The fifth-generation (5G) and beyond mobile networks aim at satisfying, in different demanding application scenarios, stringent Quality of Service (QoS) requirements, among which latency is one of the key metrics that mobile operators are supposed to optimize for mobile users. Mobile Edge Computing (MEC), an essential techniques utilized in 5G networks, brings cloud-computing capabilities to the edge of the mobile networks, especially in close proximity to mobile users, making it possible to simultaneously address the stringent latency requirements of critical services and ensure highly efficient network operation and service delivery, so as to improve user experience. MEC services, on one hand, require significant investments from both network operators and service providers in terms of deploying, operating and managing edge clouds, and on the other, provide limited computational and storage resources by design as data centers are deployed in the core of the network. During peak hours, the operator must serve a large amount of tasks from users with high demands, hence the latency requirements of different services can hardly be guaranteed. This issue can be tackled by massively deployed edge clouds that are attached to the base stations and connected to each other in a specific topology, as ultra-dense 5G-and-Beyond networks are built. In this way, the resource limitations can be solved through sharing computational and storage capabilities among multiple MEC units nearby. In this thesis, we leverage cooperation among interconnected multiple MEC units and investigate joint resource optimization considering multiple aspects of network operations, with the target of enhancing the utilization efficiency of resources to further satisfy improved QoS and reduce network operation cost. Specifically, aggregated mobile traffic and user requests are considered based on their types (e.g., voice, video, web, game, etc.) that are associated with different QoS requirements. We jointly optimize 1) where to process the traffic and requests, 2) how to route network flows and 3) how to allocate and schedule the required resources with regard to communication, computation and storage. Firstly, to serve mobile traffic, we investigate in the context of hierarchical edge networks and propose a mathematical optimization model to perform a joint slicing of mobile network and edge computation resources. The optimization aims at minimizing the total traffic latency considering operations including transmitting, outsourcing and processing user traffic, under the constraint of user tolerable latency for each class of traffic. Then, we release the constraints on hierarchical network and fixed computation capacity, and further take into account the overall budget of operators to plan and allocate the computation capabilities in edge network with an arbitrary topology. The main objective, aligned with the first one, is to further operate cost-efficient edge networks through jointly planning the availability of computational resources at the edge, slicing mobile network and edge computation resources, and routing heterogeneous traffic types to various slices. We propose an optimization model to minimize the network operation cost and the total traffic latency in the procedures of transmitting, outsourcing and processing user traffic, under the constraint of user maximum tolerable latency for each class of traffic. Finally, we focus on serving multiple classes of user requests (with starting time, deadline and duration) which require bandwidth, storage and computation resources. The main objective is to exploit the flexibility of services to requests by shifting the starting time without penalizing utility perceived by users, while, in the meantime, permitting an efficient resource utilization. We propose an optimization framework that jointly considers several key aspects of the resource allocation problem, specifically, through optimizing: admission decision, scheduling of admitted requests (also called calendaring), routing of these flows, the decision of which nodes will serve such requests, as well as the amount of processing and storage capacity reserved on the chosen nodes, with the objective of maximizing the operator's profit. The above proposed optimization models are first formulated as mixed-integer nonlinear programming (MINLP) problems, which are NP-hard. To tackle them efficiently, we perform equivalent reformulations from MINLP to mixed-integer quadratically constraint programming (MIQCP), and based on that, further propose effective heuristics to facilitate the solutions of the problems, i.e., Sequential Fixing (SF) for the edge slicing model, Neighbor Exploration and Sequential Fixing (NESF) for the edge planning model and Sequential Fixing and Scheduling (SFS) as well as a distributed algorithm based on Alternating Direction Method of Multipliers (ADMM) for the edge calendaring model. We evaluate the performance of our proposed models and heuristics in real-size network scenarios including both random geometric graphs and a realistic mobile network topology, showing the impact of all the considered parameters (i.e., different types of user traffic or requests, tolerable latency, network topology and bandwidth, computation, storage and link capacities) on both the optimal and approximate solutions. Results obtained demonstrate that near-optimal resource allocation solutions can be achieved by our heuristics in short computing time.
Le reti mobili di quinta generazione (5G) e oltre mirano a soddisfare, in diversi scenari applicativi complessi, severi requisiti di qualità del servizio (QoS). Tra gli altri, la latenza è una delle metriche chiave che gli operatori mobili dovrebbero ottimizzare per i loro utenti. Il Mobile Edge Computing (MEC), una tecnica essenziale utilizzata nelle reti 5G, porta le capacità di cloud computing ai confini delle reti mobili, specialmente in prossimità degli utenti mobili, rendendo possibile affrontare simultaneamente i severi requisiti di latenza dei servizi critici e garantire un funzionamento della rete e un'erogazione dei servizi altamente efficiente, in modo da migliorare l'esperienza dell'utente. I servizi MEC, da un lato, richiedono investimenti significativi sia da parte degli operatori di rete che dei fornitori di servizi in termini di implementazione, funzionamento e gestione dei cloud edge e, dall'altro, forniscono risorse di calcolo e archiviazione limitate poiché i data center vengono distribuiti nella core network. Durante le ore di punta, l'operatore deve svolgere una grande quantità di attività da parte di utenti con esigenze elevate, quindi i requisiti di latenza dei diversi servizi difficilmente possono essere garantiti. Questo problema può essere affrontato da edge cloud distribuiti in modo massiccio che sono collegati alle stazioni base e collegati tra loro in una topologia specifica, poiché vengono costruite reti 5G e oltre ad altissima densità. In questo modo, i limiti delle risorse locali a un'unità MEC possono essere risolti condividendo capacità di calcolo e archiviazione tra più unità MEC vicine. In questa tesi, facciamo leva sulla cooperazione tra più unità MEC interconnesse e indaghiamo sull'ottimizzazione congiunta delle risorse considerando molteplici aspetti delle operazioni di rete, con l'obiettivo di migliorare l'efficienza di utilizzo delle risorse per ottenere una migliore QoS e ridurre i costi di funzionamento della rete. Nello specifico, il traffico mobile aggregato e le richieste degli utenti vengono considerati in base ai loro tipi (ad es. Voce, video, web, gioco, ecc.) e associati a diversi requisiti di QoS. Ottimizziamo congiuntamente 1) dove elaborare il traffico e le richieste, 2) come instradare i flussi di rete e 3) come allocare e programmare le risorse richieste per quanto riguarda la comunicazione, il calcolo e l'archiviazione. In primo luogo, per servire il traffico mobile, indaghiamo nel contesto delle reti edge gerarchiche e proponiamo un modello di ottimizzazione matematica per eseguire uno slicing congiunto della rete mobile e delle risorse di calcolo edge. L'ottimizzazione mira a ridurre al minimo la latenza totale del traffico considerando le operazioni che includono la trasmissione, l'outsourcing e l'elaborazione del traffico degli utenti, sotto il vincolo della latenza tollerabile dall'utente per ciascuna classe di traffico. Quindi, rilassiamo i vincoli sulla rete gerarchica e sulla capacità di calcolo fisse e prendiamo ulteriormente in considerazione il budget complessivo degli operatori per pianificare e allocare le capacità di calcolo nella rete edge con una topologia arbitraria. L'obiettivo principale, in linea con il primo, è quello di gestire ulteriormente le reti edge efficienti in termini di costi, attraverso la pianificazione congiunta della disponibilità di risorse computazionali all'edge, la suddivisione delle risorse di rete mobile e di calcolo edge e l'instradamento di tipi di traffico eterogenei a varie sezioni. Proponiamo un modello di ottimizzazione per minimizzare il costo di funzionamento della rete e la latenza totale del traffico nelle procedure di trasmissione, esternalizzazione ed elaborazione del traffico degli utenti, sotto il vincolo della massima latenza tollerabile da parte dell'utente per ogni classe di traffico. Infine, ci concentriamo sul servire più classi di richieste degli utenti (con ora di inizio, scadenza e durata) che richiedono larghezza di banda, risorse di archiviazione e di calcolo. L'obiettivo principale è quello di sfruttare la flessibilità dei servizi alle richieste spostando l'orario di partenza senza penalizzare l'utilità percepita dagli utenti, consentendo, nel contempo, un efficiente utilizzo delle risorse. Proponiamo un framework di ottimizzazione che considera congiuntamente diversi aspetti chiave del problema di allocazione delle risorse, in particolare, attraverso l'ottimizzazione: decisione di ammissione, pianificazione delle richieste ammesse (chiamata anche calendario), instradamento di questi flussi, la decisione di quali nodi serviranno tali richieste, nonché la quantità di capacità di elaborazione e stoccaggio riservata sui nodi scelti, con l'obiettivo di massimizzare il profitto dell'operatore. I modelli di ottimizzazione proposti sopra sono inizialmente formulati come problemi di programmazione non lineare a interi misti (MINLP), che sono NP-hard. Per affrontarli in modo efficiente, eseguiamo riformulazioni equivalenti da MINLP a programmazione a vincoli quadratici a interi misti (MIQCP) e, sulla base di ciò, proponiamo ulteriori euristiche efficaci per facilitare le soluzioni dei problemi, ovvero, Sequential Fixing (SF) per lo slicing dei bordi model, Neighbor Exploration and Sequential Fixing (NESF) per il modello di pianificazione degli edge e Sequential Fixing and Scheduling (SFS), nonché un algoritmo distribuito basato su Alternating Direction Method of Multipliers (ADMM) per il modello di calendarizzazione degli edge. Valutiamo le prestazioni dei modelli proposti e le relative euristiche in scenari di rete di dimensioni reali, inclusi sia reti a grafi geometrici casuali che una topologia di rete mobile realistica, mostrando l'impatto di tutti i parametri considerati (ad esempio, diversi tipi di traffico utente o richieste, latenza tollerabile, topologia di rete e larghezza di banda, capacità di calcolo, archiviazione e collegamento) sulle soluzioni ottimali e approssimative. I risultati ottenuti dimostrano che soluzioni di allocazione delle risorse quasi ottimali possono essere ottenute dalla nostra euristica in tempi di elaborazione ridotti.
Optimization framework for resource management of mobile edge computing networks
Xiang, Bin
2020/2021
Abstract
The fifth-generation (5G) and beyond mobile networks aim at satisfying, in different demanding application scenarios, stringent Quality of Service (QoS) requirements, among which latency is one of the key metrics that mobile operators are supposed to optimize for mobile users. Mobile Edge Computing (MEC), an essential techniques utilized in 5G networks, brings cloud-computing capabilities to the edge of the mobile networks, especially in close proximity to mobile users, making it possible to simultaneously address the stringent latency requirements of critical services and ensure highly efficient network operation and service delivery, so as to improve user experience. MEC services, on one hand, require significant investments from both network operators and service providers in terms of deploying, operating and managing edge clouds, and on the other, provide limited computational and storage resources by design as data centers are deployed in the core of the network. During peak hours, the operator must serve a large amount of tasks from users with high demands, hence the latency requirements of different services can hardly be guaranteed. This issue can be tackled by massively deployed edge clouds that are attached to the base stations and connected to each other in a specific topology, as ultra-dense 5G-and-Beyond networks are built. In this way, the resource limitations can be solved through sharing computational and storage capabilities among multiple MEC units nearby. In this thesis, we leverage cooperation among interconnected multiple MEC units and investigate joint resource optimization considering multiple aspects of network operations, with the target of enhancing the utilization efficiency of resources to further satisfy improved QoS and reduce network operation cost. Specifically, aggregated mobile traffic and user requests are considered based on their types (e.g., voice, video, web, game, etc.) that are associated with different QoS requirements. We jointly optimize 1) where to process the traffic and requests, 2) how to route network flows and 3) how to allocate and schedule the required resources with regard to communication, computation and storage. Firstly, to serve mobile traffic, we investigate in the context of hierarchical edge networks and propose a mathematical optimization model to perform a joint slicing of mobile network and edge computation resources. The optimization aims at minimizing the total traffic latency considering operations including transmitting, outsourcing and processing user traffic, under the constraint of user tolerable latency for each class of traffic. Then, we release the constraints on hierarchical network and fixed computation capacity, and further take into account the overall budget of operators to plan and allocate the computation capabilities in edge network with an arbitrary topology. The main objective, aligned with the first one, is to further operate cost-efficient edge networks through jointly planning the availability of computational resources at the edge, slicing mobile network and edge computation resources, and routing heterogeneous traffic types to various slices. We propose an optimization model to minimize the network operation cost and the total traffic latency in the procedures of transmitting, outsourcing and processing user traffic, under the constraint of user maximum tolerable latency for each class of traffic. Finally, we focus on serving multiple classes of user requests (with starting time, deadline and duration) which require bandwidth, storage and computation resources. The main objective is to exploit the flexibility of services to requests by shifting the starting time without penalizing utility perceived by users, while, in the meantime, permitting an efficient resource utilization. We propose an optimization framework that jointly considers several key aspects of the resource allocation problem, specifically, through optimizing: admission decision, scheduling of admitted requests (also called calendaring), routing of these flows, the decision of which nodes will serve such requests, as well as the amount of processing and storage capacity reserved on the chosen nodes, with the objective of maximizing the operator's profit. The above proposed optimization models are first formulated as mixed-integer nonlinear programming (MINLP) problems, which are NP-hard. To tackle them efficiently, we perform equivalent reformulations from MINLP to mixed-integer quadratically constraint programming (MIQCP), and based on that, further propose effective heuristics to facilitate the solutions of the problems, i.e., Sequential Fixing (SF) for the edge slicing model, Neighbor Exploration and Sequential Fixing (NESF) for the edge planning model and Sequential Fixing and Scheduling (SFS) as well as a distributed algorithm based on Alternating Direction Method of Multipliers (ADMM) for the edge calendaring model. We evaluate the performance of our proposed models and heuristics in real-size network scenarios including both random geometric graphs and a realistic mobile network topology, showing the impact of all the considered parameters (i.e., different types of user traffic or requests, tolerable latency, network topology and bandwidth, computation, storage and link capacities) on both the optimal and approximate solutions. Results obtained demonstrate that near-optimal resource allocation solutions can be achieved by our heuristics in short computing time.| File | Dimensione | Formato | |
|---|---|---|---|
|
PhD_Thesis_BinXiang.pdf
accessibile in internet solo dagli utenti autorizzati
Dimensione
3.42 MB
Formato
Adobe PDF
|
3.42 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/176180