Service Placement Research Articles

Double Deep Q-Learning-Based Path Selection and Service Placement for Latency-Sensitive Beyond 5G Applications

Double Deep Q-Learning-Based Path Selection and Service Placement for Latency-Sensitive Beyond 5G Applications

Predictive Service Provisioning With Online Learning in Wireless Edge Networks

Mobile Edge Computing (MEC) technology can be implemented at cellular base stations, enabling flexible and configurable provisions of services for mobile users to access. Nevertheless, the conventional solutions mainly focus on <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">statical</i> service provisioning, which ignores the dynamic nature of the arriving service requests. In this work, we first conduct comprehensive data-driven observations on over 4 million service requests throughout 9,800 base stations. Our key findings suggest that users' demands intrinsically exhibit spatial and temporal patterns, which inevitably lead to performance degradation in statical service provisioning. Motivated by that, we design and implement MobiEdge, a predictive service provisioning system with online learning in wireless edge networks. We propose a graph embedding learning-based model for representation learning, thus to achieve accurate request prediction at different base stations. Then, based on the prediction of incoming service requests, we study the service provisioning reconfiguration problem, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i.e.</i> , how to jointly optimize service placement and corresponding request scheduling across dual timescales, under constraints of network resources and the total budget. By leveraging the submodular technique, we transform the research issue into a submodular function maximization problem under the <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$q$</tex-math></inline-formula> -independence system constraint, where <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$q$</tex-math></inline-formula> is a positive constant related to the ratio of coefficients in constraint conditions. On this basis, we propose a <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$1/(1+q)$</tex-math></inline-formula> approximation algorithm with rigorous theoretical analysis on the bounded maximum utility. Extensive trace-driven evaluations are conducted over networks of different scales, and MobiEdge shows remarkable performance enhancements by achieving the accuracy of up to 98% in service prediction and an average utility of 92.9% to the optimal solution in service provisioning.

Taming latency at the edge: A user-aware service placement approach

Modern network and computing infrastructures are tasked with addressing the stringent demands of today’s applications. A pivotal concern is the minimization of latency experienced by end-users accessing services. While emerging network architectures provide a conducive setting for adept orchestration of microservices in terms of reliability, self-healing and resiliency, assimilating the awareness of the latency perceived by the user into placement decisions remains an unresolved problem. Current research addresses the problem of minimizing inter-service latency without any guarantee to the level of latency from the end-user to the cluster. In this research, we introduce an architectural approach for scheduling service workloads within a given cluster, prioritizing placement on the node that offers the lowest perceived latency to the end-user. To validate the proposed approach, we propose an implementation on Kubernetes (K8s), currently one of the most used workload orchestration platforms. Experimental results show that our approach effectively reduces the latency experienced by the end-user in a finite time without degrading the quality of service. We study the performance of the proposed approach analyzing different parameters with a particular focus on the size of the cluster and the number of replica pods involved to measure the latency. We provide insights on possible trade-offs between computational costs and convergence time.

Presenting a meta-heuristic solution for optimal resource allocation in fog computing

Given that cloud computing is a relatively new field of study, there is an urgent need for comprehensive approaches to resource provisioning and the allocation of Internet of Things (IoT) services across cloud infrastructure. Other challenging aspects of cloud computing include IoT resource virtualization and disseminating IoT services among available cloud resources. To meet deadlines, optimize application execution times, efficiently use cloud resources, and identify the optimal service location, service placement plays a crucial role in installing services on existing virtual resources within a cloud-based environment. To achieve load balance in the fog computing infrastructure and ensure optimal resource allocation, this work proposes a meta-heuristic approach based on the cat swarm optimization method. For more clarity in the difference between the work presented in this research and other similar works, we named the proposed technique MH-CSO. The algorithm incorporates a resource check parameter to determine the accessibility and suitability of resources in different situations. This conclusion was drawn after evaluating the proposed solution in the ifogsim environment and comparing it with particle swarm and ant colony optimization techniques. The findings demonstrate that the proposed solution successfully optimizes key parameters, including runtime and energy usage.

Pro-active component image placement in Edge computing environments

Edge computing has attracted a lot of attention both from industry and academia in recent years and is considered as a key enabler for addressing the increasingly strict requirements of Next Generation applications. Contrary to Cloud computing, in Edge computing the computation are placed closer to the end-users into the so-called Edge, to facilitate low-latency and high-bandwidth applications and services that would not be feasible using cloud and far remote processing alone. However, the distributed, dynamic and heterogeneous environment in the Edge computing along with the diverse applications’ requirements make service placement in such infrastructure a challenging issue. One important aspect of Edge computing is the management of the placement of the applications in the network system so as to minimize each application’s runtime, given the resources of system’s devices and the capabilities of the system’s network. To this end, we propose an empirical experimental analysis, by comparing the results of different placements strategies and various edge communication networks. In particular, we model the problem of proactive placement of application images as a Minimum Vertex Cover problem. Our results demonstrate that the Greedy implementation seems to offer the best tradeoff in terms of performance, cost function and execution time.

Online Learning Aided Decentralized Multi-User Task Offloading for Mobile Edge Computing

Mobile edge computing facilitates users to offload computation tasks to edge servers for meeting their stringent delay requirements. Previous works mainly explore task offloading when system-side information is given (e.g., server processing speed, cellular data rate), or centralized offloading under system uncertainty. But both generally fall short of handling task placement involving many coexisting users in an uncertain environment. In this paper, we develop a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">multi-user</i> offloading framework considering <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">unknown yet stochastic</i> system-side information to enable a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">decentralized user-initiated</i> service placement under overlapping server coverage. Specifically, we formulate the dynamic task placement as an online multi-user multi-armed bandit process, and propose a decentralized epoch based offloading (DEBO) to optimize user rewards which are subjected under network delay. We consider both cases without and with neighboring edge feedback once users' tasks are processed, where the latter incorporates system-side information sharing among edge servers for an enhanced task placement. For both cases, we show that DEBO can gradually deduce the optimal user-server assignment during dynamic offloading, thereby achieving a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">close-to-optimal</i> service performance and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">tight</i> <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$O(\log _{2} T)$</tex-math></inline-formula> regret. Moreover, we generalize DEBO to various common scenarios such as unknown reward gap, dynamic entering or leaving of clients, and fair reward distribution, while further exploring when users' offloaded tasks require <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">heterogeneous</i> computing resources. Particularly, we accomplish a sub-linear regret for each of these instances. Real measurements based evaluations corroborate the superiority of our offloading schemes over state-of-the-art approaches in optimizing delay-sensitive rewards.

Elite‐guided equilibrium optimiser based on information enhancement: Algorithm and mobile edge computing applications

AbstractThe Equilibrium Optimiser (EO) has been demonstrated to be one of the metaheuristic algorithms that can effectively solve global optimisation problems. Balancing the paradox between exploration and exploitation operations while enhancing the ability to jump out of the local optimum are two key points to be addressed in EO research. To alleviate these limitations, an EO variant named adaptive elite‐guided Equilibrium Optimiser (AEEO) is introduced. Specifically, the adaptive elite‐guided search mechanism enhances the balance between exploration and exploitation. The modified mutualism phase reinforces the information interaction among particles and local optima avoidance. The cooperation of these two mechanisms boosts the overall performance of the basic EO. The AEEO is subjected to competitive experiments with state‐of‐the‐art algorithms and modified algorithms on 23 classical benchmark functions and IEE CEC 2017 function test suite. Experimental results demonstrate that AEEO outperforms several well‐performing EO variants, DE variants, PSO variants, SSA variants, and GWO variants in terms of convergence speed and accuracy. In addition, the AEEO algorithm is used for the edge server (ES) placement problem in mobile edge computing (MEC) environments. The experimental results show that the author’s approach outperforms the representative approaches compared in terms of access latency and deployment cost.

QoS-aware edge AI placement and scheduling with multiple implementations in FaaS-based edge computing

Resource constraints on the computing continuum require that we make smart decisions for serving AI-based services at the network edge. AI-based services typically have multiple implementations (e.g., image classification implementations include SqueezeNet, DenseNet, and others) with varying trade-offs (e.g., latency and accuracy). The question then is how should AI-based services be placed across Function-as-a-Service (FaaS) based edge computing systems in order to maximize total Quality-of-Service (QoS). To address this question, we propose a problem that jointly aims to solve (i) edge AI service placement and (ii) request scheduling. These are done across two time-scales (one for placement and one for scheduling). We first cast the problem as an integer linear program. We then decompose the problem into separate placement and scheduling subproblems and prove that both are NP-hard. We then propose a novel placement algorithm that places services while considering device-to-device communication across edge clouds to offload requests to one another. Our results show that the proposed placement algorithm is able to outperform a state-of-the-art placement algorithm for AI-based services, and other baseline heuristics, with regard to maximizing total QoS. Additionally, we present a federated learning-based framework, FLIES, to predict the future incoming service requests and their QoS requirements. Our results also show that our FLIES algorithm is able to outperform a standard decentralized learning baseline for predicting incoming requests and show comparable predictive performance when compared to centralized training.

Cost-aware Service Placement and Scheduling in the Edge-Cloud Continuum

The edge to data center computing continuum is the aggregation of computing resources located anywhere between the network edge (e.g., close to 5G antennas), and servers in traditional data centers. Kubernetes is the de facto standard for the orchestration of services in data center environments, where it is very efficient. It, however, fails to give the same performance when including edge resources. At the edge, resources are more limited, and networking conditions are changing over time. In this article, we present a methodology that lowers the costs of running applications in the edge-to-cloud computing continuum. This methodology can adapt to changing environments, e.g., moving end-users. We are also monitoring some Key Performance Indicators of the applications to ensure that cost optimizations do not negatively impact their Quality of Service. In addition, to ensure that performances are optimal even when users are moving, we introduce a background process that periodically checks if a better location is available for the service and, if so, moves the service. To demonstrate the performance of our scheduling approach, we evaluate it using a vehicle cooperative perception use case, a representative 5G application. With this use case, we can demonstrate that our scheduling approach can robustly lower the cost in different scenarios, while other approaches that are already available fail in either being adaptive to changing environments or will have poor cost-effectiveness in some scenarios.

Intelligent decision making for energy efficient fog nodes selection and smart switching in the IOT: a machine learning approach.

With the emergence of Internet of Things (IoT) technology, a huge amount of data is generated, which is costly to transfer to the cloud data centers in terms of security, bandwidth, and latency. Fog computing is an efficient paradigm for locally processing and manipulating IoT-generated data. It is difficult to configure the fog nodes to provide all of the services required by the end devices because of the static configuration, poor processing, and storage capacities. To enhance fog nodes' capabilities, it is essential to reconfigure them to accommodate a broader range and variety of hosted services. In this study, we focus on the placement of fog services and their dynamic reconfiguration in response to the end-device requests. Due to its growing successes and popularity in the IoT era, the Decision Tree (DT) machine learning model is implemented to predict the occurrence of requests and events in advance. The DT model enables the fog nodes to predict requests for a specific service in advance and reconfigure the fog node accordingly. The performance of the proposed model is evaluated in terms of high throughput, minimized energy consumption, and dynamic fog node smart switching. The simulation results demonstrate a notable increase in the fog node hit ratios, scaling up to 99% for the majority of services concurrently with a substantial reduction in miss ratios. Furthermore, the energy consumption is greatly reduced by over 50% as compared to a static node.

Online two-timescale service placement for time-sensitive applications in MEC-assisted network: A TMAGRL approach

Mobile edge computing (MEC) integrated with the Network Functions Virtualization (NFV) technique has been regarded as a promising solution for flexible services provision and user service experience improvement. However, existing service placement in such systems still faces the challenge of satisfying computing tasks with strict latency requirements, especially when massive mobile users roam around different coverage areas of edge servers. For this purpose, we first adopt a novel service placement framework that combines proactive replicas pre-deployment and reactive service migration. Based on this, we investigate the dynamic placement problem of multiple types of services achieved by the various virtualized network functions (VNFs) to minimize long-term redeployment costs in MEC-assisted systems, subject to the completion deadline of tasks and limited computing resources of edge servers. Considering that the update timescale of VNF replicas pre-deployment is different, we design a novel two-timescale multi-agent graph convolutional network-based reinforcement learning algorithm (TMAGRL) by invoking a long-timescale training layer for proactive VNF replicas placement and a short-timescale training layer for reactive VNF migration. Extensive numerical results reveal that TMAGRL, based on the designed hybrid framework, can learn a VNF placement strategy to adapt to the dynamics of the system without any prior information. Moreover, we verify its superior performance in terms of average service response latency and overall redeployment cost by comparing it with baselines.

MADDPG-Based Joint Service Placement and Task Offloading in MEC Empowered Air–Ground Integrated Networks

MADDPG-Based Joint Service Placement and Task Offloading in MEC Empowered Air–Ground Integrated Networks

Resource-aware fog service placement using deferred acceptance in edge computing

Fog computing aims at reducing network latency, bandwidth and processing delay by bringing computation resources closer to the data sources. Modern IoT applications, such as smart surveillance systems, remote health monitoring applications, ambient-aware applications etc., are time-critical and computationally intensive. For such applications, it is ideal to consider an intermediate layer of processing i.e a fog layer in addition to cloud. However, processing the workload of such IoT applications with an intermediate layer of fog nodes has several challenges. The significant challenges include the mobility of end-user nodes, dynamic patterns of workload generation leading to highly-variant resource requirements and sensitive quality of service (QOS) requirements. Hence a dynamic approach is required to precisely map the service requests from end-user devices to a suitable fog node. The optimality of a fog node for service placement depends on several factors, including location, resource utilization rates such as CPU, Memory and Disk utilization % and energy consumption. Several meta-heuristic techniques and evolutionary algorithms have been proposed in the past for optimal service placement among fog nodes. This paper investigates the applicability of a game theory-based approach called Deferred acceptance (DA) for the same. The DA algorithm, originally designed as a stable matching mechanism for the marriage market is adapted for service placement problems in fog-based IoT systems. The effectiveness of optimized service placement through the DA approach is validated experimentally. The experiment is carried out by creating a test bed that mimics a fog computing-based intelligent traffic management application. The proposed approach's performance is evaluated through parameters such as the startup time for the DA approach to produce outcomes, total execution time and handshake latency. The numerical results of the DA approach are compared with those of a simple scheduling algorithm ( FIFO) and a basic genetic algorithm-based approach. The DA approach appears to perform on par with the genetic algorithm in terms of start-up time, but exceeds the genetic algorithm with respect to stability of the outcomes produced.

Provisioning edge computing services in urban areas through the chaining of unmanned aerial vehicles: A genetic approach

AbstractEdge computing has been proved an efficient approach to provisioning computation offloading service to vehicles on road through Road‐Side Units (RSUs). However, the traffic volume on road is highly dynamic, while RSU‐based edge servers are static in terms of geographical location and computation capacity. To address this problem, this paper proposes a mobile edge server placement strategy using cruising UAVs along the roads based on the genetic algorithm. A mathematical model is first built to characterize the deployment cost of these UAV‐mounted servers and their routes. Next, a heuristic UAV‐mounted edge server deployment scheme based on K‐medoid clustering and genetic algorithms is designed. Experimental results verify that the proposed UAV deployment scheme satisfies the offloading demand of IoV nodes while reducing the total deployment cost by 17.05–48.94% compared with existing popular approaches.

The world's most remote community service? Partnership working and building community capacity on the British Overseas Territory Island of St Helena

Measuring just 47 square miles and 1950 km from its nearest neighbour, St Helena is a British Overseas Territory in the South Atlantic with a population of approximately 4500. Working in this remote setting is one of the world's smallest probation services delivering a range of interventions including community service. In this practice note the authors explore the sustained, successful partnership between St Helena Island Probation Service and the St Helena National Trust. This partnership not only provides meaningful community service placements which support vital conservation work, but simultaneously contributes to an individual's journey towards desistance through enhancing employability and supporting positive community reintegration.

Machine learning methods for service placement: a systematic review

With the growth of real-time and latency-sensitive applications in the Internet of Everything (IoE), service placement cannot rely on cloud computing alone. In response to this need, several computing paradigms, such as Mobile Edge Computing (MEC), Ultra-dense Edge Computing (UDEC), and Fog Computing (FC), have emerged. These paradigms aim to bring computing resources closer to the end user, reducing delay and wasted backhaul bandwidth. One of the major challenges of these new paradigms is the limitation of edge resources and the dependencies between different service parts. Some solutions, such as microservice architecture, allow different parts of an application to be processed simultaneously. However, due to the ever-increasing number of devices and incoming tasks, the problem of service placement cannot be solved today by relying on rule-based deterministic solutions. In such a dynamic and complex environment, many factors can influence the solution. Optimization and Machine Learning (ML) are two well-known tools that have been used most for service placement. Both methods typically use a cost function. Optimization is usually a way to define the difference between the predicted and actual value, while ML aims to minimize the cost function. In simpler terms, ML aims to minimize the gap between prediction and reality based on historical data. Instead of relying on explicit rules, ML uses prediction based on historical data. Due to the NP-hard nature of the service placement problem, classical optimization methods are not sufficient. Instead, metaheuristic and heuristic methods are widely used. In addition, the ever-changing big data in IoE environments requires the use of specific ML methods. In this systematic review, we present a taxonomy of ML methods for the service placement problem. Our findings show that 96% of applications use a distributed microservice architecture. Also, 51% of the studies are based on on-demand resource estimation methods and 81% are multi-objective. This article also outlines open questions and future research trends. Our literature review shows that one of the most important trends in ML is reinforcement learning, with a 56% share of research.

Edge server placement and allocation optimization: a tradeoff for enhanced performance

Edge server placement and allocation optimization: a tradeoff for enhanced performance

Intelligent service placement algorithm based on DDQN and prioritized experience replay in IoT-Fog computing environment

The widespread use of Internet of Things-based applications has led to the development of the fog computing paradigm, enabling seamless utilization of edge and cloud resources. The placement of IoT service requests at the appropriate fog nodes is one of the key problems in a fog-cloud environment, which affects the overall system performance and user experience. The challenges associated with service placement are the stochastic nature of service demand, heterogeneous and distributed characteristics of fog nodes. The objective of this work is to develop an effective service placement method to meet the varying resource requirements of services while optimizing two important performance parameters: service execution time and fog nodes’ energy consumption. Existing solutions are slow to adapt and neglecting the temporal patterns of service resource demand. This paper proposes a novel service placement algorithm based on double deep Q-Network and prioritized experience replay (DDQN-PER), which explores the optimal service placement strategy using the adaptive learning ability of deep reinforcement learning (DRL). The DRL agent learns the dynamic nature of the resource requirements of services to find an appropriate fog node for service placement. To evaluate the performance of the proposed algorithm, experiments are carried out using Measurement and Analysis on the WIDE Internet (MAWI) traffic traces. The proposed DDQN-PER algorithm, compared with other state-of-the-art methods and experimental results, show that the proposed method can minimize the service latency time and fog node’s energy consumption from 2.5% to 8% and 6% to 17%, respectively.

A Server Placement Algorithm for Reducing Risk and Improving Power Utilization in Data Centers

A Server Placement Algorithm for Reducing Risk and Improving Power Utilization in Data Centers

An Efficient and Autonomous Planning Scheme for Deploying IoT Services in Fog Computing: A Metaheuristic-Based Approach

The fog computing paradigm is a promising concept to overcome the exponential increase in data volume in Internet of Things (IoT) applications. This paradigm can support delay-sensitive IoT applications by extending cloud services to the network edge. However, fog computing faces challenges such as resource allocation for applications at the network edge due to limited resources as well as its heterogeneous and distributed nature. This is in line with the goals of microservice architecture and develops the placement of microservice-based IoT applications. The IoT service placement problem (SPP) on fog nodes is known as non-deterministic polynomial-time (NP)-hard. In this study, we introduce a meta-heuristic approach named SPP-differential evolution algorithm (DEA) to handle SPP, which originates from the DEA with a shared parallel architecture. The proposed method takes advantage of the scalable and deployable nature of microservices to minimize the resource utilization and delay as much as possible. SPP-DEA is developed based on monitoring, analysis, decision-making, and execution with knowledge bas (MADE-k) autonomous planning model with the aim of compromise between service cost, response time, resource utilization, and throughput. In order to address the computational complexity of the problem, we consider the resource consumption distribution and service deployment priority in the placement process. In order to evaluate the quality of placement in SPP-DEA, extensive experiments have been performed on a synthetic fog environment. The simulation results show that compared to the state-of-the-art approaches, SPP-DEA reduces the service cost and waiting time by 16% and 11%, respectively.