Search

Machine learning has found many applications in network contexts. These include solving optimisation problems and managing network operations. Conversely, networks are essential for facilitating machine learning training and inference, whether performed centrally or in a distributed fashion. To conduct rigorous research in this area, researchers must have a comprehensive understanding of fundamental techniques, specific frameworks, and access to relevant datasets. Additionally, access to training data can serve as a benchmark or a springboard for further investigation. All these techniques are summarized in this article; serving as a primer paper and hopefully providing an efficient start for anybody doing research regarding machine learning for networks or using networks for machine learning.

5G and beyond 5G are revolutionizing cellular and ubiquitous networks with new features and capabilities. The new millimeter-wave frequency band can provide high data rates for the new generations of mobile networks but suffers from NLoS caused by obstacles, which causes packet drops that mislead TCP because the protocol interprets all drops as an indication of network congestion. The principal flaw of TCP in such networks is that the root for packet drops is not distinguishable for TCP, and the protocol takes it for granted that all losses are due to congestion. This paper presents a new TCP based on deep learning that can outperform other common TCPs in terms of throughput, RTT, and congestion window fluctuation. The primary contribution of deep learning is providing the ability to distinguish various conditions in the network. The simulation results revealed that the proposed protocol could outperform conventional TCPs such as Cubic, NewReno, Highspeed, and BBR., This research was funded in part by the Spanish MCIN/AEI/ 10.13039/501100011033 through project PID2019-106808RA-I00", and by Secretaria d’Universitats i Recerca del departament d’Empresa i Coneixement de la Generalitat de Catalunya with the grant number 2021 SGR 00330., Peer Reviewed, Postprint (published version)

5G cellular communication, especially with its hugely available bandwidth provided by millimeter-wave, is a promising technology to fulfill the coming high demand for vast data rates. These networks can support new use cases such as Vehicle to Vehicle and augmented reality due to its novel features such as network slicing along with the mmWave multi-gigabit-persecond data rate. Nevertheless, 5G cellular networks suffer from some shortcomings, especially in high frequencies because of the intermittent nature of channels when the frequency rises. Non-line of sight state is one of the significant issues that the new generation encounters. This drawback is because of the intense susceptibility of higher frequencies to blockage caused by obstacles and misalignment. This unique characteristic can impair the performance of the reliable transport layer widely deployed protocol, TCP, in attaining high throughput and low latency throughout a fair network. As a result, the protocol needs to adjust the congestion window size based on the current situation of the network. However, TCP cannot adjust its congestion window efficiently, which leads to throughput degradation of the protocol. This thesis presents a comprehensive analysis of reliable end-to-end communications in 5G networks and analyzes TCP’s behavior in one of the 3GPP’s well-known s cenarios called urban deployment. Furtherm ore, two novel TCPs bas ed on artificial intelligence have been proposed to deal with this issue. The first protocol uses Fuzzy logic, a subset of artificial intelligence, and the second one is based on deep learning. The extensively conducted simulations showed that the newly proposed protocols could attain higher performance than common TCPs, such as BBR, HighSpeed, Cubic, and NewReno in terms of throughput, RTT, and sending rate adjustment in the urban scenario. The new protocols' superiority is achieved by employing smartness in the conges tions control mechanism of TCP, which is a powerful e, La comunicació cel·lular 5G, especialment amb l’amplada de banda molt disponible que proporciona l’ona mil·limètrica, és una tecnologia prometedora per satisfer l’elevada demanda de grans velocitats de dades. Aquestes xarxes poden admetre casos d’ús nous, com ara Vehicle to Vehicle i realitat augmentada, a causa de les seves novetats, com ara el tall de xarxa juntament amb la velocitat de dades mWave de multi-gigabit per segon. Tot i això, les xarxes cel·lulars 5G pateixen algunes deficiències, sobretot en freqüències altes a causa de la naturalesa intermitent dels canals quan augmenta la freqüència. L’estat de no visió és un dels problemes significatius que troba la nova generació. Aquest inconvenient es deu a la intensa susceptibilitat de freqüències més altes al bloqueig causat per obstacles i desalineació. Aquesta característica única pot perjudicar el rendiment del protocol TCP, àmpliament desplegat, de capa de transport fiable en aconseguir un alt rendiment i una latència baixa en tota una xarxa justa. Com a resultat, el protocol ha d’ajustar la mida de la finestra de congestió en funció de la situació actual de la xarxa. Tot i això, TCP no pot ajustar la seva finestra de congestió de manera eficient, cosa que provoca una degradació del rendiment del protocol. Aquesta tesi presenta una anàlisi completa de comunicacions extrem a extrem en xarxes 5G i analitza el comportament de TCP en un dels escenaris coneguts del 3GPP anomenat desplegament urbà. A més, s'han proposat dos TCP nous basats en intel·ligència artificial per tractar aquest tema. El primer protocol utilitza la lògica Fuzzy, un subconjunt d’intel·ligència artificial, i el segon es basa en l’aprenentatge profund. Les simulacions àmpliament realitzades van mostrar que els protocols proposats recentment podrien assolir un rendiment superior als TCP habituals, com ara BBR, HighSpeed, Cubic i NewReno, en termes de rendiment, RTT i ajust d’índex d’enviament en l’escenari urbà. La superioritat dels nous pro, Postprint (published version)

This is a post-peer-review, pre-copyedit version of an article published in Optical review. The final authenticated version is available online at: http://dx.doi.org/10.1007/s10043-021-00652-w., The demand for high bandwidth on the Internet is growing drastically, and one of the solutions for tackling this problem is using optical networks. Burst switching is one of the techniques that can be used in optical networks to handle high traffic. Aside from the many advantages that this technique has, it suffers from a big flaw called burst contention. Optical burst switching (OBS) is a switching technique without any buffering system. As a result, when two bursts are trying to reserve one resource, one of them drops. This drawback can have a significant impact on the performance of some protocols like TCP because they have not been designed to perform in a network without any queuing system and cannot distinguish a drop is because of the congestion or contention. In this paper, a new algorithm called AVGR (Average of RTTs) is proposed based on some mathematical equations to prevent the degradation of TCP. It tries to calculate averages for some RTTs in three different periods. Then base on the obtained results, the congestion control mechanism will be modified. The primary goal of the algorithm is to determine the current status of the network and make proper decisions based on it., This work has been funded by the Secretaria d’Universitats i Recerca del Departament d’Empresa i Coneixement de la Generalitat de Catalunya (2017 SGR 376) and the Spanish Government under Project PID2019-106808RA-I00 AEI/FEDER UE., Peer Reviewed, Postprint (author's final draft)

© 2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes,creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works., 5G era opens a new horizon toward communication with new features and capabilities. The new mobile generation comes up with a multi-gigabit per second data rate along with its huge available bandwidths provided by millimeter-wave frequency bands. Dispensing a fully connected world throughout low latency is the supreme aim for 5G networks. However, attaining high-speed, reliable communication is challenging in the new generation, especially in scenarios with numerous obstacles such as the urban one. As the frequency rises, the signal’s penetration power declines, which can mislead TCP in adjusting the sending rate because TCP cannot distinguish that a packet drop in a network is due to congestion or other shortcomings of the network such as blockage or random packet drops. This paper proposes a new TCP based on Fuzzy logic, which strives to prevent performance reduction in urban deployments. The Fuzzy rules are implemented in the congestion avoidance phase of the new protocol to adjust the sending rate intelligently and prevent blockage impacts. The ultimate aim of the protocol is to control the sending rate based on the current situation of the network so it can attain the highest possible performance. Moreover, it tries to reach its goal through low latency and keep the average sending rate as small as possible to restrain the buffer exhaustion. The extensive conducted simulations showed that the newly proposed protocol could attain higher performance compared to BBR, HighSpeed, Cubic, and NewReno in terms of throughput, RTT, and sending rate adjustment in the urban scenario., This work was supported in part by the Secretaria d’Universitats i Recerca del Departament d’Empresa i Coneixement de la Generalitat de Catalunya under Grant 2017 SGR 376, and in part by the Spanish Government under Project PID2019-106808RA-I00 AEI/FEDER UE., Peer Reviewed, Postprint (published version)

© 2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes,creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works., By deploying the millimeter-wave wide spectrum in 5G networks, the new generation is capable of providing high data rates with low latencies. However, these frequencies have intermittent characteristics as their downside, which acts as a hurdle on the way of attaining high performances. This disadvantage can lower signals’ penetration power in reaching far distances or passing materials such as vehicles, walls, and even human bodies. As a result, having a reliable end-to-end connection throughout 5G millimeter-wave networks can be challenging because this burden is on the transport layer mostly exploited protocol, TCP, which is unable to perform sufficiently due to the fluctuation of the high-frequency channels. This paper aims to analyze TCP’s behavior in one of the 3GPP’s well-known scenarios called urban deployment. The detailed investigation of TCP over 5G millimeter-wave when used in a city and the impact of different parameters such as remote servers, RLC buffer size, different congestion control algorithms, and maximum segment size are discussed thoroughly throughout the paper. The results revealed that TCP could benefit from the edge server deployment due to the shorter control loop, and increasing maximum segment size can also enhance this superiority. Moreover, individual TCP variants react to various RLC buffer sizes differently. However, in general, increased throughput can be attained by deploying larger buffers at the cost of latency., This work was supported in part by the Secretaria d’Universitats i Recerca del Departament d’Empresa i Coneixement de la Generalitat de Catalunya under Grant 2017 SGR 376, and in part by the Spanish Government under Project PID2019-106808RA-I00 AEI/FEDER UE., Peer Reviewed, Postprint (published version)

The 5G (fifth-generation) mobile networks, especially by exploiting higher bandwidth in the mmWave (millimeter wave) spectrum, is the leading candidate to be used as the coming generation for ubiquitous networks. The vast available bandwidth in mmWave can satisfy the high data rate and low latency expectations from 5G networks in order to provide new services and use cases. Although 5G mmWave networks come up with innovative and robust services, they suffer from a drawback. As the frequency rises, the penetration power and coverage area of the network decreases, so it results in having discontinuous communication between a base station and a user. This intermittent characteristic is caused due to an existing obstacle such as a car or a building on the communication path that can hurdle the establishment of a transmission, which is called NLoS (Non-Line of Sight) state. NLoS states can degrade the functionality of the network and prevent from having seamless connectivity by forcing fluctuations in the network's channels. The reason for this shortcoming is because of the susceptibility of high frequencies to the blockage that can be generated by obstacles. The intense negative effect of having a blockage in the network is on an end-to-end communication when other layers protocols such as the transport layer widely used protocol TCP (Transmission Control Protocol) are used. Having frequent disconnections in the network impairs the TCP's functionality with inducing congestion states and preventing it from achieving higher performance. In this paper, we present the performance evaluation and analysis of TCP in different situations in an urban area and find out how various conditions can affect the performance of the protocol. The simulation results indicate that conventional TCPs are not adequate enough to be exploited in 5G mmWave networks. For having them functioning in their full potential, some modifications should be made in order to adapt them to 5G mmWave networks, Peer Reviewed, Postprint (author's final draft)

5G cellular communication, especially with its hugely available bandwidth provided by millimeter-wave, is a promising technology to fulfill the coming high demand for vast data rates. These networks can support new use cases such as Vehicle to Vehicle and augmented reality due to its novel features such as network slicing along with the mmWave multi-gigabit-per-second data rate. Nevertheless, 5G cellular networks suffer from some shortcomings, especially in high frequencies because of the intermittent nature of channels when the frequency rises. Non-line of sight state, is one of the significant issues that the new generation encounters. This drawback is because of the intense susceptibility of higher frequencies to blockage caused by obstacles and misalignment. This unique characteristic can impair the performance of the reliable transport layer widely deployed protocol, TCP, in attaining high throughput and low latency throughout a fair network. As a result, the protocol needs to adjust the congestion window size based on the current situation of the network. However, TCP is not able to adjust its congestion window efficiently, and it leads to throughput degradation of the protocol. This paper presents a comprehensive analysis of reliable end-to-end communications in 5G networks. It provides the analysis of the effects of TCP in 5G mmWave networks, the discussion of TCP mechanisms and parameters involved in the performance over 5G networks, and a survey of current challenges, solutions, and proposals. Finally, a feasibility analysis proposal of machine learning-based approaches to improve reliable end-to-end communications in 5G networks is presented., This work was supported by the Secretaria d’Universitats i Recerca del Departament d’Empresa i Coneixement de la Generalitat de Catalunya under Grant 2017 SGR 376., Peer Reviewed, Postprint (published version)