Real-time Resource Research Articles

Multi-Agent Reinforcement Learning-Based Resource Management for End-to-End Network Slicing

To meet the explosive growth of mobile traffic, the 5G network is designed to be flexible and support multi-access edge computing (MEC), thereby improving the end-to-end quality of service (QoS). In particular, 5G network slicing, which allows a physical infrastructure to split into multiple logical networks, keeps the balance of network resource allocation among different service types with on-demand resource requests. However, achieving effective resource allocation across the end-to-end network is difficult due to the dynamic characteristics of slicing requests such as uncertain real-time resource demand and heterogeneous requirements. In this paper, we develop a reinforcement learning (RL)-based dynamic resource allocation framework for end-to-end network slicing with heterogeneous requirements in multi-layer MEC environments. We first design a hierarchical MEC architecture and formulate a resource allocation problem for the end-to-end network slicing as an optimization problem using the Markov decision process (MDP). Using proximal policy optimization (PPO), we develop independently-collaborative and jointly-collaborative dynamic resource allocation algorithms to maximize resource efficiency while satisfying the QoS of slices. Experimental results show that the proposed algorithms can recognize the characteristics of slice requests and coming resource demands and efficiently allocate resources with a high QoS satisfaction rate.

Optimal Multi-stage Configuration Allocation with Applications to Video Advertising

In modern applications of real-time resource allocation, for example allocating ad opportunities created by users to relevant advertisers during video streaming, a prevalent scenario is when an arriving user can be served under different configurations''. For example, in video advertising, a configuration can be an arrangement of video-ads of various lengths and formats for a subset of advertisers, showing up at different time-stamps of the video. Each selected configuration assigns a certain number of impressions to each advertiser. Depending on their long-term contracts, each advertiser is willing to pay for a certain number of impressions during the decision making horizon at a per-impression price depending on the assigned configuration. While this application is real-time, there is an inherent allowance for latency in showing ads to users, as several of these ads show up in the middle (or at the end) of the video. This provides an opportunity for ad allocator to batch the users and to make more profitable allocation decisions. While designing algorithms for the fully online variant of this setting has been studied in the literature [Devanur et al., 2016], the multi-stage variant where the allocation decisions are made batch-by-batch is not studied. Motivated by the above application, we study designing optimal competitive multi-stage configuration allocation algorithms (with preemption) in an adversarial setting, where users arrive in batches in a stage-wise fashion. The allocator then decides on an irrevocable configuration for each of the users in an arriving batch at each stage so as to maximize the total revenue. We propose a novel convex-programming based multi-stage configuration allocation algorithm that is 1-(1-1/K)^K competitive, where K is the number of stages. We further show an intimate connection between the convex programs used in each stage and a recursive functional equation (with a sequence of polynomials of decreasing degrees as its solution). Using this connection, we develop a recursive primal-dual analysis for our algorithm and extend the previous analysis framework of multi-stage fractional matching [Feng and Niazadeh, 2021] to this more general setting. Interestingly, we show there is an intimate connection between convex Lagrangian dual of each stage of our algorithm and how the primal-dual analysis needs to be set up, which might be of independent interest.

Deep Learning–Based Energy Beamforming With Transmit Power Control in Wireless Powered Communication Networks

In this paper, we propose deep learning–based energy beamforming in a multi-antennae wireless powered communication network (WPCN). We consider a WPCN where a hybrid access point (HAP) equipped with multiple antennae broadcasts an energy-bearing signal to wireless devices using energy beamforming. We investigate the joint optimization of the time allocation for wireless energy transfer (WET) and wireless information transfer (WIT) with the design for energy beams while minimizing the transmit power at the HAP for efficient use of its available resources. However, this is a non-convex problem, and it is numerically intractable to solve it. In the literature, the traditional approach to solving this problem is based on an iterative algorithm that incurs high computational and time complexity, which is not feasible for real-time applications. We study and analyze a deep neural network (DNN)-based scheme and propose a faster and more efficient approach for the fair approximation of a near-optimal solution to this problem. To train the proposed DNN, we acquire training data samples from a sequential parametric convex approximation (SPCA)-based iterative algorithm. Instead of acquiring data samples and training the DNN, which is highly complex, we use offline training for the DNN to provide a faster solution to the real-time resource allocation optimization problem. Through the simulation results, we show the proposed DNN scheme provides a fair approximation of the traditional SPCA method with low computational and time complexity.

An Energy-Aware Host Resource Management Framework for Two-Tier Virtualized Cloud Data Centers

The high energy consumption of cloud data centers is one of the key issues restricting the future development of the cloud computing industry. For two-tier virtualized data centers in which containers are deployed on VMs, this paper conducts an in-depth study of this problem from the aspect of resource management. First, the working model of two-tier virtualized data center is defined. By mapping tasks to containers and using VMs to isolate different jobs, this model improves the resource utilization of hosts while ensuring the isolation and security of jobs. Then, an energy-aware host resource management framework is constructed, which includes two algorithms. The initial static placement is a two-tier scheduling algorithm, including the load balancing alternate placement and the two-sided matching methods to complete the placement of containers to VMs and VMs to hosts respectively. The runtime dynamic consolidation algorithm takes the initial placement scheme as input, and utilizes the dynamic consolidation method to use the least active hosts to meet the real-time resource requirements of containers. Finally, simulation experiments compared with related algorithms are conducted using real workload traces. The results show that the proposed two algorithms have better performance in host resource utilization, number of active hosts, number of container migrations and SLA metric. And the entire framework achieves an energy-saving effect of 13.8% at least.

Real-Time Energy Harvesting Aided Scheduling in UAV-Assisted D2D Networks Relying on Deep Reinforcement Learning

Unmanned aerial vehicle (UAV)-assisted device-to-device (D2D) communications can be deployed flexibly thanks to UAVs' agility. By exploiting the direct D2D interaction supported by UAVs, both the user experience and network performance can be substantially enhanced at public events. However, the continuous moving of D2D users, limited energy and flying time of UAVs are impediments to their applications in real-time. To tackle this issue, we propose a novel model based on deep reinforcement learning in order to find the optimal solution for the energy-harvesting time scheduling in UAV-assisted D2D communications. To make the system model more realistic, we assume that the UAV flies around a central point, the D2D users move continuously with random walk model and the channel state information encountered during each time slot is randomly time-variant. Our numerical results demonstrate that the proposed schemes outperform the existing solutions. The associated energy efficiency game can be solved in less than one millisecond by an off-the-shelf processor using trained neural networks. Hence our deep reinforcement learning techniques are capable of solving real-time resource allocation problems in UAV-assisted wireless networks.

Online Sparse BLSTM Models for Resource Usage Prediction in Cloud Datacentres

Real time resource usage prediction is an important part of resource provisioning in a cloud data centre. As cloud workloads vary dynamically, effective resource provisioning requires prediction of future resource usage trends. The problem is highly complicated because of highly time varying nature of cloud resource workloads. Training the future resource usage prediction models once, using a fixed set of observations is not sufficient to capture the variability in cloud workloads. In this work, we propose to use gradient descent (GD) and Levenberg-Marquardt (LM) adaptation algorithms for dynamic adaptation of resource utilization prediction models. We also propose a novel sparse framework for fast online adaptation of resource usage prediction models. We propose to analyze different algorithms such as ℓ <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup> regularization, ℓ <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> regularization, optimal brain damage (OBD), optimal brain surgeon (OBS) for introducing sparsity. The proposed sparse framework for online adaptation of multivariate resource usage prediction models is validated for CPU usage prediction in the Google cluster trace and PlanetLab workload trace. A comparative analysis of different sparse frameworks shows that OBD-based LM adaptation algorithm performs better than other frameworks for online multivariate resource usage prediction in a cloud.

Reinforcement learning-based hybrid spectrum resource allocation scheme for the high load of URLLC services

Ultra-reliable and low-latency communication (URLLC) in mobile networks is still one of the core solutions that require thorough research in 5G and beyond. With the vigorous development of various emerging URLLC technologies, resource shortages will soon occur even in mmWave cells with rich spectrum resources. As a result of the large radio resource space of mmWave, traditional real-time resource scheduling decisions can cause serious delays. Consequently, we investigate a delay minimization problem with the spectrum and power constraints in the mmWave hybrid access network. To reduce the delay caused by high load and radio resource shortage, a hybrid spectrum and power resource allocation scheme based on reinforcement learning (RL) is proposed. We compress the state space and the action space by temporarily dumping and decomposing the action. The multipath deep neural network and policy gradient method are used, respectively, as the approximater and update method of the parameterized policy. The experimental results reveal that the RL-based hybrid spectrum and the power resource allocation scheme eventually converged after a limited number of iterative learnings. Compared with other schemes, the RL-based scheme can effectively guarantee the URLLC delay constraint when the load does not exceed 130%.

Analysis of Traffic Performance on Network Slicing Using Complex Network Theory

In order to provide various customized services, network slicing as a key technology for fifth-generation mobile networks enables wireless infrastructure to be sliced into multiple isolated logical networks. Thanks to the softwarization and virtualization, slices which often in competition for physical resources are realized in standardized servers. Despite the advances in the realization of network slicing, analyzing the impact of resource allocation on the traffic performance of slices still faces some challeges. In this paper, a multilayer complex network model is introduced to catch the results of resource allocation and slice deployment. This model facilitates real-time resource monitoring and topology information updating for infrastructure network providers. With this model, we investigate the impact of changing resource allocation parameters by modelling the traffic data which are transmitted within the slices. To avoid the traffic congestion caused by the resource competition, the traffic performance of slices is analyzed. The analysis results enable us to propose two cost-effective control strategies which are increasing the capacity of hub nodes and enlarging the scale of the infrastructure network by adding more servers. These strategies aim to suppress the congestion while guaranteeing the traffic performance of slices. Simulation results demonstrated the effectiveness of the proposed strategies.

SAFER model: Strategic allocation of fundamental epidemic resources.

To describe the Strategic Allocation of Fundamental Epidemic Resources (SAFER) model as a method to inform equitable community distribution of critical resources and testing infrastructure. The SAFER model incorporates a four-quadrant design to categorize a given community based on two scales: testing rate and positivity rate. Three models for stratifying testing rates and positivity rates were applied to census tracts in Milwaukee County, Wisconsin: using median values (MVs), cluster-based classification and goal-oriented values (GVs). Each of the three approaches had its strengths. MV stratification divided the categories most evenly across geography, aiding in assessing resource distribution in a fixed resource and testing capacity environment. The cluster-based stratification resulted in a less broad distribution but likely provides a truer distribution of communities. The GVs grouping displayed the least variation across communities, yet best highlighted our areas of need. The SAFER model allowed the distribution of census tracts into categories to aid in informing resource and testing allocation. The MV stratification was found to be of most utility in our community for near real time resource allocation based on even distribution of census tracts. The GVs approach was found to better demonstrate areas of need.

Profit-Based Algorithm of Joint Real-Time Task Scheduling and Resource Allocation in C-RANs

In cloud radio access networks (C-RANs), a local controller controls a set of lightweight remote radio heads (RRHs) for connecting users nearby, and creates virtual machines (VMs) for executing user tasks. In view of previous works, the strategies of minimizing system resource consumption or transmission delay are adopted in the issue of joint task scheduling and resource allocation. Different from them, this work puts forward the strategy of maximizing the profit of the joint issue. The problem is formulated as integer linear programming (ILP), and then an algorithm with bounded approximation ratio for solving the ILP is proposed. The simulation results show that the solution of the algorithm is very close to the optimal one of the ILP. In addition, another algorithm is proposed to control the tradeoff between performance and robustness of the solution under the assumption that the system resource can be expanded by a bounded factor.

Real-Time Resource Allocation for Wireless Powered Multiuser Mobile Edge Computing With Energy and Task Causality

This article considers a wireless powered multiuser mobile edge computing (MEC) system, in which a multi-antenna hybrid access point (AP) wirelessly charges multiple users, and each user relies on the harvested energy to execute computation tasks. We jointly optimize the energy beamforming and remote task execution at the AP, as well as the local computing and task offloading, aiming to minimize the total system energy consumption over a finite time horizon, subject to causality constraints for both energy harvesting and task arrival at the users. In particular, we consider a practical scenario with casual task state information (TSI) and channel state information (CSI), i.e., only the current and previous TSI and CSI are available, but the future TSI and CSI can only be predicted subject to certain errors. To solve this real-time resource allocation problem, we propose an offline-optimization inspired online design approach. First, we consider the offline optimization case by assuming that the TSI and CSI are perfectly known a-priori . In this case, the energy minimization problem corresponds to a convex problem, for which the semi-closed-form optimal solution is obtained via the Lagrange duality method. Next, inspired by the optimal offline solution, we propose a sliding-window based online resource allocation design in practical cases by integrating with the sequential optimization. Finally, numerical results show that the proposed joint wireless powered MEC designs significantly improve the system’s energy efficiency, as compared with the benchmark schemes that consider a sliding window of size one or without such joint optimization.

A Task-Aware Fine-Grained Storage Selection Mechanism for In-Memory Big Data Computing Frameworks

In-memory big data computing, widely used in hot areas such as deep learning and artificial intelligence, can meet the demands of ultra-low latency service and real-time data analysis. However, existing in-memory computing frameworks usually use memory in an aggressive way. Memory space is quickly exhausted and leads to great performance degradation or even task failure. On the other hand, the increasing volumes of raw data and intermediate data introduce huge memory demands, which further deteriorate the short of memory. To release the pressure on memory, those in-memory frameworks provide various storage schemes options for caching data, which determines where and how data is cached. But their storage scheme selection mechanisms are simple and insufficient, always manually set by users. Besides, those coarse-grained data storage mechanisms cannot satisfy memory access patterns of each computing unit which works on only part of the data. In this paper, we proposed a novel task-aware fine-grained storage scheme auto-selection mechanism. It automatically determines the storage scheme for caching each data block, which is the smallest unit during computing. The caching decision is made by considering the future tasks, real-time resource utilization, and storage costs, including block creation costs, I/O costs, and serialization costs under each storage scenario. The experiments show that our proposed mechanism, compared with the default storage setting, can offer great performance improvement, especially in memory-constrained circumstances it can be as much as 78%.

Web-Based Decision Support System for Coastal Water Quality

Sujith Kumar, S.; Panda, U.S.; Pradhan, U.K; Mishra, P., and Ramana Murthy, M.V., 2020. Web-based decision support system for coastal water quality. In: Sheela Nair, L.; Prakash, T.N.; Padmalal, D., and Kumar Seelam, J. (eds.), Oceanic and Coastal Processes of the Indian Seas. Journal of Coastal Research, Special Issue No. 89, pp. 139-144. Coconut Creek (Florida), ISSN 0749-0208.Operational systems on real-time monitoring of coastal water quality can play a key role to counter and reduce the pollution causing hazardous events in the coastal region. An automated alert system is required to provide relevant information to the coastal authorities, water quality managers and other coastal stakeholders on a real-time basis to avoid any lead time or potential human errors. This paper presents an architecture for coastal water quality prediction and alert system by integrating different platforms and services to achieve scalability, provisioning of resources in real time, simplified deployment and management of resources and applications. The forecast system was developed through data assimilation from coupled hydrodynamic water quality models and field observations and archived in an integrated Database Management System (DBMS). A web-service is programmed to fetch the data from server on demand in order to disseminate the information to the public. Web applications use an architecture that breaks applications into discrete parts, using multiple programming languages and deploying the applications on several layers of technology. The other features of the website include Query builder as per user's specification, Fetch and display latest time series data on water quality parameters based on locations and graphical representation.

Trajectory Data Driven V2V/V2I Mode Switching and Bandwidth Allocation for Vehicle Networks

Maintaining efficient and reliable communication services in vehicle networks is challenging due to the high mobility of vehicle users. One of the key issues is to deal with the large amount of channel feedback and make better use of limited wireless resources in real time. In this letter, we propose a trajectory data driven method to maximize the total user utilities, which is characterized by intelligent channel state information (CSI) feedback reduction and vehicle-to-vehicle (V2V)/vehicle-to-infrastructure (V2I) mode switching. A deep neural network (DNN) model is trained using optimization approaches and vehicles’ trajectory data. Based on the well trained model, V2V/V2I mode is selected adaptively and bandwidth is allocated optimally with partial channel feedback. Simulation results show that our proposed method outperforms the comparison baseline and can achieve 96.23% of the performance obtained by exhausting searching with full CSI feedback.

An adaptive dwell time scheduling model for phased array radar based on three-way decision

Real-time resource allocation is crucial for phased array radars to undertake multi-task with limited resources such as in the situation of multi-target tracking, in which targets need to be prioritized so that resources can be allocated accordingly and effectively. In this paper, a three-way decision-based model is proposed for adaptive scheduling of phased radar dwell time. Using the model, the threat posed by a target is measured by an evaluation function, and therefore, a target is assigned to one of the three possible decision regions, i.e., positive region, negative region, and boundary region. A different region has a various priority in terms of resource demand, and as such, a different radar resource allocation decision is applied to each region to satisfy different tracking accuracy of multi-target. In addition, the dwell time scheduling model can be further optimized by implementing a strategy for determining a proper threshold of three-way decision making to optimize the thresholds adaptively in real-time. The advantages and the performance of the proposed model has been verified by experimental simulations with comparison to the traditional two-way decision model and the three-way decision model without threshold optimization. The experiential results have demonstrated that the performance of the proposed model has a certain advantage in detecting high threat targets. 2020 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

Efficiently Consolidating Virtual Data Centers for Time-Varying Resource Demands

Data center virtualization is a flexible and efficient way to enable multiple users to share the common resources of a physical data center (DC). For efficient sharing, virtual data center (VDC) embedding is a vital problem that should be carefully addressed. However, existing studies on VDC embedding mostly assume that the capacity of each VDC is fixed, but do not consider the time-varying feature of resource demands. Considering the fact that the resource demands of most enterprise IT services exhibit the time-varying feature, resource allocation based on the fixed capacity assumption would cause a great inefficiency. To overcome this inefficiency, we propose a new VDC consolidation scheme that takes into account the time-varying feature of resource demands when embedding VDCs. We first develop a resource demand prediction model for each VDC using the Long Short-Term Memory (LSTM) neural network, which is used to predict the real-time resource demands of VDCs at different future moments. Based on the predicted resource demands, we then embed VDCs whose peaks and valleys of resource demands stagger each other onto common physical servers and links, such that the required physical resources can be minimized under the condition that all the resource demands of different VDCs are satisfied at all the different moments. An integer linear programming (ILP) model and a resource demand correlation-based heuristic algorithm are also developed for the proposed scheme. Simulation results show that the proposed consolidation scheme can significantly improve resource utilization in a DC. It can save up to 25 percent of physical servers and 29 percent of physical links used for accommodating the same requests as compared to a scheme assigning resources based on the fixed capacity assumption.

Security-Aware Dynamic Scheduling for Real-Time Optimization in Cloud-Based Industrial Applications

Nowadays, large number of cloud-based techniques have been used in industrial control systems (ICS), which also brings many security threats. The emergence of security-aware industrial control has paved the way of security-aware scheduling in cloud-based industrial applications. Actually, most cloud-based industrial applications are time sensitive, which need real-time processing. Edge cloud computing paradigm extends the computing ability of traditional cloud model with low-latency local resources. Thus, heterogeneous clouds that consist of both centralized resources and edge resources may be a promising resource model to provide both scalable and low-latency resources for cloud-based industrial applications. In view of these challenges, in this article, we propose a security-aware dynamic scheduling method for real-time resource allocation in ICS. First, a three-level security model is designed for both tasks and cloud resources in ICS, and a two-tier heterogeneous cloud architecture is introduced. Accordingly, a security-aware scheduling method based on distributed particle swarm optimization is presented for resource allocation with security concerns. To deal with the dynamics of edge resources and the mobility of mobile industrial applications, a dynamic scheduling mechanism based on dynamic workflow model is proposed for real-time optimization. Experimental results validate that the scheduling control policy proposed in this article can achieve a good balance between scheduling performance and security performance.

Fine-Grained Cloud Resource Provisioning for Virtual Network Function

The deployment of Virtualized Network Functions is expected to be dynamic and swift when using Network Function Virtualization technology. The dynamic nature of workload from users requires the resource allocation of underlying infrastructure to be flexible to cope with the changes. Existing works investigated elastic NFV solutions by dynamically creating and dismantling Virtual Machine (VM) replicas, while maintaining balanced workload among VMs. However, those solutions are coarse-grained which may cause unnecessary resource over-provisioning as different network functions consume different amount of resources. In this paper, we present ElasticNFV, a dynamic and fine-grained cloud resource provisioning solution for VNF. ElasticNFV takes real-time resource demand of multiple service chains and allocates resources through an elastic provision mechanism. When a scaling conflict occurs, ElasticNFV provides a two-phase minimal migration algorithm to optimize the migration time and embedding cost of VNF instances. We implement ElasticNFV on top of the KVM platform to provide elastic VM for each VNF instance and Open vSwitch to form elastic intra-cloud network with virtual links between VNF instances. Our evaluation results show that ElasticNFV can improve VNF performance significantly, and achieve high resource utilization and fast migration time with low cost.

Optimisation of phonetic aware speech recognition through multi-objective evolutionary algorithms

Recent advances in the availability of computational resources allow for more sophisticated approaches to speech recognition than ever before. This study considers Artificial Neural Network and Hidden Markov Model methods of classification for Human Speech Recognition through Diphthong Vowel sounds in the English Phonetic Alphabet rather than the classical approach of the classification of whole words and phrases, with a specific focus on both single and multi-objective evolutionary optimisation of bioinspired classification methods. A set of audio clips are recorded by subjects from the United Kingdom and Mexico and the recordings are transformed into a static dataset of statistics by way of their Mel-Frequency Cepstral Coefficients (MFCC) at sliding window length of 200ms as well as a reshaped MFCC timeseries format for forecast-based models. An deep neural network with evolutionary optimised topology achieves 90.77% phoneme classification accuracy in comparison to the best HMM that achieves 86.23% accuracy with 150 hidden units, when only accuracy is considered in a single-objective optimisation approach. The obtained solutions are far more complex than the HMM taking around 248 seconds to train on powerful hardware versus 160 for the HMM. A multi-objective approach is explored due to this. In the multi-objective approaches of scalarisation presented, within which real-time resource usage is also considered towards solution fitness, far more optimal solutions are produced which train far quicker than the forecast approach (69 seconds) with classification ability retained (86.73%). Weightings towards either maximising accuracy or reducing resource usage from 0.1 to 0.9 are suggested depending on the resources available, since many future IoT devices and autonomous robots may have limited access to cloud resources at a premium in comparison to the GPU used in this experiment.

Design of Airport Wireless Bird Repellent Monitoring System

At present, the status detection of the bird driving equipment in the airport is completed by manual periodic inspection. This method can’t grasp the status of the bird repelling equipment in real time and cost human resources. In view of this situation, a remote monitoring system based on STM32 is proposed. The system uses the embedded STM32 as the control and data processing core to collect the signals of the temperature sensor, pressure sensor, power supply voltage and sound sensor of the bird-repelling equipment. It is uploaded to the host computer through ZigBee technology to analyze the working conditions of the bird-repelling equipment and the situation of the gas gun. Through the test of the design system, the results show that the detection accuracy of the system can meet the actual need, the correct rate of detecting sound signal is 100%. In the environment of two concrete buildings, the wireless transmission distance is within 800m, and the error rate is 0.