As soon as it reaches the original state, it remains stable for longer time as compared to legacy until packet loss or time out. We demonstrate the effectiveness of our algorithm with the help of live air experiments (performed in Samsung R&D India, Bangalore) and NS3 based simulation experiments. Through our experiments, we show that our algorithm outperforms the legacy congestion control algorithms (like CUBIC, RENO, TCPW) and the existing CLTCP algorithm in terms of goodput, intra algorithm fairness and inter algorithm fairness maintaining the scalability and friendliness.