We study the performance of the Message Queuing Telemetry Transport Protocol (MQTT) over QUIC. QUIC has been recently proposed as a new transport protocol, and it is gaining relevance at a very fast pace, favored by the support of key players, such as Google. It overcomes some of the limitations of the more widespread alternative, TCP, especially regarding the overhead of connection establishment. However, its use for Internet of Things (IoT) scenarios is still under consideration. In this paper we integrate a GO-based implementation of the QUIC protocol with MQTT, and we compare the performance of this combination with that exhibited by the more traditional MQTT/TLS/TCP approach. We use Linux Containers and we emulate various wireless network technologies by means of the ns3 simulator. The results of an extensive measurement campaign, show that QUIC protocol can indeed yield good performances for typical IoT use cases.

QUIC, fostered by Google and under standardization in the IETF, integrates some of HTTP/s, TLS, and TCP functionalities over UDP. One of its main goals is to facilitate transport protocol design, with fast evolution and innovation. However, congestion control in QUIC is still severely jeopardized by packet losses, despite implemented loss recovery mechanisms, whose behavior strongly depends on the Round Trip Time. In this paper, we design and implement rQUIC, a framework that enables FEC within QUIC protocol to improve its performance over wireless networks. The main idea behind rQUIC is to reduce QUIC's loss recovery time by making it robust to erasures over wireless networks, as compared to traditional transport protocol loss detection and recovery mechanisms. We evaluate the performance of our solution by means of extensive simulations over different type of wireless networks and for different applications. For LTE and Wifi networks, our results illustrate significant gains of up to 60% and 25% savings in the completion time for bulk transfer and web browsing, respectively.

Quick UDP Internet Connections (QUIC) is an experimental transport protocol designed to primarily reduce connection establishment and transport latency, as well as to improve security standards with default end-to-end encryption in HTTPbased applications. QUIC is a multiplexed and secure transport protocol fostered by Google and its design emerged from the urgent need of innovation in the transport layer, mainly due to difficulties extending TCP and deploying new protocols. While still under standardisation, a non-negligble fraction of the Internet's traffic, more than 7% of a European Tier1-ISP, is already running over QUIC and it constitutes more than 30% of Google's egress traffic [1].

The presence of IoT in current networking scenarios is more relevant every day. IoT covers a wide range of applications, ranging from wearable devices to vehicular communications. With the consolidation of Industry 4.0, IIoT (Industrial IoT) environments are becoming more common. Communications in these scenarios are mostly wireless, and due to the lossy nature of wireless communications, the loss of information becomes an intrinsic problem. However, loss recovery schemes increase the delay that characterizes any communication. On the other hand, both reliability (robustness) and low delay are crucial requirements for some applications in IIoT. An interesting strategy to improve both of them is the use of Network Coding techniques, which have shown promising results, in terms of increasing reliability and performance. This work focuses on a possible new coding approach, based on systematic network coding scheme with overlapping generations. We perform a thorough analysis of its behavior. Based on the results, we draw out a number of conclusions for practical implementations in wireless networks, focusing our interest in IIoT environments.

We consider the transmission of packets across a lossy end-to-end network path so as to achieve low in-order delivery delay. This can be formulated as a decision problem, namely deciding whether the next packet to send should be an information packet or a coded packet. Importantly, this decision is made based on delayed feedback from the receiver. While an exact solution to this decision problem is challenging, we exploit ideas from queueing theory to derive scheduling policies based on prediction of a receiver queue length that, while suboptimal, can be efficiently implemented and offer substantially better performance than state of the art approaches. We obtain a number of useful analytic bounds that help characterise design trade-offs and our analysis highlights that the use of prediction plays a key role in achieving good performance in the presence of significant feedback delay. Our approach readily generalises to networks of paths and we illustrate this by application to multipath trans port scheduler design.

Multicast services are believed to play a relevant role in next wireless networking scenarios. In this paper we exploit Tunable Sparse Network Coding techniques to increase reliability of multicast communications. We show that the proposed network coding scheme yields a better performance than state-of-the-art solutions, which are traditionally based on retransmissions. We first use a model to analytically compare the two approaches. Then, we validate and broaden this analysis by means of an experimental campaign over a testbed deployed with Commercial Of-The-Shelf devices. This platform, comprising low cost devices (Raspberry-PI), allows us to assess the feasibility of the proposed solution, which offers a relevant gain in terms of performance.

Despite their high computational complexity, Random Linear Network Coding (RLNC) techniques have been shown to offer a good robustness against packet erasure wireless channels. Some approaches have been recently proposed to reduce such computational burden, for both encoder and decoder elements. One of those approaches are the so- called Tunable Sparse Network Coding (TSNC) techniques, which advocate limiting the number of packets that are combined to build a coded packet. They also propose dynamically adapting the corresponding sparsity level, as the transmission evolves, although an optimum tuning policy has not been yet found. In this paper we present a TSNC implementation that exploits a novel analytical model to estimate the probability of generating an innovative packet (linearly independent combination), given the current status at the decoder. Taking advantage of the model's accuracy, the proposed scheme offers a better trade-off between computational complexity and network performance. Furthermore, we broaden the analysis of TSNC techniques by thoroughly assessing their behavior over wireless networks using the ns-3 platform. The results yield a remarkable complexity reduction (approx. 3.33x less complexity), without jeopardizing network performance.

Random linear network coding has been shown to offer an efficient communication scheme, leveraging a remarkable robustness against packet losses. However, it suffers from a high-computational complexity, and some novel approaches, which follow the same idea, have been recently proposed. One of such solutions is sparse network coding (SNC), where only few packets are combined with each transmission. The amount of data packets to be combined can be set from a density parameter/distribution, which could be eventually adapted. In this paper, we present a semi-analytical model that captures the performance of SNC on an accurate way. We exploit an absorbing Markov process, where the states are defined by the number of useful packets received by the decoder, i.e., the decoding matrix rank, and the number of non-zero columns at such matrix. The model is validated by the means of a thorough simulation campaign, and the difference between model and simulation is negligible. We also include in the comparison of some more general bounds that have been recently used, showing that their accuracy is rather poor. The proposed model would enable a more precise assessment of the behavior of SNC techniques.

Random Linear Network Coding (RLNC) is a technique that provides several benefits. For instance, when applied over wireless mesh networks, it can be exploited to ease routing solutions as well as to increase the robustness against packet losses. Nevertheless, the complexity of the decoding process and the required overhead might jeopardize its performance. There is a trade-off when deciding the field and block sizes; larger values decrease the probability of transmitting linearly dependent packets, but they also increase both the required overhead and the decoding complexity. In order to overcome these limitations, we propose a sliding window enhancement; a fixed number of packets (fewer than the block size) is combined within every transmission, and the decoding process can therefore take advantage of the algebra with sparse matrices. The paper presents an analytical model, which is first validated and later broaden by means of an extensive simulation campaign carried out over the ns-3 simulator.

Random Linear Network Coding (RLNC) has been shown to be a technique with several benefits, in particular when applied over wireless mesh networks, since it provides robustness against packet losses. On the other hand, Tunable Sparse Network Coding (TSNC) is a promising concept, which leverages a trade-off between computational complexity and goodput. An optimal density tuning function has not been found yet, due to the lack of a closed-form expression that links density, performance and computational cost. In addition, it would be difficult to implement, due to the feedback delay. In this work we propose two novel tuning functions with a lower computational cost, which do not highly increase the overhead in terms of the transmission of linear dependent packets compared with RLNC and previous proposals. Furthermore, we also broaden previous studies of TSNC techniques, by means of an extensive simulation campaign carried out using the ns-3 simulator. This brings the possibility of assessing their performance over more realistic scenarios, e.g considering MAC effects and delays. We exploit this implementation to analyze the impact of the feedback sent by the decoder. The results, compared to RLNC, show a reduction of 3.5 times in the number of operations without jeopardizing the network performance, in terms of goodput, even when we consider the delay effect on the feedback sent by the decoder.

In this work we focus on the application of an intra-flow linear random coding scheme over wireless mesh networks. We propose a cross-layer technique to balance the load between relaying nodes, considering the quality of the wireless links. We assess the performance of our proposal by means of an extensive simulation campaign carried out over the ns-3 platform, exploiting a channel model based on a Hidden Markov Process, which accurately mimics the bursty behavior that is observed over real indoor channels. The results show the benefits of the proposed cross-layer technique. In addition, we also discuss the trade-off (between accuracy and overhead) that needs to be considered when obtaining the information that such scheme rely on. Our simulations yield a 10% performance gain when exploiting the link quality information, as it is already obtained by various routing protocols.