Current congestion control algorithms ignore the application-layer delay, where the untransmitted data waiting at source nodes degrades the delay performance of services. Moreover, differentiated priorities are necessary for the application services with various delay requirements, especially for mission-critical services. Different from the existing works only considering the network delay, in this paper, by adding the flow queueing delay at source nodes, we formulate the network utility maximization (NUM) problem with additional deterministic service delay constraints. We propose a general TCP-based two-timescale congestion window control (TCWC) framework with delay-aware priority to enhance traditional algorithms. Specifically, to handle the obstacle of new delay constraints, we transform them to the time-average stability of virtual queues. By solving the new NUM problem via Lyapunov optimization, we design a short-term congestion window adjustment strategy in each time slot. To further guarantee the service delay, we apply extreme value theory (EVT) to evaluate the priorities of different flows, and determine the long-term control of window update rates. We deploy the proposed framework in three classic algorithms including NewReno, Vegas and DCTCP. In addition, simulation results show that our TCWC framework can significantly reduce the average service delay and provide deterministic guarantees compared with time-aware TCP congestion control algorithms such as TIMELY and BBRv2.

Low earth orbit satellite networks (LEOSNs) are increasingly favored for providing ubiquitous Internet access. However, the dynamic characteristics in LEOSNs pose two challenging issues to congestion control algorithms (CCAs) in transport layer: 1) time-varying link capacity when links remain connected, leading to a continuous mismatch between CCA's sending rate and capacity, and 2) brief but significant link interruptions during satellite handovers. To address these issues, this paper proposes a two-phase CCA called Creo, in which each phase tackles one of the two aforementioned issues in LEOSNs individually. In Creo's connected phase where links remain connected, we design a deep reinforcement learning framework, which captures complex patterns of highly variable link capacities in LEOSNs to generate dynamically adaptive congestion control strategies. In Creo's handover phase where links suffer interruptions, we introduce a handover-aware process, which leverages cross-layer notifications to notify TCP sender in advance of handover occurrence to instruct sender when to stop and resume sending at precise timestamps. Extensive simulation results show that, compared to other CCAs, Creo consistently tracks time-varying capacity, reduces average handover recovery time by 60.5%, and overall achieves a 55% average throughput improvement while maintaining low latency and low delay jitter.

The high mobility in Vehicular Ad-hoc Networks (VANETs) significantly affects the reliability of data transmission. To solve this problem, Named Data Networking (NDN)-based VANETs are proposed, utilizing in-network caching and named-based forwarding to overcome the dual challenges of mobility and connectivity. Although in-network caching enhances content availability, a strategy that accurately locates and efficiently utilizes the cached content in VANETs with highly dynamic environments is still lacking. In this paper, we propose a novel NDN-based cache-aware adaptive forwarding (CAAF) strategy for VANETs. CAAF proactively predicts content locations and ensures reliable content retrieval by adaptively selecting forwarding nodes that prioritize fast delivery and stable transmission. Specifically, we design a content information table for each vehicle to record information about the Interest packets it receives. Furthermore, these tables are updated periodically across all vehicles and a prediction model is used to predict real-time in-network caching during the update interval. Subsequently, we execute a filter mechanism to sieve candidate forwarding vehicles that satisfy both the accessibility and stability requirements. These candidates are then evaluated using a multi-attribute decision-making method across diverse parameters to determine the optimal forwarding node. Our extensive simulation results demonstrate that the proposed CAAF outperforms the state-of-the-art forwarding strategy regarding content retrieval delay and Interest satisfaction ratio across diverse scenarios.

Multipath TCP (MPTCP) has gained popularity to enhance data transmission. From the last decade, proposed MPTCP enhancement schemes for congestion control, path management, and packet scheduling, have been used to benefit transmission performance. However, despite their efforts, they are exigent with a comprehensive understanding of real-world performance to guide the implementation of MPTCP to a more complex wireless network. To that end, we conduct a measurement-driven study of MPTCP enhancement schemes, providing insights and in-depth demonstrations of their performance with a comprehensive real-world platform. Our finding indicates that the enhancement schemes struggle to consistently maintain high performance at all times. One can achieve optimal efficiency in its specific scenarios, but suffers extreme degradation at times. To eliminate this transmission uncertainty in wireless networks, we further propose an adaptive control framework OLSch to integrate different schemes, emphasizing their strengths to provide consistently high performance. To be specific, OLSch is implemented with different scheduling schemes and leverages an online-learning-driven approach to choose one that best fits the current network conditions. Evaluations show that OLSch obviously improves the stability of transmission in harsh network scenarios, eliminates performance degradation, and increases the 95% tail throughput by 1.45×-2.39×.

Congestion control (CC) is crucial in data center networks (DCNs), providing high throughput and low latency transmission services for diverse applications. Existing CC schemes typically rely on iterative rate adjustment at ends, and suffer from performance issues such as slow convergence, throughput fluctuations, and fairness defects. Explicit rate control (ERC) promises to address these challenges by allowing switches to directly allocate rates for each flow, freeing senders from heuristic detection of available bandwidth. However, current ERC-based schemes employ inefficient feedback control to regulate the allocated rates, resulting in sub-optimal performance. In this paper, we propose RGuide, a fast and accurate CC scheme based on ERC. RGuide can calculate accurate fair share rates in real-time at switches with the consideration of low latency, and utilize the rate to guide host adjustments instead of the need for end-to-end iteration processes. We meticulously design the ERC trigger conditions, enabling switches to recognize the different congestion states of flows and rectify flows that deviate from the fair share rate at sub-RTT timescales. We conduct actual testbed experiments and extensive simulations to evaluate RGuide comprehensively. The results demonstrate the significant advantages of RGuide in terms of convergence speed, throughput stability, and fairness.

Real-time video streaming requires high throughput and low delivery time for enhanced user’s Quality of Experience (QoE). This motivates the use of multi-path transmission to improve performance. However, ensuring target performance within specified deadlines and priorities for video frames is particularly crucial for real-time communication and video quality, especially in scenarios with limited resources. To address this challenge, we propose a novel framework,vStreamPth, to guarantee high-quality real-time video streaming through multi-path transmission. For essential quality assurance,vStreamPthincorporates key requirement indicators that guide the transmission decisions of video frames across predefined multiple paths. In this framework, lightweight and robust decision-making is achieved through the collaboration of application-oriented and network-oriented data scheduling. Specifically, it employs robustness estimation to maintain the non-blocking delivery of frames, and further applies online fine-tuning to correct variations caused by changes in end-to-end transmission and multi-path network conditions. We implement a prototype ofvStreamPthin Linux user space and conduct a thorough evaluation. Experimental results demonstrate the absolute improvement ofvStreamPthin achieving high QoE and deadline satisfaction ratio compared to existing multi-path solutions.

A Link Flooding Attack (LFA) is a type of link-aimed Distributed Denial of Service (DDoS) attack that can overwhelm the Internet critical links to cut off connections with lots of low-rate, seemingly benign traffic. To defend against such threats, a promising solution involves mitigating the attack through load balancing. However, adaptive attacks employ two effective means to circumvent existing load balancing strategies. The first is the frequent changing of targets, known as rolling attacks. Rolling attacks exploit the delay between attack detection feedback and the mitigation of load balancing, depleting the defender’s resources. The second is the strategical selection of target links to create the worst-case scenario for load balancing algorithms. To address these challenges, we propose LinkDam. Specifically, LinkDam adopts a proactive approach by tracking and predicting potential victim links, providing defense against all targets of rolling attacks. Subsequently, we introduce a robust load balancing strategy to prevent the exploitation of selected link combinations. Additionally, LinkDam introduces a partial deployment approach, demanding a mere 40% of nodes be programmable (i.e., SDN nodes) while maintaining an acceptable 10% performance reduction from the maximum achievable. The experimental results indicate that LinkDam surpasses an 80% accuracy threshold, and exhibits a 57% higher tolerance to attack budgets compared to state-of-the-art solutions.

Link-flooding attacks (LFAs) pose a significant challenge to Internet availability by attacking critical network links with high volumes of seemingly legitimate traffic. In response, researchers have developed network topology obfuscation (NTO) to safeguard critical links. However, state-of-the-art NTO defenses are coarse-grained, leading to less efficient security and usability. In addition, once under attack, NTO schemes cannot identify the attacker’s bot and launch counter-defensive measures. To address these issues, this paper introduces SpiderNet, which employs advanced obfuscation techniques to secure critical links while using strategically created honeypot links for effective bot identification. When adversaries probe the network, SpiderNet captures their probing behavior and deliberately feeds back misinformation about honeypot links. By analyzing the attack patterns directed at these decoy targets, SpiderNet correlates them with adversarial probing activities to effectively identify the bots. Our experiments demonstrate that SpiderNet is more robust than state-of-the-art NTO schemes in terms of security and usability, while also being capable of identifying LFA bots.

Named Data Networking (NDN) has emerged as a promising network paradigm for the future Internet. It revolutionizes content retrieval by decoupling it from specific locations, thereby overcoming the limitations of traditional IP addressing and significantly enhancing data delivery efficiency. Additionally, NDN’s stateful forwarding plane for routers enables robust aggregation of identical requests, bolstering resistance against Distributed Denial of Service (DDoS) attacks. Despite these advancements, NDN remains vulnerable to the Interest Flooding Attack (IFA), wherein excessive requests from attackers can compromise transmission quality by depleting router resources. In the current landscape, researchers have proposed various strategies aimed at improving the accuracy, timeliness, and cost-effectiveness of defenses against IFA attacks, presuming stable user behavior. However, several challenges persist in effectively countering IFA attacks, including the need to ensure transmission quality throughout users’ lifecycles, eliminate attacks at their origin, and adapt to dynamic user behaviors. In response to these challenges, this paper presents the Hybrid Proactive Reactive Defense Scheme (HPR-DS). HPR-DS employs distinct proactive and reactive modules for resource management and user behavior analysis, respectively, at intermediate and edge nodes. It employs time series analysis to gauge evolving resource requirements and maintains separate resource pools for each content. Additionally, HPR-DS utilizes multidimensional data clustering to accurately identify attackers. Simulation results demonstrate the superior performance of HPR-DS in safeguarding user transmission quality throughout the entirety of their lifecycle and in enhancing detection precision in dynamic network environments.

Load balancing in data center networks (DCNs) is a crucial and complex undertaking. Multi-path TCP (MPTCP) has been proposed as a cost-effective solution that aims to distribute workloads and improve network resource utilization. However, it can escalate buffer occupancy and undermine burst tolerance, particularly in scenarios involving incast short flows. To address these limitations, we propose a novel multi-path congestion control algorithm, RateMP, to optimize bandwidth utilization efficiency while ensuring burst tolerance in DCNs. RateMP employs a hybrid window and rate control loop with coupled gradient projection adjustment, enabling fast and fine-grained bandwidth allocation and accelerating convergence. Additionally, RateMP eliminates the limitation of cwnd with under-rate pacing to protect incast and busty flows. We prove that RateMP is Lyapunov stable and asymptotically stable, and show the improvement of RateMP through a kernel-based implementation and extended large-scale simulations. RateMP keeps high bandwidth utilization, cuts RTT by 2x and reduces flow completion times (FCT) by 45% in incast scenarios compared to existing algorithms.

As multiple datacenters are established in different geographical locations, some applications run on cross-datacenter networks. In order to improve the service quality, service providers establish dedicated links between datacenters to take advantage of the high performance of RDMA. However, existing RDMA congestion control algorithms are designed for intra-datacenter networks. Due to the long distance between data-centers, cross-datacenter networks have higher latency, which leads to long feedback loop that prevents timely adjustments at the traffic source. Meanwhile, excessive congestion signals are generated due to untimely adjustments, which makes existing RDMA congestion control algorithms unable to accurately deal with congestion like in cross-datacenter networks. In addition, the long-haul link connecting datacenters has large bandwidth delay product (BDP), which brings great buffer pressure to switches. In this paper, link-segmented congestion control (LSCC) is proposed to avoid congestion through segmented link control. LSCC builds a segmented feedback loop between egress switches connecting to the long-haul link, which provides timely congestion feedback and greatly reduces the buffer pressure of switches. Evaluations based on DPDK implementation and large-scale simulation show that LSCC can reduce the average flow completion time (FCT) by 30%-65% and 31%-56% in realistic datacenter load and cross-datacenter load, respectively.

The utilization of Multi-path TCP (MPTCP) has been demonstrated to provide superior transport-layer support for data center networks (DCNs) due to its exceptional resource utilization and load-balancing capabilities. However, the substantial path diversity can make it challenging to utilize network resources to their full potential in DCNs. This paper focuses on studying the resource allocation issue of MPTCP from a resource optimization perspective. Based on theoretical analysis, we propose FMPTCP, which uses a feedback-based congestion control algorithm (FCC) and a feedback-based multi-path routing algorithm (FMP) to jointly achieve high bandwidth utilization and low round-trip time (RTT) in DCNs. The FCC algorithm utilizes probabilistic explicit congestion notification (ECN) to provide feedback on path congestion degree, and uses a gradient descent method to adjust the congestion window for optimal resource utilization and load balancing under a fixed routing topology. On the other hand, the FMP algorithm employs a hop-by-hop feedback mechanism to notify in-network congestion and path delay information, allowing for transparent multi-path routing for MPTCP flows. Our extensive simulations demonstrate that FMPTCP enables effective network resource utilization, which not only enhances overall throughput but also reduces transmission latency for DCNs.

Nowadays, Remote Direct Memory Access (RDMA) is gaining popularity in data centers for low CPU overhead, high throughput, and ultra-low latency. As one of the state-of-the-art RDMA Congestion Control (CC) mechanisms, HPCC leverages the In-band Network Telemetry (INT) features to achieve accurate control and significantly shortens the Flow Completion Time (FCT) for short flows. However, there exists redundant INT information increasing the processing latency at switches and affecting flows’ throughput. Besides, its end-to-end feedback mechanism is not timely enough to help senders cope well with bursty traffic, and there still exists a high probability of triggering Priority-based Flow Control (PFC) pauses under large-scale incast. In this paper, we propose a Congestion-Aware (CA) control mechanism called CACC, which attempts to push CC to the theoretical low INT overhead and PFC pause delay. CACC introduces two CA algorithms to quantize switch buffer and egress port congestion, separately, along with a fine-grained window size adjustment algorithm at the sender. Specifically, the buffer CA algorithm perceives large-scale congestion that may trigger PFC pauses and provides early feedback, significantly reducing the PFC pause delay. The egress port CA algorithm perceives the link state and selectively inserts useful INT data, achieving lower queue sizes and reducing the average overhead per packet from 42 bytes to 2 bits. In our evaluation, compared with HPCC, PINT, and Bolt, CACC shortens the average and tail FCT by up to 27% and 60.1%, respectively.

With the development of datacenter networks (DCNs) towards high bandwidth and low latency, the demands of high-level datacenter applications are heading towards high performance and high reliability, which makes traffic congestion one of the most notable problems in DCNs and brings new challenges to transport protocols. Proactive transport protocols are gaining prevalence due to their ability to provide accurate feedback and precise end-to-end control, while multipath transmission is having a broader application space in the multi-path topology of large-scale DCNs. However, these advanced transport protocols aim to improve their performance by addressing some specific congestion problems, but fail to handle multiple congestion problems caused by incast, high workload and load imbalance. Their performance in terms of flow completion time (FCT), delay, robustness, and balance still has room for further improvement. In this paper, we propose ProactMP, a novel proactive multipath transport protocol for further improvement of datacenter communications. ProactMP utilizes the rich resources of parallel paths in modern DCN and spreads the load across available network paths to improve network efficiency. ProactMP deploys a credit-based bandwidth allocation strategy to achieve low delay and zero packet loss, and overcommits receiver downlinks to ensure high link utilization. We have implemented ProactMP in the Linux system. Our testbed experiments show that ProactMP outperforms the TCP variants, MPTCP variants and a leading proactive transport protocol in FCT, link utilization, fairness and latency.

Named Data Networking (NDN) stands out as a promising Information Centric Networking architecture capable of facilitating large-scale content distribution through in-network caching and location-independent data access. However, attackers can easily inject poisoned content into the network, called content poisoning attacks, which leads to a substantial deterioration in user experience and transmission efficiency. In existing schemes, routers fail to determine the contamination source of received poisoned content, leading to the inability to accurately identify attacker nodes. Besides, attackers’ dynamic behaviors and network instability could disrupt identification results. In this paper, we propose a Secure and Lightweight scheme against content poisoning attacks based on Probing (SLP), where a proactive and reliable probing protocol is designed to identify adversaries quickly and precisely. In SLP, a router sends specifically chosen interest packets to probe a suspicious node, so that the returned corresponding content can straightly reflect its trustworthiness without other nodes’ interference. In addition, a hypothesis testing algorithm is developed to analyze the returned content, which can exclude the impact of transmission errors and adapt to dynamic attackers. Moreover, we utilize users’ feedback to avoid unnecessary probing costs on unaffected routers, with its reliability guaranteed by an efficient cuckoo-filter-based feedback validation mechanism. Security analysis shows that SLP achieves resistance against content poisoning attacks and malicious feedback. The experimental results demonstrate that SLP makes users hardly be affected by attacks and brings in only slight overhead.

Named Data Networking (NDN), with a receiver-driven connectionless communication paradigm, naturally supports content delivery from multiple sources via multiple paths. In a dynamic environment, sources and paths may change unexpectedly and are uncontrollable for consumer, which requires flexible rate control and real-time multi-path management, still lacking investigations. To address this issue, we propose an Adaptive Multi-source Multi-path Congestion Control (AMM-CC) scheme based on online learning. AMM-CC explores source/path distribution with continuous micro-experiments and abstracts the empirically experienced performance by meticulously designed two-level utility functions. Specifically, AMM-CC enables each consumer to optimize a local transmission-level utility function that fuses multi-source characteristics, including congestion level and source weights. Then, a sub-gradient descent method is designed to adjust transmission rate adaptively and achieve fine-grained control. Moreover, AMM-CC coordinates consumer with the forwarding module to ensure efficient and on-time multi-path management. It enables consumer to determine congestion gap among multiple paths by a path-level utility that sensitively captures changes and congestion on each path. Then, consumer further notifies the forwarding module in achieving precise traffic transferring. We conducted comprehensive evaluations in dynamic scenario with various content distribution using the NDN simulator, ndnSIM. The evaluation results demonstrate that AMM-CC can adapt to flexible content acquisition from multi-sources and significantly improve bandwidth utilization of multi-path compared with state-of-the-art schemes.

With potential advantages over TCP/IP for content delivery, mobility, and security, Named Data Networking (NDN) has become a promising architecture for the next-generation network. However, its poor performance in reliable transmission is still an unsolved problem. Many existing schemes in NDN employ inaccurate retransmission timeouts calculated with RTTs from diverse content sources to detect packet loss, which is lagging and may deteriorate transmission performance. Besides, after identifying the loss, the consumer costly resends the request to recover it, further increasing recovery time. In this paper, we propose an in-network Proactive Loss Recovery (PLR) scheme, which provides an efficient in-network method for timely detection and proactive recovery of lost packets. Deployed on each router, PLR detects the loss by monitoring queue status and sends high-priority explicit feedback to notify consumers of loss events timely. Meanwhile, lost packets are stored in each router's cache and will be retransmitted at an adaptive rate based on the detected remaining bandwidth. The simulation shows that PLR can vastly reduce the number of retransmissions on consumers, and the content completion time can be decreased by up to 21.8% compared with the baseline.

In data center networks (DCNs), numerous congestion control schemes utilize explicit congestion notification (ECN) to achieve low average queue delay. Such schemes generally mark packets based on the current queue length exceeding a marking threshold. However, due to the delay of ECN feedback, the queue length may further increase before the congestion notification is delivered to senders, which may lead to uncontrollable maximum queue length when bursts occur. In this paper, we propose an early ECN marking scheme based on prediction, E-ECN, to control the maximum queue length in DCNs. E-ECN uses predicted queue length rather than the current to indicate congestion with an advance time which offsets the hysteresis of ECN. We theoretically and experimentally demonstrate that early marking does not impact the throughput with appropriate selection of the advance time, and we provide guidelines for the selection in DCNs. Our simulation results show that E-ECN achieves shorter average queue delay and controllable maximum queue length in general with a bandwidth utilization guarantee. E-ECN greatly reduces queue overflow and improves the robustness of DCNs.

With Remote Direct Memory Access (RDMA) extended to commercial Ethernet, modern Data Center Networks (DCNs) carry both traditional TCP and RDMA, to support diversified application requirements. RDMA flows are guaranteed lossless transmission through Priority-based Flow Control (PFC), while TCP flows are generally lossy traffic with packet loss. However, TCP is prone to excessively occupy the shared buffer, frequently triggering PFC pause frames and overflows at switches, damaging the performance of RDMA, which expose the vulnerability of existing buffer management policies. In this paper, we propose L2BM, a buffer management algorithm for shared-memory switches to support dynamic hybrid traffic. L2BM utilizes the average occupying time of packets in each ingress queues, to perceive the congestion states timely at ingress ports, allocating the ingress pool fairly and flexibly. Based on the perception, L2BM allocates more buffer for ingress queues with faster drain and lower congestion degrees to absorb micro-burst and reduce pause frames, less buffer for long-occupied queues to prevent excessive injection. As a result, L2BM achieves low tail latency, high burst traffic absorption capacity and low buffer occupancy. Evaluations show that L2BM enable to cut the tail latency of RDMA traffic by 50% at high workloads, reduce the buffer occupancy by 40% and decrease average query delay by 57%, while ensuring few PFC pause frames and maintaining good performance of TCP flows.

The distribution of traffic shows a characteristic of different flow sizes in Data Center Networks (DCNs), which requires diverse demands for data transmission. However, most existing congestion control schemes treat all the flows equivalently and have a consistent control logic, which cannot meet the diverse demands of applications. In this paper, we propose FACC, a flow-size-aware congestion control scheme. In FACC, we design a distinguished congestion control logic to assign the transmission demands of different kinds of flows in the network. To meet the diverse demands, FACC provides an adaptable congestion window (cwnd) adjustment by assigning customized weights with a well-designed flow-size-aware reward function. Simulation results show that FACC can reduce the average FCT and the 99- th percentile FCT slowdown of short flows by 35% and 23% compared to the state-of-the-art congestion control schemes in DCNs, respectively.

As a promising implementation of Information Centric Networking, Named Data Networking (NDN) can facilitate content distribution with in-network caching and location-independent data access. However, the reliance on caches makes NDN vulnerable to content poisoning attacks, which waste network resources and decrease transmission efficiency. Most mitigating schemes follow the pattern that each content is repeatedly verified individually in each router and all producers have the same status, which wastes computation resources and degrades network performance. In this paper, we propose a Reputation-based Probabilistic Batch Verification (RPBV) scheme to address the issue, in which producers’ reputation is estimated according to verification results to distinguish different producers. We provide an adaptive probabilistic verification method based on reputation to avoid a lot of unnecessary verification operations. At the same time, we adopt an efficient batch verification algorithm to simultaneously verify multiple content, which reduces the overhead greatly. With the above mechanisms implemented only on the edge router to avoid repeated verification, we provide an optional probabilistic verification method on intermediate routers to strengthen the security. The extensive simulations show that RPBV achieves much lower computation overhead and shorter content retrieval time than the traditional schemes.

Hoh Xil is an uninhabited extremity secluded on the Tibetan Plateau hinterland. A complete mapping of ground motion variation in Hoh Xil is essential for in-depth understanding of terrain’s responses to climate change on the Tibetan Plateau. However, the inaccessibility and extremely harsh environment impeded extensive field investigations on landform alteration and its formative process. Such difficulty can be resolved by Interferometric Synthetic Aperture Radar (InSAR), which enables a broad detection of subtle permafrost motions at millimeter precision. This study, for the first time, accomplished a Multi-temporal InSAR (MT-InSAR) deformation mapping from 2015 to 2020 in Hoh Xil, with a wide coverage of about 200,000 km2. 1,592 Sentinel-1 images were processed based on the small baseline subset (SBAS) technique. The results show that Hoh Xil was experiencing dynamic permafrost disturbances. Thawing permafrost with both the linear subsidence rate higher than 2 mm/yr and the periodic amplitude over 2 mm was primarily detected in areas of flat or gentle slopes. The InSAR cumulative deformation is highly correlated with permafrost thawing depth. Significant lag times were identified between seasonal oscillation of InSAR deformation and land surface temperature (LST). Thermokarst landforms of retrogressive thaw slumps and thermokarst lakes broadly formed and dynamically evolved as a consequence of permafrost degradation. Particularly, widespread thawing permafrost characterized by the spatial clustering of thermokarst lakes appeared to occur in areas adjacent to large lakes. The discovered dynamic permafrost disturbances in Hoh Xil manifested even the secluded Tibetan Plateau hinterland was facing the threat of climate change.

Data Center Networks (DCNs) suffer from synchronized bursts for network topology and parallel applications, leading to buffer overflows at switches and increasing network delay. To overcome this problem, some congestion control algorithms like DCTCP use Explicit Congestion Notification (ECN) to notify in-network congestion and reduce switch buffer occupancy. However, the traditional Additive Increase Multiplicative Decrease (AIMD) method causes high fluctuation of round-trip time (RTT) in DCNs. Some intelligent congestion control algorithms designed for Internet can achieve great flexibility, but are not applicable in DCNs for a lack of accurate congestion feedback. In this paper, we analyze the deficiencies of utilizing RTT as congestion signals and the applicability of learning algorithms in DCNs. Then, we propose DECC, a smart TCP congestion control algorithm for DCNs, which combines Deep Reinforcement Learning (DRL) with ECN to achieve high bandwidth utilization as well as low queuing delay. DECC fully utilizes precise in-network feedback and formulates several QoS requirements to a multi-objective function. Meanwhile, it decouples cwnd adjustment with DRL decision making to gradually learn the optimal congestion control policy in real-time. We evaluate the performance of DECC in various scenarios. Simulation results show that DECC can reduce the queue length at bottleneck switches by more than 50% compared to DCTCP, while maintaining high bandwidth utilization and reducing Flow Completion Time (FCTs) under burst traffic.

Existing congestion control algorithms for MPTCP that care about only long flow transmission aim at the Congestion-Avoidance (CA) phase and they need a long time to reach convergence states. We verified that the exponential growth of congestion window (cwnd) in the uncoupled Slow-Start (SS) leads to not only unfairness to TCP but also buffer overflow due to burst data. Moreover, these algorithms cannot support fair bandwidth sharing among TCP/MPTCP flows before reaching convergence at the bottleneck, which may reduce the transmission efficiency of short flows and even hurts long flows. In this paper, we propose a Throughput Consistency Congestion Control (TCCC) algorithm consisting of Coupled Slow-Start (CSS) and Aggressive Congestion Avoidance (ACA). To prevent packet loss caused by excessive burst data, CSS couples the increment of subflows’ cwnd and reset the ssthresh value to safely move the flows to CA when it achieves expected throughput. Based on CSS, ACA periodically detects path states and allocates the same throughput increment as the best TCP to subflows to achieve fair bandwidth share in CA. Finally, we implement TCCC in both NS3 and real testbed. The results show that TCCC reduces retransmissions, improves transmission efficiency, and maintains better fairness.

The upgrade of network devices to be equipped with multiple network interfaces makes it possible to improve network throughput performance through multipath transmission protocols, especially multipath TCP (MPTCP). However, so far the mostly used MPTCP protocols have a common limitation, namely the rigid and conservative method. They have been designed with little consideration of the fact that real networks are dynamic and the network status changes frequently, thus leading to the poor performance of current MPTCP in many realistic scenarios. In this paper, we propose a lightweight multipath congestion control algorithm based on online learning, named MP-OL. MP-OL models congestion control as a multi-armed bandit problem, and adjusts the sending rate of each subflow flexibly and adaptively through online learning. Therefore, MP-OL possesses the capability of suiting various network scenarios, and can achieve fairness and high performance in dynamic network environment. It can also flexibly switch between online learning and traditional method, which reduces the computational complexity while ensuring the learning efficiency, thus making MP-OL easy to deploy and use. As the experimental results demonstrated, compared with the leading MPTCP variants, MP-OL achieves significant improvements in fairness and link utilization, and shows better resilience to non-congestion loss and better adaptability to unstable network conditions. In real networks, MP-OL also obtains better throughput performance.

Multipath TCP is designed to utilize multiple network paths to achieve improved throughput and robustness against network failure. These features are supposed to make MPTCP preferable to single-path TCP in mobile networks. However, it fails to achieve the expected performance in practice. A key challenge of using MPTCP in mobile networks is how to effectively spread packets over heterogeneous and unstable network paths to mobile devices with limited buffers. If packets are not sent in an effective way, MPTCP may only provide equal or even lower throughput than single-path TCP. Several packet scheduling algorithms have been designed to tackle this challenge. Unfortunately, they still cannot achieve the expected performance in dynamic scenarios such as mobile networks. In this paper, we propose an Online-Learning Assisted Packet Scheduler (OLAPS) to solve the packet scheduling problem by modeling it as a multi-armed bandit problem. Over time, OLAPS can adaptively learn from current network conditions to make the best scheduling policy to provide the highest possible throughput in a dynamic environment. Moreover, when the inbuilt reward monitor detects the mismatch between network conditions and the learned policy, OLAPS aborts the outdated policy and switches to a new one swiftly. We implement OLAPS as a Linux kernel module and evaluate it over a wide range of ns-3 -simulated network conditions. The results show that OLAPS retains MPTCP’s ability to provide higher throughput and also significantly improves the throughput performance of MPTCP when other in-kernel schedulers suffer a dramatic throughput decline.

Multipath TCP (MPTCP) improves the bandwidth utilization in wireless network scenarios, since it can simultaneously utilize multiple interfaces for data transmission. However, with the fast growth of mobile devices and applications, link interruptions caused by handoffs still lead to drastic performance degradation in such scenarios. Typically, a series of packet losses on part of the links will block the transmission of the entire connection when handoff occurs. This paper proposes an Experience-driven Adaptive Redundant packet scheduler (EdAR) for MPTCP, aiming at achieving seamless handoffs in mobile networks. EdAR enables flexibly scheduling redundant packets with an experience-driven learning-based approach in the face of drastic network environment changes for multipath performance enhancement. To enable accurate learning and prediction, both the network environment and the best course of actions are jointly learned via a Deep Reinforcement Learning (DRL) agent, which we design with a hybrid structure to deal with the complexity of system states. Furthermore, both offline and online learning are utilized to allow the agent to adapt to different and changing network environments. Evaluation results show that EdAR outperforms the state-of-the-art MPTCP schedulers in most network scenarios. Specifically in mobile networks with frequent handoffs, EdAR brings$2\times $improvement in terms of the overall goodput.

A QUIC (Quick UDP Internet Connections) protocol is designed to improve Hypertext Transfer Protocol (HTTP) traffic and carries a non-negligible portion of the traffic in the current Internet. As its extension, Multipath QUIC (MPQUIC) provides higher bandwidth and smoother network handover by using multiple network interfaces simultaneously. However, to improve HTTP traffic, there are still some issues not yet carefully addressed in the existing MPQUIC, and packet scheduling is a vital one among the issues. Specifically, existing methods fail to respond to the stream prioritization of HTTP Version 2 (HTTP/2), leading to unsatisfying web page load performance. Besides, managing asymmetric and dynamic network paths is also a challenging issue, which may result in Head-of-Line (HoL) blocking and excessive buffer usage if not effectively handled. In this paper, we present a stream-aware per-packet scheduler, HoL Blocking Eliminating Scheduler (HBES), to improve the performance of MPQUIC in mobile networks. Firstly, HBES provides a fair allocation of aggregated bandwidth for different streams based on their priority. Then, it keeps stream data arriving at the receiver in order by estimating packet arrival time to mitigate HoL blocking and excessive buffer usage. We implement HBES and evaluate its performance in various network scenarios. Experimental results verify the superiority of HBES in reducing stream completion time and buffer occupation over those existing MPQUIC schedulers.

As a promising implementation of Information-Centric Networking (ICN), Named Data Networking (NDN) has potential advantages over the TCP/IP network in content distribution, mobility support, etc. However, the research on NDN is still in its infancy, and congestion control, NDN’s most important functional element, poses many challenges, such as congestion detection, excessive window reduction for non-congested paths, and unfairness. In this paper, we propose an Intelligent Edge-Aided Congestion Control (IEACC) scheme for the NDN network based on Deep Reinforcement Learning (DRL). The proposed IEACC provides a proactive congestion detector that utilizes intermediate routers to transmit accurate congestion information along the path to consumers through data packets. Furthermore, considering the multi-source transmission in NDN, IEACC divides data packets into different congestion degrees by a lightweight clustering algorithm and provides suitable inputs for DRL, thereby obtaining a reasonable transmission rate. Then, it distributes the estimated bandwidth resources to consumers with transmission needs to maintain fairness. Finally, we implement our proposed scheme in the simulation platform and evaluate the performance in different scenarios. The results show that it can improve data transmission rate, reduce packet loss, and maintain fairness compared with others.

360-degree virtual reality videos greatly improve the video experience by providing users with a more immersive and interactive environment than standard streaming video. However, 360-degree videos suffer from bandwidth limits. Existing bandwidth-efficient solutions mainly focus on spatially cutting 360-degree video into tiles, and only provide video content in the Field-of-View (FoV) of users with high quality to reduce bandwidth consumption. Although existing tile-based schemes can reduce the bandwidth consumption, the bandwidth and transmission delay provided by a single-path TCP may still not meet the high requirements of 360-degree videos. Multipath TCP (MPTCP) allows a TCP connection to operate across multiple paths simultaneously and becomes highly attractive to support the mobile devices with various radio interfaces to aggregate multipath bandwidth and improve the throughput. In this paper, by taking the advantage of MPTCP, we propose an MPTCP-based adaptive streaming framework for 360-degree Virtual Reality videos, named MP-VR. MP-VR dynamically selects the appropriate tile bitrate according to the bandwidth and transmission delay of different subflows. Then it schedules the video segments to subflows to improve QoE of users. We conduct experiments on a testbed in our lab and simulations on NS-3. Evaluation results show that MP-VR outperforms existing tile-based strategies when network fluctuations or errors in FoV predictions occur.

Multipath TCP (MPTCP) utilizes multiple paths for simultaneous data transmission to enhance performance. However, existing MPTCP protocols are still far from satisfactory in wireless networks because of their loss-based congestion control and the difficulty of managing multiple subflows. To overcome these problems, we redesign the coupled congestion control algorithm and scheduler to boost MPTCP in wireless heterogeneous networks. The main purpose is to promote transmission rate under lossy networks, while also provide stability when networks suffer physical link changes and asymmetric links. In this paper, inspired by Bottleneck Bandwidth and Round-trip propagation time (BBR), we first propose Coupled BBR that utilizes detected bandwidth to adjust the sending rate within an MPTCP connection. Coupled BBR provides high loss tolerance as well as balanced congestion among MPTCP subflows. Then, to further improve the performance, we propose an Adaptively Redundant and Predictive packet (AR&P) scheduler to improve adaptability and keep in-order packet delivery in highly dynamic network scenarios. Based on Linux kernel implementation and experiments in both testbed and real network scenarios, we show that the proposed scheme not only provides high throughput in wireless networks, but also improves robustness and reduces out-of-order packets in some harsh circumstances.

It is a known issue that low-latency communication is hard to achieve when using multiple network interfaces with asymmetric capacity and delay (e.g., LTE and WLAN) simultaneously. A main underlying cause of this issue is that the packets with lower sequence number are stalled on a high-latency path, thus the early arriving packets with higher sequence number become “out-of-order (OFO)” packets. These OFO packets may excessively consume receiver’s buffer, causing long reordering delay and unnecessary packet retransmission. In this paper, we present a novel design of packet scheduling for Multipath TCP (MPTCP), called OverLapped Scheduler (OLS), able to tackle the OFO-packet problem more effectively. OLS can guarantee sufficient throughput on demand of upper layer applications, and utilizes the remaining bandwidth to reduce OFO-packets. To do so, OLS schedules packets according to their arrival time and sends a controlled number of redundant packets to avoid the impact of inaccurate arrival-time estimations due to network jitter. We implement OLS in a Linux kernel, and the experiments show that in asymmetric networks with or without jitter, OLS can effectively reduce OFO-packets and transmission latency while maintaining a sufficient throughput, which makes it fully capable to meet the requirements of applications such as live video streaming.

Nowadays, mobile devices are equipped with multiple interfaces connected to various networks, which makes it possible to aggregate bandwidth in actual application. Multipath TCP (MPTCP) is one of the transport protocols that uses multiple interfaces simultaneously and provides robust and efficient data transmission. In practice, MPTCP will interact with various network facilities. Among them, Content Delivery Network (CDN) is a popular one, which is a widely distributed network system deployed across the Internet. Using MPTCP in CDN could provide better performance for users, however, we find that CDN may not give full play to its functions when working with MPTCP. Because the Default Server Selection (DSS) mechanism in CDN only obtains servers optimal in single path connection scenarios, it may not provide the globally optimal server for MPTCP. In this paper, we propose a new algorithm called Server Selection for MPTCP (SSMP), which utilizes all available multi-homed sources to provide the globally optimal performance. SSMP modifies the DNS mechanism to return the optimal server for each available interface by the origin strategy and further selects the globally optimal server for both elephant and mice flows. We compare SSMP with DSS through experiments under video streaming and file download scenarios with both stable and variable environments. Our results show that SSMP consistently utilizes available paths more efficiently than DSS, particularly for servers with a great gap in server quality.

Permafrost in the Qinghai-Tibet Plateau (QTP) has seasonal dynamic changes and annual degradation trends that could affect the stability of artificial infrastructures such as the Qinghai-Tibet Railway (QTR). In this paper, the seasonal surface deformation along the QTR in Wudaoliang permafrost area is detected by the StaMPS-InSAR method using long-term Sentinel-1 images from March 2017 to June 2018, and the spatiotemporal variation characteristics of the surface deformation are further analyzed. The beginning of thawing or freezing stages are different between the northeastern section and southwestern section of the QTR in the study area. The results with distinct seasonal trend indicate that the maximum thawing subsidence of the QTR, at about 12.8 mm, happened in earlier September 2017. In addition, the maximum freezing uplift happened in January next year and may reach to about 7.2 mm. Also, the difference between maximum thawing subsidence and freezing uplift could show the annual degradation trend. This application demonstrates InSAR is a promising tool for monitoring surface deformation over wide permafrost areas.

In order to be TCP-friendly, the original Multipath TCP (MPTCP) congestion control algorithm is always restricted to gain no better throughput than a traditional single-path TCP on the best path. However, it is unable to maximize the throughput over all available paths when they do not go through a shared bottleneck. Also, bottleneck fairness based solutions detect the bottleneck and conduct different congestion control algorithms at different bottleneck sets to increase throughput while remaining fair to single TCP. However, existing solutions generally detect shared bottlenecks through delay correlation and loss correlation between two flows, which often lead to misjudgement in dynamic and complex network scenarios. Therefore, in this paper, we first propose a new Shared Bottleneck based Congestion Control scheme, called SB-CC, which leverages ECN (Explicit Congestion Notification) mechanism to detect shared bottlenecks among subflows and estimate the congestion degree of each subflow. Then, with the congestion degree, SB-CC balances the loads among all subflows, and smooths out congestion window fluctuation. Also, in order to prevent throughput degradation due to out-of-order packets, we propose a Shared Bottleneck based Forward Prediction packet Scheduling scheme, called SB-FPS. SB-FPS distributes data according to the window size changes of each subflow, and thus could more accurately schedule data in shared bottleneck scenarios. We implement our proposed scheme in the Linux kernel and simulation platform to evaluate the performance in different scenarios. Measurement results indicate that our scheme can detect the bottleneck more accurately and improve the overall network performance while still keeping bottleneck fairness.

For video streaming, Adaptive BitRate (ABR) algorithms are usually used to improve end-to-end user’s Quality of Experience (QoE). Many of the state-of-the-art ABR algorithms are based on simplified models, leading to conservative predictions of real situations. To optimize the QoE in complex end-to-end transmission environments, ABR algorithms based on Deep Reinforcement Learning (DRL) has shown a great improvement compared to traditional algorithms. However, the slow convergence of existing DRL-based ABR algorithms limits the QoE performance under dynamic video streaming environments. In this paper, we propose Fastconv, a novel DRL-based ABR algorithm that has a fast convergence speed to ensure a satisfactory QoE performance. Our work can be mainly divided into two parts. First, we preprocess the input data with large fluctuation in order to obtain the steady input and reduce the indeterminacy of convergence. Second, in order to reduce the structural complexity of the neural network itself and the number of parameters, we propose a neural network architecture based on multiplexed convolution kernel. Experiment results based on a real traced mobile dataset have demonstrated that Fastconv outperforms both the traditional and DRL-based ABR algorithms in terms of QoE.

Information Centric Networking (ICN), a new paradigm of Internet infrastructure, aims to better accommodate users' rapid growing demand for content delivery and optimize bandwidth utilization. Although the in-network cache feature of ICN facilitates the dissemination of content to users, it also poses new challenges on access control for content and network resource. Moreover, it is common that the access privilege of content dynamically change over time. However, existing access control mechanisms in ICN cannot support the publication and distribution of such time-sensitive content. In this paper, we propose a time- sensitive, lightweight, and secure access control mechanism, called TSLS, to solve this problem. We introduce broadcast encryption combined with time tokens for content providers to protect content confidentiality, and only authorized users satisfying the time limitation have capability to decrypt and access the content. Besides, a fast lightweight challenge-response verification is implemented at the edge routers to block unauthorized request from injecting into the network. The responses of authorized users are forwarded to content providers for pre-distribute popular content at in-network caches in advance. Our security analysis shows that TSLS possesses the properties of data confidentiality, unforgeability, anonymity, and DoS/DDoS attacks resistance. Our simulation results indicate that our proposed TSLS is an efficient mechanism with low computation cost and network delay.

Multipath TCP (MPTCP) is designed to offer higher throughput than single-path TCP, and meanwhile MPTCP flow is fair to concurrent TCP flows at the bottleneck. Although the coupled congestion control in current MPTCP can achieve the goals by coupling different subflows, it only focuses on Congestion Avoidance but each subflow still behaves like an independent TCP flow in Slow-Start. However, during Slow-Start, MPTCP is unfair to concurrent TCP flows as it uses more network resources at the shared bottleneck than single-path TCP. Worse still, since the exponential growth of multiple concurrent subflows' congestion windows often results in serious buffer overflow and packet loss at the shared bottleneck, the latency of short flows using MPTCP is often not as good as using TCP. This leads to the fact that MPTCP cannot satisfy the above design goals when handling short flows. To address this issue, we present a Coupled Slow-Start (CSS) Algorithm in this paper. CSS links the exponential growth of subflows' congestion windows to ensure the fairness and reduce the burstiness of Slow- Start. To reduce the packet loss, CSS resets the Slow-Start Threshold (ssthresh) of different subflows for MPTCP to safely move to Congestion Avoidance when it achieves its expected throughput. Our simulation shows that CSS can lower short flows' latency of up to 45% and significantly reduce the packet loss in two typical network environments, meanwhile CSS is TCP-friendly at the shared bottleneck. Simulation results also indicate that CSS can perform at least as well as original MPTCP for the bulk data transfer in common network environments.