Search

In this paper, a power minimization problem using base station sleeping is proposed for heterogeneous cloud radio access networks (H-CRANs) taking into account the computing delay constraints. In the proposed system, which is modeled using M/M/k queues, the edge device coexists with the small base station (SBS) to provide computing capabilities beside the central cloud. In general, the SBS sleeping is governed by the availability of resources provided the macro base station (MBS) which is in charge of accommodating offloaded users from sleeping SBSs. However, switching off lightly loaded SBSs can impose significant burdens on cloud servers. Here, the proposed sleeping scheme allows SBSs serving more computing tasks to remain active in order to fulfill the task completion deadlines requested by mobile users and to keep the cloud response time within a predefined limit. In other words, the proposed scheme aims to save power by undertaking a centralized selection of active and sleeping SBSs taking into account the delay constraints of both cloud and mobile devices. First, we consider a disjoint cloud-edge system, where computing services can be provided by either the cloud or the edge device, and aim to minimize the number of active SBSs. The problem is formulated as a 0-1 knapsack problem with SBS utilization considered as the weight while the ratio of computing tasks to all incoming tasks is considered as the value of that SBS. In this problem, which is solved using dynamic programming, SBSs processing less computing tasks are given higher values; and as a result, higher chance to sleep compared to others. Second, a shared computing system is proposed whereby active SBSs (edge devices) contribute to the total computing capability. Here, an exhaustive search approach is used to achieve the optimal power saving. We also proved that the shared computing system performs better in terms of response time compared to the disjoint system depending on the number of active SBSs.

In this paper, an energy saving scheme is presented whereby unused virtual machines (VMs) in distributed edge devices enter a sleep mode taking into account the quality of service (QoS) experienced by mobile users. The virtual machines in the system under consideration form a shared VM pool assisted by the software-defined networking (SDN) technology. The energy saving optimization problem aims to minimize the number of active VMs without violating the QoS requirements, and is solved using the square-root staffing rule and the Halfin-Whitt function. In addition, a traffic management strategy in overloaded edge devices is carried out to maximize the number of accommodated users taking into account the fronthaul capacity which impacts task migration among edge devices. Results show the effectiveness of the proposed scheme in saving remarkable amount of energy while satisfying the QoS requirements. It is also shown that the VM sleeping scheme exhibits higher energy saving and less task migration compared to the entire edge device sleeping scheme due to its enhanced flexibility.

The ever-increasing growth of connected smart devices and Internet of Things (IoT) verticals is leading to the crucial challenges of handling the massive amount of raw data generated by distributed IoT systems and providing timely feedback to the end-users. Although existing cloud computing paradigm has an enormous amount of virtual computing power and storage capacity, it might not be able to satisfy delaysensitive applications since computing tasks are usually processed at the distant cloud-servers. To this end, edge/fog computing has recently emerged as a new computing paradigm that helps to extend cloud functionalities to the network edge. Despite several benefits of edge computing including geo-distribution, mobility support and location awareness, various communication and computing related challenges need to be addressed for future IoT systems. In this regard, this paper provides a comprehensive view on the current issues encountered in distributed IoT systems and effective solutions by classifying them into three main categories, namely, radio and computing resource management, intelligent edge-IoT systems, and flexible infrastructure management. Furthermore, an optimization framework for edge-IoT systems is proposed by considering the key performance metrics including throughput, delay, resource utilization and energy consumption. Finally, a Machine Learning (ML) based case study is presented along with some numerical results to illustrate the significance of ML in edge-IoT computing.

In this paper, a sparse code multiple access (SCMA)-based edge computing scheme is proposed for Internet-of- Things (IoT) systems. The aim of implementing SCMA, which is a nonorthogonal multiple access resource allocation technique, is to improve network connectivity and maximize data rate provision. The proposed edge-IoT system is investigated under different SCMA configurations to explore the various performance aspects such as connectivity, throughput, task completion time, and complexity. First, the problem is formulated as a data rate maximization problem for SCMA-based heterogeneous networks under power constraints. Then, the problem is subdivided into a power allocation problem, which is solved using the water filling approach, and a codebook allocation problem that is solved using a heuristic algorithm. The results show that the SCMA scheme can significantly improve the IoT performance compared to the conventional orthogonal frequencydivision multiple access resource allocation scheme in terms of connectivity, throughput, and task completion time provided that SCMA configurations are suitable with IoT processing capabilities to avoid undesired detection latency. Natural Sciences and Engineering Research Council of Canada CGIAR Tubitak

With the growing popularity of smart applications that contain computing-intensive tasks, the provision of radio and computing resources with high quality is becoming more and more challenging. Moreover, supporting network scalability is crucial to accommodate the massive numbers of connected devices. In this thesis, we present effective energy saving strategies that consider the utilization of network elements such as base stations and virtual machines, and implement on/off mechanisms taking into account the quality of service (QoS) required by mobile users. Moreover, we investigate the performance of a NOMA-based resource allocation scheme in the context of Internet of Things aiming to improve network scalability and reduce the energy consumption of mobile users. The system model is mainly built upon the M/M/k queueing system that has been widely used in most relevant works. First, the energy saving mechanism is formulated as a 0-1 knapsack problem where the weight and value of each small base station is determined by the utilization and proportion of computing tasks at that base station, respectively. The problem is then solved using the dynamic programming approach which showed significant energy saving performance while maintaining the cloud response time at desired levels. Afterwards, the energy saving mechanism is applied on edge computing to reduce the amount of under-utilized virtual machines in edge devices. Herein, the square-root staffing rule and the Halfin-Whitt function are used to determine the minimum number of virtual machines required to maintain the queueing probability below a threshold value. On the user level, reducing energy consumption can be achieved by maximizing data rate provision to reduce the task completion time, and hence, the transmission energy. Herein, a NOMA-based scheme is introduced, particularly, the sparse code multiple access (SCMA) technique that allows subcarriers to be shared by multiple users. Not only does SCMA help provide higher data rates but also increase the number of accommodated users. In this context, a power optimization and codebook allocation problems are formulated and solved using the water-filling and heuristic approaches, respectively. Results show that SCMA can significantly improve data rate provision and accommodate more mobile users with improved user satisfaction.

Edge computing provides cloud-like services at the network edge near mobile users. Due to the prosperity of smart applications that involve computing-intensive tasks, edge devices are intended to provide sufficient amounts of resources in order to accommodate the increasing computing demands. However, computing resources could also suffer being underutilized which leads to both resource and energy wastage. In this paper, heterogeneous virtual machine (VM) allocation in edge computing is considered to cope with the different computing demands at each edge device. To this end, an unsupervised machine learning technique, namely, the K-means is used to cluster incoming tasks into three different categories according to their processing requirements. Afterwards, tasks belonging to each cluster will be allocated the appropriate type of VMs to better utilize the computing resources. Results show the effectiveness of the proposed scheme in clustering computing tasks and improving resource utilization in edge devices.

In this paper, a supervised machine learning approach, namely, the decision tree is used to classify IoT applications based on their delay requirements. The decision-tree is trained and tested using simulated datasets to classify tasks into delay-sensitive and delay-insensitive based on the application features such as type and location. Delay-sensitive tasks are generally related to applications such as medical, manufacturing, and connected vehicles that require high service quality and short response time. Once delay-sensitive tasks are recognized, a prioritized scheduling mechanism is implemented to reduce the queueing delay at edge devices. Here, a two-class priority queueing system is used to model the scheduling mechanism at the edge device. Results show the effectiveness of machine learning in identifying delay-sensitive tasks that will experience shorter queueing delay at the edge device to enable high quality edge computing services.

In this paper, an edge device sleeping mechanism is proposed to save energy in cooperative edge computing networks. Energy saving in the proposed scheme is obtained by implementing a software- defined networking (SDN)-assisted interactive On/Off operation on edge devices taking into account the quality of service (QoS) experienced by end-users. First, an optimization problem is formulated to reduce the number of active edge devices under the queueing probability constraint. Herein, edge devices are modeled as M/M/k queueing systems, whereas the square-root staffing rule is used to maintain the queueing probability below desired levels. Then, a load balancing mechanism is carried out to reduce the variations in resource utilization among edge devices. To this end, a discrete-time Markov chain (DTMC)-based algorithm is implemented to achieve the intended load balancing. Results show the effectiveness of the proposed scheme in achieving energy saving and maintaining the queueing delay at controlled levels.

Fully cloudified mobile network infrastructure, which is featured by the joint deployment of heterogeneous cloud radio access networks and edge computing, will successfully cope with the data deluge by densely deploying virtualized wireless base stations and servers while providing the system design with high flexibility, reliability, availability, and scalability. On the other hand, the massive replication of the wireless and computing infrastructure will significantly increase the energy footprint to prohibitive levels. In order to gain actionable insights on energy-efficiency for a fully cloudified mobile network infrastructure, this paper first presents a comprehensive survey of the recent research breakthroughs on each building block of the system, namely: remote radio heads, baseband unit pool, fronthaul, backhaul, HetNet, and edge and cloud computing. Next, we consolidate the discussion with the challenges and open issues of a joint operation.

Future 5G networks are expected to have the capabilities of providing extremely high data rates, seamless coverage, massive number of connected devices, low latency, etc., in order to support the internet of things era. The dynamic performance of 5G networks is a key feature for controlling the dense and rapidly changing communication environment. Technical issues such as limited frequency resources, interference, energy consumption, and network management are the main challenges facing 5G networks. This article presents a comprehensive study of 5G networks architecture, technologies, challenges, and possible solutions based on recent advances in technology and research.

The massive numbers of connected devices in the IoT era impose a real challenge in the management of both communication and computing resources. Moreover, the competition of those devices on the limited resources will inherently raise the delay experienced by users. To this end, we propose a priority service provision scheme to reduce the latency experienced by delay-sensitive services. Here, incoming tasks are classified into delay-sensitive and delay-insensitive whereby priority classes are assigned using a matching theory approach. Then, the queue delay experienced by each class is investigated at the computing node i.e., the edge device, and the communication node i.e., the small base station (SBS). To maintain high quality of experience (QoE) in regard with time delay for all tasks, a dynamic priority scheme is proposed and controlled using a heuristic algorithm. The goal of the dynamic priority scheme is to minimize the delay at the communication node (SBS) for users requesting non-computing tasks (e.g., regular phone calls) by promoting their class when the delay exceeds a threshold value. The combined delay experienced at both communication and computing nodes is compared using the prioritized, non-prioritized, and the dynamic priority schemes. Results show that undertaking a dynamic priority service provision can achieve significant reduction in the amount of delay experienced by users.

In this paper, a base station sleeping mechanism is proposed for cloud-fog computing networks. Motivated by the capability of heterogeneous cloud radio access networks (H-CRANs) to efficiently control all network nodes, we aim to minimize the power consumption of small base stations (SBSs) taking into account the delay incurred by tasks offloaded from sleeping SBSs to the central cloud. In the proposed model, computing tasks can be processed either at the network edge (i.e., SBS) or at the central cloud. All computing and non-computing tasks of sleeping SBSs are offloaded to the macro base station (MBS), while computing tasks are further offloaded to the cloud. The probabilities of queueing in the MBS and the cloud given that a particular SBS is sleeping are calculated prior to making the decision of SBS sleeping. Therefore, according to whether more power saving or less delay is preferred, SBSs that impose less queueing probability on the MBS or cloud are forced to enter the sleep mode, respectively. Results show that taking computing tasks into consideration in SBS sleeping can reduce the computing response time at the cloud.

In this paper, a user association and bandwidth allocation approach is proposed for heterogeneous networks (HetNets) using fuzzy logic controllers. Due to the heterogeneous nature of user demands, we categorize the incoming mobile users into low, medium, or high based on their data rate requirements. Similarly, the bandwidth utilization in a given small cell base station (SBS) is quantified and evaluated. Based on the per user demand and bandwidth availability, the controller decides whether a particular user should be associated with that SBS or offloaded to the macro base station (MBS). Moreover, the controller adjusts the fraction of bandwidth allocated to each user based on users data rate requirement and availability of resources towards maximizing the total data rate in the network. The proposed scheme, which is performed by SBSs in a distributed manner, is investigated and compared with two other approaches; namely, the best signal-to-interference-plus-noise ratio (SINR) which is considered as the baseline approach in the literature, and a greedy-based approach where priority in association is given to users demanding higher data rates. Our approach shows promising results regarding the improvement of data rate, bandwidth utilization, and blocking ratio, with an increased number of offloaded users.

The architecture of cloud radio access networks (C-RANs) is envisioned as an attractive paradigm of 5G that takes advantages of both centralized baseband and coordinated multi-point(CoMP) processing in radio access networks. In C-RAN the data rate provisioning can be significantly improved due to the fractional frequency reuse performed by small cells. The dense deployment of small cells, however, incurs severe inter-tier and inter-cell interference turning the user association into a more challenging problem. Moreover, multi-cell association problem occurs in CoMP and control/user planes (C/U planes) splitting based C-RAN system where users are associated with more than one cells to support joint-transmission and reception method. In this paper, we consider multi-cell user association approach taking into account the data rate and aggregated interference of mobile users. We propose the posterior probability based user association and power allocation (P2UPA) method that depends on prior knowledge of the channel state information (CSI). The objective of the proposed method is to maximize the sum data rate of small cell users while maintaining the constraints of aggregated interference, power consumption, and data rate among small cell users. Finally, the sum data rate and energy efficiency performance of P2UPA are evaluated through simulations.

Providing seamless wireless coverage is essential for maintaining reliable communications in cellular networks. However, in extraordinary circumstances such as catastrophes, cellular infrastructure could be severely damaged; therefore, coverage is lost. For this reason, device-to-device (D2D) communication emerged as a promising technology to maintain wireless connectivity especially for search and rescue missions. This paper presents a review on recent advances in D2D communications regarding public safety applications such as search and rescue missions, coverage extension, and road safety. In search and rescue missions, discovery of devices in impacted areas can be achieved by other devices that have access to the cellular network. Moreover, the close proximity among mobile devices enables high data rate provisioning thus allowing rescue teams to send real-time high quality video and image reports to control centers. Cellular coverage can also be extended using relays such as drones that could play a significant role in coverage extension in extraordinary conditions. Furthermore, the safety of roads can be highly improved with vehicle-type communications, whereby information among vehicles are exchanged to avoid collisions.

In emergency situations such as catastrophes, the cellular infrastructure can be partially or totally damaged. For rescue teams to transmit and receive on-sight reports using live video and high quality images, the provisioning of seamless coverage with high data rates is inevitable. As cellular networks continue to evolve into more heterogeneous structures regarding the size of deployed cells, the implementation of traffic offloading strategies is essential to achieve high data rates and more power saving. In regard to network modeling, stochastic geometry has been considered in the literature as an efficient approach for modeling the locations of base stations (BSs). Moreover, stochastic geometry provides tractable mathematical expressions for key network parameters such as coverage probability and average data rate. In heterogeneous networks (HetNets), density and power of BSs in a given tier influence the volume of traffic offloaded to other tiers. However, these two parameters are generally constrained together in practical scenarios. In other words, given a BS density, the power per BS must be adjusted to satisfy the inter-tier traffic flow to maximized system throughput. Furthermore, in D2D networks the fraction of spectrum allocated to D2D links also affects the net achieved throughput. In this paper, we formulate the problem of jointly optimizing the power of small BSs (SBSs), SBS density, and the fraction of spectrum allocated for D2D communications in order to maximize system throughput for critical conditions such as rescue missions.

This paper investigates the problem of user‐centric cooperative edge caching in content delivery networks to leverage service provisioning at the network edge and to improve the quality of...