Edge Resources Research Articles

On Optimizing Resources for Real‐Time End‐to‐End Machine Learning in Heterogeneous Edges

ABSTRACTDeploying end‐to‐end ML applications on edge resources becomes a viable solution to achieve performance and data regulations. With the microservice architecture, these applications can scale dynamically, improving service availability under dynamic workloads. However, orchestrating multiple end‐to‐end ML applications within heterogeneous edge environments must deal with numerous challenges while sharing computing resources. Prevalent orchestration tools/frameworks supporting edge ML serving are inefficient in provisioning methods due to constrained resources, diverse resource demands and utilization patterns. In this work, we present a provisioning method to optimize resource utilization for end‐to‐end ML applications on a heterogeneous edge. By profiling all microservices within the application, we estimate scales and allocate them on desired hardware platforms with sufficient resources when considering their runtime utilization patterns. We also provide several practical analyses on runtime monitoring metrics to detect and mitigate resource contentions, guaranteeing performance. The experiments with three real‐world ML applications demonstrate the practicality of our method on a heterogeneous edge cluster of Raspberry Pis and Jetson Developer Kits.

GP4ESP: a hybrid genetic algorithm and particle swarm optimization algorithm for edge server placement

Edge computing has attracted wide attention due to its ultra-low latency services, as well as the prevalence of smart devices and intelligent applications. Edge server placement (ESP) is one of the key issues needed to be addressed for effective and efficient request processing, by deciding which edge stations to equip with limited edge resources. Due to NP-hardness of ESP, some works have designed meta-heuristic algorithms for solving it. While these algorithms either exploited only one kind of meta-heuristic search strategies or separately perform two different meta-heuristic algorithms. This can result in limit performance of ESP solutions due to the “No Free Lunch” theorem. In addition, existing algorithms ignored the computing delay of edge servers (ESs) on request process, resulting in overestimation of the service quality. To address these issues, in this article, we first formulate ESP problem with the objective of minimizing the overall response time, considering heterogeneous edge servers with various service capacity. Then, to search effective or even the best ESP solutions, we propose a hybrid meta-heuristic algorithm (named GP4ESP) by taking advantage of both the powerful global search ability of genetic algorithm (GA) and the fast convergence of particle swarm optimization (PSO). GP4ESP effectively fuses the merits of GA and PS by integrating the swarm cognition of PSO into the evolutionary strategy of GA. At last, we conducted extensive simulation experiments to evaluate the performance of GP4ESP, and results show that GP4ESP achieves 18.2%–20.7% shorter overall response time, compared with eleven up-to-date ESP solving algorithms, and the performance improvement is stable as the scale of ESP is varied.

Optimizing point-of-sale services in MEC enabled near field wireless communications using multi-agent reinforcement learning

In the next-generation communication system, near-field communication (NFC) is a key enabler of contactless transactions, including mobile payments, ticketing, and access control. With the growing demand for contactless solutions, NFC technology will play a pivotal role in enabling secure and convenient payment experiences across various sectors. In contrast, Internet of Things (IoT) devices such as phones’ Point of Sale (PoS) constitute limited battery life and finite computational resources that act as a bottleneck to doing the authentication in a minimal amount of time. Because of this, it garnered considerable attention in both academic and industrial realms. To overcome this, in this work we consider the Multiple Mobile Edge Computing (MEC) as an effective solution that provides extensive computation to PoS connected to it. To address the above, this work considers the PoS-enabled multi-MEC network to guarantee NFC communication reliably and effectively. For this, we formulate the joint optimization problem to maximize the probability of successful authentication while minimizing the queueing delay by jointly optimizing the computation and communication resources by utilizing a multi-agent reinforcement learning optimization approach. Through extensive simulations based on real-world scenarios, the effectiveness of the proposed approach was demonstrated. The results demonstrate that adjusting the complexity and learning rates of the model, coupled with strategic allocation of edge resources, significantly increased authentication success rates. Furthermore, the optimal allocation strategy was found to be crucial in reducing latency and improving authentication success by approximately 9.75%, surpassing other approaches. This study highlights the importance of resource management in optimizing MEC systems, paving the way for advancements in establishing secure, efficient, and dependable systems within the Internet of Things framework.

Enhancing driver attention and road safety through EEG-informed deep reinforcement learning and soft computing

This paper introduces a transformative edge computing-based approach for enhancing driver attention and road safety using EEG-driven deep reinforcement learning (DRL). As driver inattention remains a significant factor in accidents, real-time cognitive state monitoring enabled by in-vehicle edge devices offers new promise. Our method leverages EEG data collected from drivers using headsets, analyzing signals related to visual attention. Edge computing resources in the vehicle extract features and classify attention levels using Deep Q-Network (DQN) and Proximal Policy Optimization (PPO) models trained to approximate optimal driving decisions. A novel reward structure combining driving performance and attention guides the models. Our edge computing-powered framework reacts within critical time latencies to maximize attention through interventions adapting to the driving environment. Results demonstrate the effectiveness of this approach, with PPO agent on edge devices achieving high average rewards up to 489,752.4 and 99.3% reward as accuracy in classifying attention states, thereby significantly outperforming traditional methods. This underscores edge computing’s potential to enable real-time integration of neuroscience and AI, advancing road safety. The edge resources deliver time-critical analysis and adaptation, while connectivity to the fog and cloud allows optimizing and learning at scale across populations. This research pioneers a new epoch for road safety powered by edge intelligence.

Multi-agent reinforcement learning for task offloading with hybrid decision space in multi-access edge computing

Multi-access Edge Computing (MEC) has become a significant technology for supporting the computation-intensive and time-sensitive applications on the Internet of Things (IoT) devices. However, it is challenging to jointly optimize task offloading and resource allocation in the dynamic wireless environment with constrained edge resource. In this paper, we investigate a multi-user and multi-MEC servers system with varying task request and stochastic channel condition. Our purpose is to minimize the total energy consumption and time delay by optimizing the offloading decision, offloading ratio and computing resource allocation simultaneously. As the users are geographically distributed within an area, we formulate the problem of task offloading and resource allocation in MEC system as a partially observable Markov decision process (POMDP) and propose a novel multi-agent deep reinforcement learning (MADRL) -based algorithm to solve it. In particular, two aspects have been modified for performance enhancement: (1) To make fine-grained control, we design a novel neural network structure to effectively handle the hybrid action space arisen by the heterogeneous variables. (2) An adaptive reward mechanism is proposed to reasonably evaluate the infeasible actions and to mitigate the instability caused by manual configuration. Simulation results show the proposed method can achieve 7.12%−20.97% performance enhancements compared with the existing approaches.

Resource Allocation Based on Task Priority and Resource Consumption in Edge Computing

The computational power of Internet of Things (IoT) devices is usually low, which makes it necessary to process data and extract relevant information on devices with higher processing capacity. Edge Computing emerged as a complementary solution to cloud computing, providing devices at the network edge with computational resources to handle the data processing and analysis that constrained IoT devices eventually cannot perform. This solution allows data processing closer to the IoT devices, reducing latency for IoT applications. However, the resource constraints of edge nodes, which have lower computational power than the cloud nodes, make resource allocation and processing massive requests challenging. This study proposes an edge resource allocation mechanism based on task priority and machine learning. The proposed approach efficiently allocates resources for IoT requests based on their task priorities while monitoring the resource consumption of edge nodes. This study evaluates the performance of different classification algorithms by using well-known metrics for classifying models. The most efficient classifier achieved an accuracy of 92% and a precision of 90%. The results indicate good performance when using this classifier in the evaluated approach. The proposed mechanism demonstrated that resource management can be done more efficiently with significantly lower resource utilization when compared to an allocation method based only on distance. The study tested different scenarios regarding the number of requests, edge nodes, and a proposed failure mechanism to schedule failed node tasks to functional nodes. This failure control mechanism is a significant contribution of the proposal. Therefore, the proposed method in this study can become a valuable tool for efficient resource management with reduced computational cost and efficient resource allocation.

A joint optimization of resource allocation management and multi-task offloading in high-mobility vehicular multi-access edge computing networks

Vehicular communications have advanced data exchange and real-time services in intelligent transportation systems by exploiting advanced communication between vehicles and infrastructure. The emergence of Multi-access Edge Computing (MEC) has further elevated this field by utilizing distributed edge resources near vehicles for low-latency data processing and high-reliability communication. In this dynamic environment, adequate resource allocation and task offloading are pivotal to ensure superior performance, lower latency, and efficient network resource utilization, enhancing Quality of Service (QoS) and overall driving experience and safety. This paper presents a developed vehicular network and offloading mechanism, introducing a resource management model with real-time allocation and load balancing. The proposed method integrates task prioritization, multi-agent collaboration, context-aware decision-making, and distributed learning to optimize network performance. The introduced optimized algorithm initializes Q-networks and target networks, sets up an experience replay buffer, and configures agents with local state representations. Agents use an ε-greedy policy for action selection, update Q-values through experience replay, and prioritize tasks based on urgency while sharing state information for collaborative decision-making. Evaluations through simulation demonstrate optimized performance, enhancing efficiency in vehicular MEC networks compared to baseline and the other well-known algorithms.

An online cost optimization approach for edge resource provisioning in cloud gaming

Cloud gaming (CG), as an emergent computing paradigm, is revolutionizing the gaming industry. Currently, cloud gaming service providers (CGSPs) begin to integrate edge computing with cloud to provide services, with the aim of maximizing gaming service revenue while considering the costs incurred and the benefits generated. However, it is non-trivial to maximize gaming service revenue, as future requests are not known beforehand, and poor resource provisioning may result in exorbitant costs. In addition, the edge resource provisioning (ERP) problem in CG necessitates a trade-off between cost and inevitable queuing issues in CG systems. To address this issue, we propose ERPOL (ERP Online), a convenient and efficient approach to formulate ERP strategies for CGSPs, without requiring any future information. The performance of ERPOL has been theoretically validated and experimentally evaluated. Experiments driven by real-world traces show that it can achieve significant cost savings. The proposed approach has the potential to transform how CGSPs manage their infrastructure.

Collaborative learning-based inter-dependent task dispatching and co-location in an integrated edge computing system

Recently, several edge deployment types, such as on-premise edge clusters, Unmanned Aerial Vehicles (UAV)-attached edge devices, telecommunication base stations installed with edge clusters, etc., are being deployed to enable faster response time for latency-sensitive tasks. One fundamental problem is where and how to offload and schedule multi-dependent tasks so as to minimize their collective execution time and to achieve high resource utilization. Existing approaches randomly dispatch tasks naively to available edge nodes without considering the resource demands of tasks, inter-dependencies of tasks and edge resource availability. These approaches can result in the longer waiting time for tasks due to insufficient resource availability or dependency support, as well as provider lock-in. Therefore, we present EdgeColla, which is based on the integration of edge resources running across multi-edge deployments. EdgeColla leverages learning techniques to intelligently dispatch multi-dependent tasks, and a variant bin-packing optimization method to co-locate these tasks firmly on available nodes to optimally utilize them. Extensive experiments on real-world datasets from Alibaba on task dependencies show that our approach can achieve optimal performance than the baseline schemes.

Reinforcement-Learning-Assisted Service Function Chain Embedding Algorithm in Edge Computing Networks

Edge computing networks are critical infrastructures for processing massive data and providing instantaneous services. However, how to efficiently allocate resources in edge computing networks to meet the embedding requirements of service function chains has become an urgent problem. In this paper, we model the resource allocation problem in edge computing networks as a service function chain embedding problem model, aiming to optimize the resource allocation through reinforcement learning algorithms to achieve the goals of low cost, high revenue, and high embedding rate. In this paper, the basic concepts of edge computing network and service function chain are elaborated, and the resource allocation problem is transformed into a service function chain embedding problem by establishing a mathematical model, which provides a foundation for the subsequent algorithm design. In this paper, a service function chain embedding algorithm based on reinforcement learning is designed to gradually optimize the resource allocation decision by simulating the learning process. In order to verify the effectiveness of the algorithm, a series of simulation experiments are conducted in this paper and compared with other algorithms. The experimental results show that the service function chain embedding algorithm based on reinforcement learning proposed in this paper exhibits superior performance in resource allocation. Compared with traditional resource allocation methods, the algorithm achieves significant improvement in terms of low cost, high revenue, and high embedding rate.

Drift-based task management in support of pervasive edge applications

The Internet of Things (IoT) creates a sprawling network where multiple devices can interact, enabling a variety of devices to communicate, collect information, and carry out tasks in support of services that end users may enjoy. Edge Computing, known for its lower latency compared to the Cloud, has sparked interest in managing the execution of tasks within a huge ecosystem that acts as a cover upon the IoT infrastructure. The primary challenge lies in maximizing the use of limited edge resources while minimizing response time, prompting numerous research endeavors. However, existing efforts often overlook shifts in the collected data that may affect the execution of tasks and the production of knowledge. This paper focuses on developing a mechanism that considers data and concept drifts to optimize the management of tasks. The ultimate goal is to maximize the accuracy levels while optimizing resource utilization, through a tasks’ offloading scheme that accounts for distribution-based similarity, offering substantial benefits in managing these constrained resources efficiently. The shifts in data can help determine if the execution of a task is efficient in a specific node and become part of a reasoning model that guides the offloading decisions. Finally, we evaluate our model using a large set of experimental scenarios.

Optimizing Edge Resources in Intelligent Railway Construction: A Two-Level Game Approach

Optimizing Edge Resources in Intelligent Railway Construction: A Two-Level Game Approach

Reliablity and Security for Fog Computing Systems

Fog computing (FC) is a distributed architecture in which computing resources and services are placed on edge devices closer to data sources. This enables more efficient data processing, shorter latency times, and better performance. Fog computing was shown to be a promising solution for addressing the new computing requirements. However, there are still many challenges to overcome to utilize this new computing paradigm, in particular, reliability and security. Following this need, a systematic literature review was conducted to create a list of requirements. As a result, the following four key requirements were formulated: (1) low latency and response times; (2) scalability and resource management; (3) fault tolerance and redundancy; and (4) privacy and security. Low delay and response can be achieved through edge caching, edge real-time analyses and decision making, and mobile edge computing. Scalability and resource management can be enabled by edge federation, virtualization and containerization, and edge resource discovery and orchestration. Fault tolerance and redundancy can be enabled by backup and recovery mechanisms, data replication strategies, and disaster recovery plans, with a residual number system (RNS) being a promising solution. Data security and data privacy are manifested in strong authentication and authorization mechanisms, access control and authorization management, with fully homomorphic encryption (FHE) and the secret sharing system (SSS) being of particular interest.

Device-edge collaborative occluded face recognition method based on cross-domain feature fusion

The lack of facial features caused by wearing masks degrades the performance of facial recognition systems. Traditional occluded face recognition methods cannot integrate the computational resources of the edge layer and the device layer. Besides, previous research fails to consider the facial characteristics including occluded and unoccluded parts. To solve the above problems, we put forward a device-edge collaborative occluded face recognition method based on cross-domain feature fusion. Specifically, the device-edge collaborative face recognition architecture gets the utmost out of maximizes device and edge resources for real-time occluded face recognition. Then, a cross-domain facial feature fusion method is presented which combines both the explicit domain and the implicit domain facial. Furthermore, a delay-optimized edge recognition task scheduling method is developed that comprehensively considers the task load, computational power, bandwidth, and delay tolerance constraints of the edge. This method can dynamically schedule face recognition tasks and minimize recognition delay while ensuring recognition accuracy. The experimental results show that the proposed method achieves an average gain of about 21% in recognition latency, while the accuracy of the face recognition task is basically the same compared to the baseline method.

An offloading and pricing mechanism based on virtualization in edge–cloud computing

Integrate users, edge servers and cloud into a system can improve resource utilization and quality of service (QoS). The users can send the same request to multiple Edge Servers. However, edge servers have limited scope, which may result in repeated offloading tasks for the same user. In this paper, we consider the cloud resource migration in the form of virtual machines (VMs) to achieve high computing efficiency and low latency requirements. Obviously, the migration of VMs causes changes in the number and price of edge resources, which further affects the task offloading strategy. Therefore, it is necessary to optimize the VMs migration strategy and resource pricing strategy successively to achieve the optimal offloading strategy suitable for users. First, based on the competition between edge servers, a static VMs migration algorithm (SVMMA) is designed to formulate the strategy and price of VMs migration. Then, according to the users’ resource needs and the resource possession of different edge servers, a resource partitioning approach (RPA) have been proposed. Finally, two different QoS functions are evaluated, representing the QoS matching degree between tasks and multiple resource combination blocks (MRCBs). Wherein, a dynamic matching and pricing of tasks algorithm (DMPT) is conducted to obtain the optimal offloading strategy and price of the task. Experiments show that our algorithms can improve the offloading rate of tasks and achieve the relative balance among user, cloud server, and edge server benefits.

On incentivizing resource allocation and task offloading for cooperative edge computing

Cooperative edge computing serves as an effective solution to provide reliable and elastic edge computing services through pooling geographically proximate edge resources and efficiently allocating them to users. The incentive mechanism is critical for realizing cooperative edge computing. In this paper, we propose a virtual machine (VM) or container based resource allocation scheme and two market-based mechanisms to stimulate collaboration among edge servers (ESs). Our first mechanism is a novel two-stage double auction based approach where edge resources are allocated to services/VMs at the first stage and then distributed among ESs at the second stage via a two-sided multi-unit auction. To make it more practical, we use markups to set the ask and bid prices for the participants in the market. Moreover, with the concept of virtual sellers and virtual buyers, we cast the multi-unit auction as single-unit one so that the auction becomes tractable. We further propose a price-based decentralized mechanism in case there is no central authority. With an appropriately constructed objective function, we formulate the resource allocation task as a network utility maximization (NUM) problem and leverage the dual-based decomposition theory for a distributed solution. We theoretically prove that both mechanisms guarantee properties such as budget balance and individual rationality for all ESs. Numerical studies with real-world dataset are presented to demonstrate the superior performance of our mechanisms over baselines.

5G and edge: A reinforcement learning approach for Virtual Network Embedding with cost optimization and improved acceptance rate

5G technologies are fueling a revolution across numerous industries, including manufacturing, healthcare, and entertainment, by enabling the development and deployment of novel applications at the network’s edge. To meet the demanding service level agreements of these industries, a dynamic and adaptable infrastructure strategy that combines cloud and edge computing models is needed. This hybrid approach offers the benefits of both centralized cloud processing and decentralized edge computing, optimized for responsiveness and efficiency. A key element for success is an orchestration mechanism that dynamically allocates resources to ensure the infrastructure can adapt to fluctuating demands in real time, optimizing resource utilization and meeting SLA requirements. Among these mechanisms, virtual network embedding (VNE) and dynamic resource management (DRM) have emerged as tools for defining where and how edge technology should be used. However, current VNE approaches struggle to adapt to real-time fluctuations in demand across geographically distributed edge resources. This work introduces a novel resource allocation algorithm, the VNE-CRS, which uses an Artificial Intelligence technique called Reinforcement Learning to orchestrate resources across multiple domains. This approach benefits from the strength of Reinforcement Learning: its ability to consider the entire problem from beginning to end while incorporating various aspects of 5G Quality of Service Indicators for optimal decision-making. Experiments were conducted to evaluate the performance of VNE-CRS against state-of-the-art algorithms for multi-domain edge environments. Results have shown that employing Reinforcement Learning techniques for VNE resource allocation yields performance gains of 12.32 percentage points in comparison with the GRC algorithm and 28.80 percentage points in comparison with the base edge environment, presenting an acceptability rate closer to the Public Cloud environment with all benefits of edge environment. In conclusion, VNE-CRS offers an efficient solution for resource allocation in 5G environments, achieving superior performance and transforming the VNE architecture into a comprehensive orchestration system that optimizes infrastructure utilization for strategic long-term benefits.

A cloud-edge collaborative task scheduling method based on model segmentation

With the continuous development and combined application of cloud computing and artificial intelligence, some new methods have emerged to reduce task execution time for training neural network models in a cloud-edge collaborative environment. The most attractive method is neural network model segmentation. However, many factors affect the segmentation point, such as resource allocation, system energy consumption, load balancing, and network Bandwidth allocation. Some segmentation methods consider the shortest task execution time, which ignores the utilization of resources at the edge and can result in resource waste. Additionally, these factors are difficult to measure, which presents a challenge in calculating the best segmentation point to achieve the goal of maximum resource utilization and minimum task execution time. To solve this problem, this paper proposes a cloud-edge collaborative task scheduling method based on model segmentation (CECMS). This method first analyzes the factors affecting the segmentation point of the model and then obtains accurate factors that affect the segmentation point calculation through the pre-execution method. Furthermore, a multi-objective solution algorithm is improved to calculate the optimal model segmentation point. And tasks are separately offloaded to the edge and cloud based on the optimal model segmentation point. Finally, the experiments are conducted to verify the effectiveness of this method. Finally, the effectiveness of the CECMS method was verified through simulation experiments. Compared with the Dynamic Adaptive DNN Surgery (DADS) method and an adaptive DNN inference acceleration framework algorithm with end–edge–cloud collaborative computing algorithm (ADC), CECMS achieves the same effectiveness as DADS and ADC in optimizing task execution time by comprehensively considering the utilization of edge resources and minimizing task execution time, while also effectively ensuring resource utilization.

Harnessing idle edge resources for execution of cloud services: A comprehensive review

AbstractThis article presents a comprehensive investigation into voluntary resource sharing at the edge of the internet for cloud service execution. Through a systematic review, 460 scientific articles published between 2019 and 2023 were analyzed, compiling the primary recent contributions in the field. The study presents a taxonomy of the results and highlights the state of the art. Despite the growing interest in the topic in recent years due to the expansion of Edge and Fog Computing, numerous challenges and open research gaps were identified for the effective utilization of idle resources for cloud service execution at the edge. This paper delves into these challenges, emphasizing the need for extensive research in areas such as resource management, interoperability, incentive model establishment, and security and privacy concerns.

Risk Assessment Edge Contract for Efficient Resource Allocation

The rapid growth of edge devices and mobile applications has driven the adoption of edge computing to handle computing tasks closer to end-users. However, the heterogeneity of edge devices and their limited computing resources raise challenges in the efficient allocation of computing resources to complete services with different characteristics and preferences. In this paper, we delve into an edge scenario comprising multiple Edge Computing Servers (ECSs), multiple Device-to-Device (D2D) Edge Nodes (ENs), and multiple edge devices. In order to address the resource allocation challenge among ECSs, ENs, and edge devices in high-workload environments, as well as the pricing of edge resources within the resource market framework, we propose a Risk Assessment Contract Algorithm (RACA) based on risk assessment theory. The RACA enables ECSs to assess risks associated with local users by estimating their future revenue potential and updating the contract autonomously at present and in the future. ENs acquire additional resources from ECSs to efficiently complete local users’ tasks. Simultaneously, ENs can also negotiate reasonable resource requests and pricing with ECSs by a Stackelberg game algorithm. Furthermore, we prove the unique existence of Nash equilibrium in the established game, implying that equilibrium solutions can stably converge through computational methods in heterogeneous environments. Finally, through simulation experiments on the dataset, we demonstrate that risk assessment can better enhance the overall profit capability of the system. Moreover, through multiple experiments, we showcase the stability of the contract’s autonomous update capability. The RACA exhibits better utility in terms of system profit capabilities, stability in high-workload environments, and energy consumption. This work provides a more dynamic and effective solution to the resource allocation problem in edge systems under high-workload environments.