pr\u00e9sentent<\/p>\n\n\n\n
Autoris\u00e9 \u00e0 pr\u00e9senter ses travaux en vue de l\u2019obtention du Doctorat de l’Institut Polytechnique de Paris, pr\u00e9par\u00e9 \u00e0 T\u00e9l\u00e9com SudParis en :<\/p>\n\n\n\n
le VENDREDI 11 JUILLET 2025 \u00e0 9h30<\/p>\n\n\n\n
\u00e0<\/p>\n\n\n\n
Amphith\u00e9\u00e2tre 10
T\u00e9l\u00e9com SudParis, 9 Rue Charles Fourier, 91000 Evry-Courcouronnes FR<\/p>\n\n\n\n
Membres du jury :<\/strong><\/p>\n\n\n\n Mme Hind CASTEL<\/strong>, Full professor, Telecom Sudparis, FRANCE – Directeur de these Invit\u00e9s : <\/strong><\/p>\n\n\n\n M. Badii JOUABER, Professeur, T\u00e9l\u00e9com SudParis, Co-directeur de th\u00e8se<\/p>\n\n\n\n R\u00e9sum\u00e9 :<\/strong><\/p>\n\n\n\n Les d\u00e9ploiements modernes de r\u00e9seaux adoptent de plus en plus un mod\u00e8le multi-locataire, dans lequel les propri\u00e9taires d’infrastructures louent des ressources informatiques et r\u00e9seau \u00e0 des op\u00e9rateurs de r\u00e9seaux (NOs: Network Operators) servant divers fournisseurs de services (SPs: Service Providers). Ces SPs peuvent ex\u00e9cuter des applications critiques en termes de latence et sensibles \u00e0 la fiabilit\u00e9 (e.g., autonomous driving), ainsi que des services IoT \u00e0 grande \u00e9chelle ou des services gourmands en m\u00e9dias (e.g., video streaming). Garantir des d\u00e9lais stricts tout en minimisant la consommation d\u2019\u00e9nergie dans cet environnement h\u00e9t\u00e9rog\u00e8ne constitue un d\u00e9fi majeur, car chaque service impose des exigences vari\u00e9es sur les ressources partag\u00e9es. Pour traiter ces enjeux de mani\u00e8re coh\u00e9rente, cette th\u00e8se int\u00e8gre trois volets de recherche compl\u00e9mentaires \u2014 le placement de fonctions virtuelles, le dimensionnement des ressources, et un cadre probabiliste g\u00e9n\u00e9ral de latence \u2014 dans un cadre unifi\u00e9 d\u2019optimisation multi-locataire. Premi\u00e8rement, nous abordons le probl\u00e8me de haut niveau du placement des fonctions virtuelles dans un sc\u00e9nario de slicing \u00e0 grande \u00e9chelle, en respectant la capacit\u00e9 limit\u00e9e des ressources r\u00e9seau et informatiques. Nous d\u00e9veloppons des algorithmes d\u2019apprentissage capables de converger vers une strat\u00e9gie de d\u00e9ploiement optimale sous incertitude, en mod\u00e9lisant le placement comme un probl\u00e8me du bandit manchot. Une analyse th\u00e9orique d\u00e9montre que ces algorithmes atteignent un regret born\u00e9 et respectent les contraintes de faisabilit\u00e9 avec une probabilit\u00e9 \u00e9lev\u00e9e. Deuxi\u00e8mement, \u00e0 partir du placement obtenu, nous nous int\u00e9ressons au dimensionnement des ressources, o\u00f9 l\u2019on cherche \u00e0 optimiser l\u2019efficacit\u00e9 \u00e9nerg\u00e9tique tout en satisfaisant des contraintes strictes de latence. Nous proposons une approche originale repr\u00e9sentant chaque slice comme un r\u00e9seau de Jackson, permettant de mod\u00e9liser les interd\u00e9pendances entre ses composants. Gr\u00e2ce au Dynamic Voltage Frequency Scaling (DVFS), nous s\u00e9lectionnons les fr\u00e9quences CPU appropri\u00e9es pour r\u00e9duire la consommation d\u2019\u00e9nergie dynamique sans compromettre les exigences de latence, assurant ainsi le respect simultan\u00e9 des contraintes de capacit\u00e9 et de d\u00e9lai. Enfin, nous \u00e9tendons nos solutions \u00e0 une classe plus g\u00e9n\u00e9rale de probl\u00e8mes multi-locataires en introduisant un cadre probabiliste appel\u00e9 chance constrained multi-armed bandits (CCMAB). Cette g\u00e9n\u00e9ralisation int\u00e8gre des contraintes probabilistes dans la prise de d\u00e9cision, avec des applications telles que le offloading de t\u00e2ches dans les syst\u00e8mes distribu\u00e9s. Chaque locataire doit satisfaire des contraintes de chance sur les d\u00e9lais stochastiques tout en minimisant la consommation globale d\u2019\u00e9nergie. Nous prouvons l\u2019optimalit\u00e9 de l\u2019algorithme propos\u00e9 et validons ses performances par des simulations. En r\u00e9sum\u00e9, cette th\u00e8se propose une approche coh\u00e9rente et hi\u00e9rarchis\u00e9e combinant un placement guid\u00e9 par la capacit\u00e9 limit\u00e9e des ressources, un dimensionnement des ressources sensible \u00e0 la latence, et une prise de d\u00e9cision probabiliste dans un seul cadre destin\u00e9 aux environnements multi-locataires h\u00e9t\u00e9rog\u00e8nes. Cette int\u00e9gration offre non seulement de solides garanties th\u00e9oriques, mais aussi des algorithmes pratiques pour une utilisation efficace des ressources et une fiabilit\u00e9 assur\u00e9e face \u00e0 des exigences de services vari\u00e9es. Abstract :<\/strong><\/p>\n\n\n\n Modern network deployments increasingly adopt a multi-tenant model, in which infrastructure owners lease computing and networking resources to Network Operators (NOs) serving diverse Service Providers (SPs). These SPs may run latency-critical and reliability-sensitive applications (e.g., autonomous driving) as well as large-scale IoT or media-intensive services (e.g., video streaming). Achieving strict latency guarantees while minimizing power consumption in this heterogeneous environment poses a core challenge, given that each service imposes diversity demands on shared resources. To tackle these issues cohesively, this thesis integrates three complementary research components\u2014virtual function placement, resource dimensioning, and a general probabilistic latency framework\u2014within a unified multi-tenant optimization framework. First, we address the upper-level problem of virtual function placement in a large-scale network slicing scenario, focusing primarily on ensuring feasible capacity usage across the underlying network and computing resources. We develop learning-based algorithms capable of converging to an optimal deployment strategy under uncertainty, by modeling the placement decision as a multi-armed bandit problem. Theoretical analysis confirms that these algorithms achieve bounded regret and respect feasibility constraints with high probability. Second, building upon the chosen placement, we turn to resource dimensioning, where we optimize for power efficiency while meeting strict latency requirements in terms of chance constraints. We employ a novel approach by representing each slice as a Jackson Network, thereby capturing the interdependencies between various slice components. Through Dynamic Voltage Frequency Scaling (DVFS), we select appropriate CPU frequencies to reduce dynamic power consumption without compromising latency requirements. This ensures that both resource capacity and latency constraints are satisfied. Finally, we extend our solutions to a broader class of multi-tenant problems by introducing a general probabilistic latency framework called chance constrained multi-armed bandits (CCMAB). This generalization incorporates probabilistic constraints into multi-tenant decision-making, enabling applications such as task offloading between edge and cloud. Each tenant must satisfy chance constraints on stochastic delays while minimizing overall power consumption. We prove the optimality of the proposed algorithm and validate its performance through simulations. In summary, this thesis presents a cohesive, multi-layered approach that combines capacity-driven placement, latency-aware resource dimensioning, and probabilistic decision-making into a single framework for heterogeneous multi-tenant environments. This integration not only provides strong theoretical guarantees but also proposes practical algorithms for efficient resource utilization and reliability assurance across diverse service demands.<\/p>\n","protected":false},"excerpt":{"rendered":" L’Ecole doctorale : Ecole Doctorale de l’Institut Polytechnique de Paris et le Laboratoire de recherche SAMOVAR – Services r\u00e9partis, Architectures, Mod\u00e9lisation, Validation, Administration des R\u00e9seaux pr\u00e9sentent l\u2019AVIS DE SOUTENANCE de Monsieur Wei HUANG Autoris\u00e9 \u00e0 pr\u00e9senter ses travaux en vue de l\u2019obtention du Doctorat de l’Institut Polytechnique de Paris, pr\u00e9par\u00e9 \u00e0 T\u00e9l\u00e9com SudParis en : […]<\/p>\n","protected":false},"author":4,"featured_media":0,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ocean_post_layout":"","ocean_both_sidebars_style":"","ocean_both_sidebars_content_width":0,"ocean_both_sidebars_sidebars_width":0,"ocean_sidebar":"","ocean_second_sidebar":"","ocean_disable_margins":"enable","ocean_add_body_class":"","ocean_shortcode_before_top_bar":"","ocean_shortcode_after_top_bar":"","ocean_shortcode_before_header":"","ocean_shortcode_after_header":"","ocean_has_shortcode":"","ocean_shortcode_after_title":"","ocean_shortcode_before_footer_widgets":"","ocean_shortcode_after_footer_widgets":"","ocean_shortcode_before_footer_bottom":"","ocean_shortcode_after_footer_bottom":"","ocean_display_top_bar":"default","ocean_display_header":"default","ocean_header_style":"","ocean_center_header_left_menu":"","ocean_custom_header_template":"","ocean_custom_logo":0,"ocean_custom_retina_logo":0,"ocean_custom_logo_max_width":0,"ocean_custom_logo_tablet_max_width":0,"ocean_custom_logo_mobile_max_width":0,"ocean_custom_logo_max_height":0,"ocean_custom_logo_tablet_max_height":0,"ocean_custom_logo_mobile_max_height":0,"ocean_header_custom_menu":"","ocean_menu_typo_font_family":"","ocean_menu_typo_font_subset":"","ocean_menu_typo_font_size":0,"ocean_menu_typo_font_size_tablet":0,"ocean_menu_typo_font_size_mobile":0,"ocean_menu_typo_font_size_unit":"px","ocean_menu_typo_font_weight":"","ocean_menu_typo_font_weight_tablet":"","ocean_menu_typo_font_weight_mobile":"","ocean_menu_typo_transform":"","ocean_menu_typo_transform_tablet":"","ocean_menu_typo_transform_mobile":"","ocean_menu_typo_line_height":0,"ocean_menu_typo_line_height_tablet":0,"ocean_menu_typo_line_height_mobile":0,"ocean_menu_typo_line_height_unit":"","ocean_menu_typo_spacing":0,"ocean_menu_typo_spacing_tablet":0,"ocean_menu_typo_spacing_mobile":0,"ocean_menu_typo_spacing_unit":"","ocean_menu_link_color":"","ocean_menu_link_color_hover":"","ocean_menu_link_color_active":"","ocean_menu_link_background":"","ocean_menu_link_hover_background":"","ocean_menu_link_active_background":"","ocean_menu_social_links_bg":"","ocean_menu_social_hover_links_bg":"","ocean_menu_social_links_color":"","ocean_menu_social_hover_links_color":"","ocean_disable_title":"default","ocean_disable_heading":"default","ocean_post_title":"","ocean_post_subheading":"","ocean_post_title_style":"","ocean_post_title_background_color":"","ocean_post_title_background":0,"ocean_post_title_bg_image_position":"","ocean_post_title_bg_image_attachment":"","ocean_post_title_bg_image_repeat":"","ocean_post_title_bg_image_size":"","ocean_post_title_height":0,"ocean_post_title_bg_overlay":0.5,"ocean_post_title_bg_overlay_color":"","ocean_disable_breadcrumbs":"default","ocean_breadcrumbs_color":"","ocean_breadcrumbs_separator_color":"","ocean_breadcrumbs_links_color":"","ocean_breadcrumbs_links_hover_color":"","ocean_display_footer_widgets":"default","ocean_display_footer_bottom":"default","ocean_custom_footer_template":"","ocean_post_oembed":"","ocean_post_self_hosted_media":"","ocean_post_video_embed":"","ocean_link_format":"","ocean_link_format_target":"self","ocean_quote_format":"","ocean_quote_format_link":"post","ocean_gallery_link_images":"on","ocean_gallery_id":[],"footnotes":""},"categories":[286,402],"tags":[],"class_list":["post-6874","post","type-post","status-publish","format-standard","hentry","category-fractualites-ennews-fr","category-seminaires-ness-2013-fr","entry"],"_links":{"self":[{"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/posts\/6874","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/users\/4"}],"replies":[{"embeddable":true,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/comments?post=6874"}],"version-history":[{"count":1,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/posts\/6874\/revisions"}],"predecessor-version":[{"id":6875,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/posts\/6874\/revisions\/6875"}],"wp:attachment":[{"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/media?parent=6874"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/categories?post=6874"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/tags?post=6874"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
M. Andre-luc BEYLOT<\/strong>, Full professor, Institut de Recherche en Informatique de Toulouse, FRANCE – Rapporteur
M. Yassine HADJADJ-AOUL<\/strong>, Full professor, IRISA\/INRIA, University of Rennes, FRANCE – Examinateur
Mme Nihal PEKERGIN<\/strong>, Full professor, l’UFR des Sciences et Technologie, FRANCE – Examinateur
M. Nazim AGOULMINE<\/strong>, Full professor, University of evry, FRANCE – Examinateur
M. Andrea ARALDO<\/strong>, Associate Professor, Telecom Sudparis, FRANCE – Co-encadrant de these
M. Abdel LISSER<\/strong>, Full professor, Universit\u00e9 Paris Saclay – CentraleSupelec, FRANCE – Examinateur
Mme V\u00e9ronique VEQUE<\/strong>, Full professor, L2S, CentraleSup\u00e9lec, FRANCE – Rapporteur<\/p>\n\n\n\n<\/h2>\n\n\n\n
\u00ab Efficacit\u00e9 \u00e9nerg\u00e9tique et garanties de latence dans un environnement multi-locataire \u00bb<\/h2>\n\n\n\n
pr\u00e9sent\u00e9 par Monsieur Wei HUANG<\/h2>\n\n\n\n
<\/p>\n\n\n\n