In the LTE mobile system, all cells use the same set of frequencies. This means that a user could experience interference from other cells. A method that has been studied in order to reduce this interference and thereby increase data rate or system throughput is to coordinate scheduling between cells. Good results of this have been found in different studies. However, the interference is generally assumed to be known. Studies using estimated interference and simulating more than one cluster of cells have found almost no gain. This thesis will focus on how to use information from coordinated scheduling and other traffic estimates to do better interference estimation and link adaption. The suggested method is to coordinate larger clusters and use the coordination information, as well as estimates of which cells will be transmitting, to make estimates of interference from other cells. The additional information from interference estimation is used in the link adaptation. Limitations in bandwidth of the backhaul needed to send data between cells are considered, as well as the delay it may introduce. A limitation of the scope is that MIMO or HetNet scenarios have not been simulated. The suggested method for interference estimation and link adaptation have been implemented and simulated in a system simulator. The method gives a less biased estimate of SINR, but there are no gains in user bit rate. The lesser bias is since the method is better at predicting high SINR than the base estimate is. The lack of gains regarding user bit rate may result from the fact that in the studied scenarios, users where not able to make use of the higher estimated SINR since the base estimate is already high. The conclusion is that the method might be useful in scenarios where there are not full load, but the users either have bad channel quality or are able to make use of very high SINR. Such scenarios could be HetNet or MIMO scenarios, respectively.

The resource demand on embedded devices is constantly growing. This is caused by the sheer explosion of software based functions in embedded systems, that are growing far faster than the resources of the single-core and multi-core embedded processors. As one of the limitation is the computing power of the processors we need to explore ways to use this resource more efficiently. We identified that during the run-time of the embedded devices the resource demand of the software functions is permanently changing dependent on the device situation. To enable an embedded device to take advantage of this dynamic resource demand, the allocation of the software functions to the processor must be handled by a scheduler that is able to evaluate the resource demand of the software functions in relation to the device situation. This marks a change in embedded devices from static defined software systems to dynamic software systems. Above that we can increase the efficiency even further by extending the approach from a single device to a distributed or networked system (many-core system). However, existing approaches to deal with dynamic resource allocation are focused on individual devices and leave the optimization potential of manycore systems untouched. Our concept will extend the existing Hierarchical Asynchronous Multi-Core Scheduler (HAMS) concept for individual devices to many-core systems. This extension introduces a dynamic situation aware scheduler for many-core systems which take the current workload of all devices and the system-situation into account. With our approach, the resource efficiency of an embedded many-core system can be increased. The following paper will explain the architecture and the expected results of our concept.

A challenge posed by future computer architectures is the efficient exploitation of their many and sometimes heterogeneous computational cores. This challenge is exacerbated by the multiple facilities for data movement and sharing across cores resident on such platforms. To answer the question of how systems software should treat heterogeneous resources, this dissertation describes an approach that (1) creates a common manageable pool for all the resources present in the platform, and then (2) provides virtual machines (VMs) with multiple `personalities', flexibly mapped to and efficiently run on the heterogeneous underlying hardware. A VM's personality is its execution context on the different types of available processing resources usable by the VM. We provide mechanisms for making such platforms manageable and evaluate coordinated scheduling policies for mapping different VM personalities on heterogeneous hardware. Towards that end, this dissertation contributes technologies that include (1) restructuring hypervisor and system functions to create high performance environments that enable flexibility of execution and data sharing, (2) scheduling and other resource management infrastructure for supporting diverse application needs and heterogeneous platform characteristics, and (3) hypervisor level policies to permit efficient and coordinated resource usage and sharing. Experimental evaluations on multiple heterogeneous platforms, like one comprised of x86-based cores with attached NVIDIA accelerators and others with asymmetric elements on chip, demonstrate the utility of the approach and its ability to efficiently host diverse applications and resource management methods.

Increasing global competition in the business world and heightened expectations of customers have forced companies to consider not only the pricing or product quality, but reliability and timeliness of the deliveries as well. In manufacturing-centric industries such as automotive and electronics, distribution and inventory costs constitute the second and third largest cost components following the production costs. Therefore, industrial and logistics companies need to continuously search for ways to lower the inventory level and distribution cost. This trend has created a closer interaction between the different stages of a supply chain, and increased the practical usefulness of the integrated models.This thesis considers two categories of integrated scheduling problems. One is Integrated Scheduling of Production-Distribution-Inventory problems (ISPDI problems) and the other is Integrated Scheduling of Production-Inventory-Distribution-Inventory problems (ISPIDI problems). Jobs are first processed on a single machine in the production stage, and then delivered to a pre-specified customer by a capacitated transporter. Each job has a distinct due date, and must be delivered to customer before this due date. Each production batch requires a setup cost and a setup time before the first job of this batch is processed. Each round trip between the factory and customer requires a delivery cost as well as a delivery time. Moreover, it is assumed that a job which is completed before its departure date or delivered to the customer before its due date will incur a corresponding inventory cost. Our objective is to minimize the total cost involving setup, inventory and delivery costs while guaranteeing a certain customer service level.For ISPDI problems, we firstly provide a mixed integer programming model for the case of multi-product, single-stage situation, and develop an improved Genetic algorithm (GA) for solving it. Then, we extend this model to a single-product, multi-stage model, and provide two methods, dominance-related greedy algorithm and GA, for solving it. For ISPIDI problems, we establish a general non-linear model for the case of single-product situation and devise a special case from the general model. Then we provide an optimality property between the production and delivery schedules for the special case. Finally, a heuristic approach is developed for solving it. For each problem under study, in order to evaluate the performance of the proposed algorithms, some interesting lower bounds on the corresponding objective functions are established according to different methods such as Lagrangian relaxation method, classical bin-packing based method. Computational results show the efficiency of the proposed models and algorithms in terms of solution quality and running time.

La coordination des activités dans une chaîne logistique a suscité récemment beaucoup d'attention dans larecherche de gestion. Dans une chaîne logistique typique, des produits finis sont produits et transportés pour êtrestockage temporairement ou arriver directement chez clients. Pour réaliser la représentation opérationnelleoptimale, la coordination et l'intégration de la production, de la livraison, et du stockage devient très importante.L'étude récente a considéré le coût de stockage de client avec le coût fixe de transport ou la taille fixe des lots, cequi est irréaliste.Dans cette thèse, nous étudions la coordination de traitement des produits en lots et l'ordonnancement des lots, quiinclut la coordination du traitement en lots des produits dans les groupes après l'étape de production,l'ordonnancement des lots après les productions exigent la livraison du fournisseur ainsi que le stockage desproduits aux clients. Ce travail focalise sur les cas du simple-fournisseur/plusieurs-clients et le cas de simplefournisseur/plusieurs-transporteurs. Pour le premier scénario avec plusieurs-clients, deux modèles illustrent letransfert des lots aux clients. Dans le premier modèle, nous avons considéré un simple-fournisseur/plusieurs-clientsavec un transporteur disponible pour servir les clients sans considération du problème de tournée de véhicule. Puis,dans le deuxième modèle, nous avons considéré un simple-fournisseur/plusieurs-clients avec plusieurs transportersdisponibles pour servir les clients. Différentes hypothèses sont proposées et comparées dans le dernier chapitre.Pour ce qui concerne le deuxième scénario, nous avons étudié le cas du simple-fournisseur avec plusieurstransporteurs disponibles pour servir un seul client. Dans ce scénario, des modèles avec les véhicules homogèneset hétérogènes sont étudiés. Tout le coût du système est calculé en additionnant de tout le coût de la livraison et destockage pour les différents clients et transporteurs qui se sont dans le système à étudier. Le nombre des produitsdans les lots peut être inégal et les lots sont limités seulement par la capacité du transporteur utilisé. Le coût destockage chez les clients dépend de chaque client, la distance entre le fournisseur et leurs clients dépend del'emplacement de client, qui est le cas du coût de livraison également qui dépend de l'emplacement du client. Dansle cas des multi transporteurs, le coût de livraison dépend du transporteur utilisé.Dans chaque modèle, nous présentons ce qui suit : procédures de solution pour résoudre chaque modèle, plusieursexemples numériques pour soutenir des résultats mathématiques et pour clarifier le problème, et comparaisons desperformances parmi différents résultats. La future extension de cette recherche peut considérer des contraintes detemps et de coût de chargement dans l'étape de production, la considération du tourné de véhicule avec un cout destockage chez les clients. / The coordination of logistics activities in a supply chain has recently received a lot of attention in operationsmanagement research. In a typical supply chain, finished products are produced and either shipped to be temporarystorage or arrived directly on time to the customers. To achieve optimal operational performance, the coordinationand integration of production, delivery, and storage is an important consideration. The recent study consideredcustomer storage cost with fixed transportation cost or fixed batch size, which is unrealistic. In this thesis, we studythe coordinate of batching and scheduling activities, which includes the coordination of batching of products inbatches after the production stage, the coordination of scheduling, customer(s) orders which require the deliveryfrom the supplier, and the storage of products at the customer(s). This study focus on single-supplier/multi-customerscenario and single-supplier/multi-transporter scenario. For the first scenario with multi-customer, two modelsillustrate the transferring of batches to the customer. Where in the first model, we considered a singlesupplier/ multicustomerwith one capacitated transporter available to serve the customers without the vehicle routingconsideration. Then, in the second model, we considered a single-supplier/multi-customer with multi-transportavailable to serve the customers. In this case different assumption is proposed and compared in the last chapter.Concerning the second scenario, we studied the case of single-supplier with multi-transporter available to serve asingle customer. In this scenario, models with homogeneous and heterogeneous vehicles are studied. The totalsystem cost is calculated by summing the total delivery and storage cost for different customers and transporters inthe system. The number of products by batch is unequal and they are limited only by the capacity of the transporterused. The storage cost of the customers depends on the customer destination, the distance between the supplierand their customers depends on the customer location, which is the case of the delivery cost also which depends onthe customer¿s location. In the case of the multi-transporters, the delivery cost depends on the transporter used.In each model, we present the following: solution procedures to solve each model, many numerical examples tosupport mathematical findings and to clarify the problem under study, and performance comparisons amongdifferent findings. The future extension of this research may involve considering setup time and cost constraints inthe production stage, the vehicle routing consideration with inventory in the multi-customer case.

The resource demand on embedded devices is constantly growing. This is caused by the sheer explosion of software based functions in embedded systems, that are growing far faster than the resources of the single-core and multi-core embedded processors. As one of the limitation is the computing power of the processors we need to explore ways to use this resource more efficiently. We identified that during the run-time of the embedded devices the resource demand of the software functions is permanently changing dependent on the device situation. To enable an embedded device to take advantage of this dynamic resource demand, the allocation of the software functions to the processor must be handled by a scheduler that is able to evaluate the resource demand of the software functions in relation to the device situation. This marks a change in embedded devices from static defined software systems to dynamic software systems. Above that we can increase the efficiency even further by extending the approach from a single device to a distributed or networked system (many-core system). However, existing approaches to deal with dynamic resource allocation are focused on individual devices and leave the optimization potential of manycore systems untouched. Our concept will extend the existing Hierarchical Asynchronous Multi-Core Scheduler (HAMS) concept for individual devices to many-core systems. This extension introduces a dynamic situation aware scheduler for many-core systems which take the current workload of all devices and the system-situation into account. With our approach, the resource efficiency of an embedded many-core system can be increased. The following paper will explain the architecture and the expected results of our concept.

The next generation wireless networks i.e. 5G aim to provide multi-Gbps data traffic, in order to satisfy the increasing demand for high-definition video, among other high data rate services, as well as the exponential growth in mobile subscribers. To achieve this dramatic increase in data rates, current research is focused on improving the capacity of current 4G network standards, based on Long Term Evolution (LTE), before radical changes are exploited which could include acquiring additional/new spectrum. The LTE network has a reuse factor of one; hence neighbouring cells/sectors use the same spectrum, therefore making the cell edge users vulnerable to inter-cell interference. In addition, wireless transmission is commonly hindered by fading and pathloss. In this direction, this thesis focuses on improving the performance of cell edge users in LTE and LTE-Advanced (LTE-A) networks by initially implementing a new Coordinated Multi-Point (CoMP) algorithm to mitigate cell edge user interference. Subsequently Device-to-Device (D2D) communication is investigated as the enabling technology for maximising Resource Block (RB) utilisation in current 4G and emerging 5G networks. It is demonstrated that the application, as an extension to the above, of novel power control algorithms, to reduce the required D2D TX power, and multihop transmission for relaying D2D traffic, can further enhance network performance. To be able to develop the aforementioned technologies and evaluate the performance of new algorithms in emerging network scenarios, a beyond-the-state-of-the-art LTE system-level simulator (SLS) was implemented. The new simulator includes Multiple-Input Multiple-Output (MIMO) antenna functionalities, comprehensive channel models (such as Wireless World initiative New Radio II i.e. WINNER II) and adaptive modulation and coding schemes to accurately emulate the LTE and LTE-A network standards. Additionally, a novel interference modelling scheme using the 'wrap around' technique was proposed and implemented that maintained the topology of flat surfaced maps, allowing for use with cell planning tools while obtaining accurate and timely results in the SLS compared to the few existing platforms. For the proposed CoMP algorithm, the adaptive beamforming technique was employed to reduce interference on the cell edge UEs by applying Coordinated Scheduling (CoSH) between cooperating cells. Simulation results show up to 2-fold improvement in terms of throughput, and also shows SINR gain for the cell edge UEs in the cooperating cells. Furthermore, D2D communication underlaying the LTE network (and future generation of wireless networks) was investigated. The technology exploits the proximity of users in a network to achieve higher data rates with maximum RB utilisation (as the technology reuses the cellular RB simultaneously), while taking some load off the Evolved Node B (eNB) i.e. by direct communication between User Equipment (UE). Simulation results show that the proximity and transmission power of D2D transmission yields high performance gains for a D2D receiver, which was demonstrated to be better than that of cellular UEs with better channel conditions or in close proximity to the eNB in the network. The impact of interference from the simultaneous transmission however impedes the achievable data rates of cellular UEs in the network, especially at the cell edge. Thus, a power control algorithm was proposed to mitigate the impact of interference in the hybrid network (network consisting of both cellular and D2D UEs). It was implemented by setting a minimum SINR threshold so that the cellular UEs achieve a minimum performance, and equally a maximum SINR threshold to establish fairness for the D2D transmission as well. Simulation results show an increase in the cell edge throughput and notable improvement in the overall SINR distribution of UEs in the hybrid network. Additionally, multihop transmission for D2D UEs was investigated in the hybrid network: traditionally, the scheme is implemented to relay cellular traffic in a homogenous network. Contrary to most current studies where D2D UEs are employed to relay cellular traffic, the use of idle nodes to relay D2D traffic was implemented uniquely in this thesis. Simulation results show improvement in D2D receiver throughput with multihop transmission, which was significantly better than that of the same UEs performance with equivalent distance between the D2D pair when using single hop transmission.

L'augmentation de la concurrence économique internationale et les attentes accrues des clients ont imposé aux entreprises de prendre en compte non seulement le prix ou la qualité du produit, mais également la fiabilité et la rapidité des livraisons. Dans les industries ayant une composante manufacturière dominante telles que l'automobile et l'électronique, la distribution et les coûts de stockage constituent les deuxième et troisième catégories de coûts les plus importantes après les coûts de production. Par conséquent, les entreprises industrielles et de logistique recherchent continuellement des méthodes pour réduire le niveau des stocks et les coûts de distribution. Cette tendance a créé une interaction plus forte entre les différentes étapes de la chaîne logistique, et augmente de ce fait l'utilité pratique des modèles intégrés.Cette thèse considère deux catégories de problèmes d'ordonnancement intégré. La première catégorie est l'ordonnancement intégré de la production, distribution et stockage (Integrated Scheduling of Production-Distribution-Inventory, ISPDI) et la deuxième est l'ordonnancement intégré de la production, stockage, distribution et stockage (Integrated Scheduling of Production-Inventory-Distribution-Inventory, ISPIDI). Au niveau de la production, les tâches à réaliser sont traitées sur une seule machine et regroupées par lot de production, ce qui nécessite un coût et un temps de réglage. Elles doivent ensuite être livrées à un client prédéfini par un transporteur à capacité limitée, avant des dates dues données. Chaque aller-retour du transporteur entre l'usine et le client implique un coût de livraison et des délais de livraison. De plus, on suppose que les tâches qui sont terminées avant leur date de départ ou qui sont livrées au client avant leur date due entraînent un coût de stockage supplémentaire. Notre objectif est de minimiser le coût total comprenant les coûts de reglage, de stockage et de transport, tout en garantissant un niveau de service donné pour le client.Pour les problèmes ISPDI, nous avons d'abord fourni un modèle de programmation mixte entière pour le problème multi-produits, à un seul niveau, et avons développé un algorithme génétique amélioré pour le résoudre. Puis, nous avons modifié ce modèle pour prendre en compte le cas mono-produit, multi-niveau, et avons proposé deux méthodes, un algorithme hybride et un algorithme génétique, pour le résoudre. Pour les problèmes ISPIDI, nous avons établi un modèle général non-linéaire dans le cas mono-produit, et avons traité un cas spécifique du cas général. Puis nous avons démontré une propriété d'optimalité qui lie les ordonnancements de production et de livraison dans le cas particulier, pour finalement proposer une approche heuristique pour le résoudre. Pour chaque problème étudié et afin d'évaluer la performance des algorithmes proposés, des limites inférieures intéressantes sur les fonctions objectifs correspondantes ont été établies selon des méthodes différentes telles que la méthode de relaxation lagrangienne ou des méthodes basées sur les bornes inférieures du problème de bin packing. Les résultats des expérimentations montrent l'efficacité des modèles et algorithmes proposés en termes de qualité de la solution et de temps d'exécution. / Increasing global competition in the business world and heightened expectations of customers have forced companies to consider not only the pricing or product quality, but reliability and timeliness of the deliveries as well. In manufacturing-centric industries such as automotive and electronics, distribution and inventory costs constitute the second and third largest cost components following the production costs. Therefore, industrial and logistics companies need to continuously search for ways to lower the inventory level and distribution cost. This trend has created a closer interaction between the different stages of a supply chain, and increased the practical usefulness of the integrated models.This thesis considers two categories of integrated scheduling problems. One is Integrated Scheduling of Production-Distribution-Inventory problems (ISPDI problems) and the other is Integrated Scheduling of Production-Inventory-Distribution-Inventory problems (ISPIDI problems). Jobs are first processed on a single machine in the production stage, and then delivered to a pre-specified customer by a capacitated transporter. Each job has a distinct due date, and must be delivered to customer before this due date. Each production batch requires a setup cost and a setup time before the first job of this batch is processed. Each round trip between the factory and customer requires a delivery cost as well as a delivery time. Moreover, it is assumed that a job which is completed before its departure date or delivered to the customer before its due date will incur a corresponding inventory cost. Our objective is to minimize the total cost involving setup, inventory and delivery costs while guaranteeing a certain customer service level.For ISPDI problems, we firstly provide a mixed integer programming model for the case of multi-product, single-stage situation, and develop an improved Genetic algorithm (GA) for solving it. Then, we extend this model to a single-product, multi-stage model, and provide two methods, dominance-related greedy algorithm and GA, for solving it. For ISPIDI problems, we establish a general non-linear model for the case of single-product situation and devise a special case from the general model. Then we provide an optimality property between the production and delivery schedules for the special case. Finally, a heuristic approach is developed for solving it. For each problem under study, in order to evaluate the performance of the proposed algorithms, some interesting lower bounds on the corresponding objective functions are established according to different methods such as Lagrangian relaxation method, classical bin-packing based method. Computational results show the efficiency of the proposed models and algorithms in terms of solution quality and running time.