R.1 Méthodologie de recherche. . . R-6 R.2 Économie d’énergie et délais des trames (Du et Dumax) US en fonction de la charge
du trafic US ρu. . . R-9 R.3 Architecture de référence proposée pour la planification des flux de travail dans un
environnement de cloud multi-locataire. . . R-11 R.4 Illustration de l’évaluation du mobile-edge et de la cloudification des réseaux 5G. . . R-14 R.5 a) Efficacité du temps de réponse vs. Taille du paquet de déchargement; b) Délai
moyen des paquets du trafic coexistant entre H2H et MEC; c) Consommation d’énergie contre la charge du trafic H2H pour les scénarios MEC et non-MEC; d) Durée de vie de la batterie des périphériques vs. Charge du trafic déchargé. (TL: Charge de trafic
H2H). . . R-16 R.6 Architecture des infrastructures de communication CM-FiWi avec partage de back-
haul en fibre. . . R-20 R.7 Impact de la charge de trafic H2H sur les performances de retard de paquets MEC
dans un scénario sauts multiples. . . R-21 R.8 Probabilité de connectivité FiWi de STAk contre la probabilité d’échec de backhaul
optique p avec différents schémas de survie. . . R-21 R.9 Taxonomie des technologies et techniques de ’internet tactile. . . R-24 R.10 Rendement moyen du temps de réponse du déchargement vs. La taille du paquet
déchargé. . . R-25 1.1 Typical topology of TDM-PON system. . . 2 1.2 Evolution and capacity trend of PONs (different options of downstream and upstream
capacity shown for the IEEE 802.3ca NG-EPON is still under discussion). . . 3 1.3 Schematic block diagram of RoF system (E-O: Electrical-to-Optical; O-E: Optical-
to-Electrical; BPF: Band-pass filter; SMF: Standard single mode fiber). . . 6 1.4 The evolution of wired and wireless technologies with projections. Source: M.
Décina [69], based on data by Bell Labs, G. Fettweis and S. Alamouti [28]. . . 7 1.5 Overview of cloud computing: Characteristics, challenges, service delivery, and de-
ployment models. . . 11 1.6 Research methodology. . . 18 2.1 FiWi network architecture with EPON backhaul and WLAN-based front-end. . . 33 2.2 Illustration of ECO-FiWi for FiWi access networks. . . 35 2.3 M/G/1 queueing model of US scheduling: (a) Illustration of US scheduling; (b)
2.4 Frame arrival and its reservation-and-vacation time: (a) During ONU1’s data inter- val; (b) During ONU1’s reservation interval; (c) During ONUj’s data interval; (d) During ONUj’s reservation interval; and (e) During ONU-AP’s sleep interval. . . . . 47 2.5 M/G/1 queueing model of DS scheduling: (a) Illustration of data and reservation
intervals; (b) Illustration of reservation time difference between the conventional polling system (upper part) and DS scheduling (lower part). . . 48 2.6 Illustration of upper bounds on US and DS frame delays in ECO-FiWi networks. . . 51 2.7 Potential energy saving with varying cycle time Tc: Integrated vs. Conventional
ONU (η vs. ηnosta). . . 53
2.8 Polling cycle time Tc as function of ρu, RT T , and Tg. . . 53 2.9 Energy saving η and US frame delays (Du and Dmaxu ) as function of US traffic load ρu. 54
2.10 Energy saving (η) and US frame delays (Du and Dmaxu ) as function of PON RT T . . 56
2.11 Energy saving (η) and US frame delays (Du and Dmaxu ) as function of number of STAs per ONU-AP (M ). . . . 56 2.12 DS-based cycle time Tcdand exhaustive service threshold for DS transmission ρthd . . . 57 2.13 DS end-to-end frame delays (Dd and Dmaxd ). . . 58
3.1 SIPHT [140] – an example of scientific workflow. . . 65 3.2 Proposed reference architecture of workflow scheduling in a multi-tenant cloud envi-
ronment. . . 69 3.3 Representation of a cloud workflow by a direct acyclic graph (DAG). . . 73 3.4 Illustration of scheduling multiple workflows. . . 80 3.5 Mean makespan vs. total number of workflow. . . 84 3.6 Cumulative distribution function (CDF) of makespan for different scheduling schemes. 85 3.7 Average execution time of scheduling schemes. . . 85 3.8 Resource utilization. . . 86 3.9 Average tardiness vs. number of workflows. . . 86 3.10 99th percentile makespan of different scheduling schemes for the SIPHT workflow
application. . . 88 3.11 99thpercentile makespan of different scheduling schemes for the CyberShake workflow
application. . . 88 4.1 Generic network architecture of cloudlet enhanced FiWi access networks for mobile-
edge computing. . . 95 4.2 Illustration of the proposed unified resource management scheme for MEC integrated
FiWi access networks. . . 97 4.3 Format of extended REPORT message. . . 98 4.4 Packet delay components of both conventional H2H and MEC traffic. . . 106 4.5 Illustration of the sleep mode at an STA: a) when tasks are executed locally at STA;
b) when tasks are offloaded onto a cloudlet. . . 112 4.6 Polling cycle time Tc vs. conventional FiWi traffic load for different guard time Tg
and number of ONU-APs. . . 117 4.7 Impact of FiWi traffic load on packet delay of both FiWi and MEC traffic. . . 118 4.8 Impact of FiWi traffic load ρh2h on cloudlet response timeRcloudlet. . . 118 4.9 Impact of polling cycle time Tc on average offload gain-overhead ratio γ. . . 119
4.10 Impact of offloaded traffic load ρmecon the average response time of cloudlet Rcloudlet with different number of ONU-APs. . . 119
4.11 Impact of offloaded traffic load with different numbers of ONU-APs on the average offload gain-overhead ratioγ. . . 120 4.12 Impact of offload packet size on the average response time efficiency δ. . . 120 4.13 Impact of offload packet size on average energy efficiencyη. . . 121 4.14 Impact of offloaded traffic load ρou on average battery life Bl with different number
of ONU-APs. . . 121 4.15 Experimental testbed setup for cloudlet enhanced FiWi networks with MEC capa-
bilities: 1) optical fiber loops; 2) passive splitter at remote node of EPON; 3) cable connecting the gateway for core network; 4) cable connecting APs; 5) WLAN access point; 6) Ethernet cable connecting an ONU; 7) Ethernet cable connecting cloudlet; 8) cloudlet server hosting OpenStack++ platform; 9) running VM instance in Open- Stack++ on the cloudlet server; 10) STA/edge device running edge application. . . . 122 4.16 Comparison of experimental and analytical results: a) Impact of offload packet size
on average response time efficiency, b) impact of offloaded traffic load on average response time of cloudlet. . . 124 5.1 Illustration of mobile edge computing and cloudification of 5G networks. . . 129 5.2 Overview of mobile edge computing: benefits, potential use cases, enabling technolo-
gies, and challenges (LWA: LTE and WiFi aggregation, LAA: licensed assisted access, LTE-U: LTE Unlicensed, LWI: LTE/WiFi interworking, CDN: content delivery net- work). . . 131 5.3 MEC over FiWi network architectures: a) MEC over Ethernet-based FiWi networks
and MEC over 4G LTE-based FiWi networks; b) coexistence of MEC and C-RAN over FiWi enhanced 4G LTE HetNets. . . 137 5.4 Illustration of TDMA-based unified resource management scheme and sleep mode for
MEC over WLAN-based FiWi networks. . . 140 5.5 a) response time efficiency vs. offload packet size; b) mean packet delay of co-existing
H2H and MEC traffic; c) energy consumption vs. H2H traffic load for MEC and non- MEC scenarios; d) battery life of edge devices vs. offloaded traffic load. (TL: H2H traffic load). . . 143 6.1 Architecture of CM-FiWi communications infrastructures with fiber backhaul sharing.150 6.2 Illustration of the proposed unified resource management scheme for CM-FiWi network.152 6.3 Components of packet delay of both conventional H2H and MEC traffic. . . 161 6.4 Components of packet delay of cloud traffic. . . 163 6.5 Illustration of optical backhaul reliability, MEC reliability, and wireless protection
by utilizing redundancy techniques. . . 168 6.6 Delay performance for single-hop scenario. . . 171 6.7 Impact of H2H traffic load on MEC packet delay performance in multi-hop scenario. 171 6.8 Delay performance as function of polling cycle time. . . 172 6.9 Response time efficiency vs. offloaded packet size. . . 172 6.10 Offload gain-overhead ratio vs. offloaded traffic load. . . 173 6.11 CCR vs. offloaded traffic load. . . 173 6.12 FiWi connectivity probability of STAk vs. optical backhaul failure probability p with
different survival schemes. . . 174 6.13 Mean end-to-end delay vs H2H traffic load with/without IF scenario. . . 174
7.1 The Tactile Internet: a) revolutionary leap of the Tactile Internet (in compliance with ITU-T Technology Watch Report [46]); b) the three lenses of IoT, 5G, and the Tactile Internet: Commonalities and differences. . . 179 7.2 Taxonomy of enabling Tactile Internet technologies and techniques. . . 183 7.3 FiWi enhanced LTE-A HetNets performance: a) Average end-to-end delay vs. ag-
gregate throughput for different WiFi offloading ratio (WOR); b) FiWi connectivity probability of a mobile user vs. EPON fiber link failure probability p. . . . 185 7.4 Generic architecture of cloudlet enhanced FiWi access networks for mobile-edge com-
puting. . . 186 7.5 Average offload response time efficiency vs. offload packet size. . . 187 7.6 Two-tier cloud robotics systems architecture based on M2M and M2C communications.190