Vijayaraj et al 2024a - Revision history

Rimni at 08:00, 31 May 2024

2024-05-31T08:00:11Z

Rimni at 07:59, 31 May 2024

2024-05-31T07:59:28Z

Rimni at 07:58, 31 May 2024

2024-05-31T07:58:38Z

Rimni at 07:58, 31 May 2024

2024-05-31T07:58:20Z

Rimni at 07:57, 31 May 2024

2024-05-31T07:57:59Z

Rimni at 07:57, 31 May 2024

2024-05-31T07:57:30Z

Rimni at 07:57, 31 May 2024

2024-05-31T07:57:06Z

Vijayveeramani: The paragraph is modified that is started before Equation 10.

2024-05-31T06:53:33Z

The paragraph is modified that is started before Equation 10.

Rimni at 11:49, 30 May 2024

2024-05-30T11:49:32Z

Show changes

Rimni at 11:10, 30 May 2024

2024-05-30T11:10:27Z

@@ Line 438: / Line 438: @@
 Using ZOA to provide well-informed options for job situations and reconfiguration, this article explored intelligent planning task delivery. The problem is solved by providing a system design that allows for careful planning and reconfiguration in a big data setting. We create a quantitative approach to lower the total cost of timeliness for all jobs. Furthermore, a model for optimization is proposed that incorporates a timetable and agents that may be reconfigured. The agency's characteristics, operations, and incentives are aimed at the two services. Using cloud-edge computing in a big data setting, it claims to provide a high-performance computing strategy that makes good use of available resources and evenly distributes processing loads. The suggested protocol also utilized the ZOA method to discover effective keys with the best possible local besides global fitness functions. In order to maintain a steady connection with little latency, the fitness function makes use of the energy and reliability aspects of the channel. On top of that, network edges protect data storage and gathering from network risks while small hardware devices are dispersed and limited incur little costs. Based on the results of the experiments, it is evident that the ZOA reduces energy usage by up to 18.74J compared to other models like MOA, SLOA, and BOA. However, the proposed model achieved low performance in minimizing the makespan due to high interia weight of the proposed zebra optimization algorithm. This must be addressed in future work and also tries to minimize the high computational time.
-===7.1. Future Work ===
+===7.1 Future Work ===
 The future of this project depends on investigating and resolving issues related to its computational complexity and cost efficiency. Implementing the provided methodologies for Big Data requests multi-cloud surroundings is also within the realm of possibility.

@@ Line 257: / Line 257: @@
 The planned pipeline has been dissected and its process is shown in [[#img-1|Figure 1]]. Big data environment mobile user devices, base stations, cloud nodes, and fog nodes make up this system. The scheduler's steps for the suggested strategy are as follows: (1) A appeal is sent to the fog node with the mobile data. This (2) creates the fog node and sends the job to the main station. (3) The base station estimates, schedules, and assigns tasks to the nodes in the cloud. (4) At a certain interval, the cloud node will submit the completed job. Users start the process, and nodes in the cloud and fog keep in touch with each other. The job that was passed from the fog node is collected by the station. After receiving the job, the base station breaks it down into its component parts, makes an estimate, and then provides the aggregate result. The findings are combined by the base station once the work is decomposed. The task is then scheduled and sent. At a certain moment, the aggregated outcome is applied in the base station. Users get the final, combined result when it is computed. The request is sent to the fog node using the suggested architecture, which uses the smartphone network. After leaving the fog node, the job is directed to station. Through the base station, the job is sent and received. The base station sends the deconstructed job to the cloud node. The work is done in the cloud node, and then the output is performed. The reconfiguration agent comes after the architectural step.
-===4.1. Reconfiguration agent===
+===4.1 Reconfiguration agent===
 When the judgment unit uses the most recent state characteristics to create a reconfiguring activity. At each <math display="inline"> T </math> step, the RA reaps the advantages of both the prior reconfiguring activity and the transactions that came before it. Rearranging the reconfiguring action is done using the reconfiguration agent. Using the reconfiguration agent, the reconfiguring procedure may be rearranged. Here we will go over the steps to create the reward, training, and state components.
-====4.1.1. Reward====
+====4.1.1 Reward====
 Keep in mind that reducing the total delay cost for all network jobs is the objective of the problem being researched. Reducing the total cost of delay should be the goal of any reconfiguration effort. What this means is that every reconfiguration procedure has to keep the total cost of configuration step to a minimum. Maximizing the cumulative reward for an episode is the goal of a deep learning agent. So, the compensation for each reconfiguration phase is defined as the inverse of the tardy was just applied per second at that stage.
@@ Line 280: / Line 280: @@
 The first and ultimate stages of reconfiguration are denoted as <math display="inline">r'</math> and <math display="inline">r''</math>, respectively. The reconfiguration judgment awards and distributes the cumulative prizes from each episode. What follows is an explanation of the reconfiguration judgment.
-====4.1.2. Reconfiguration judgment====
+====4.1.2 Reconfiguration judgment====
 To decide whether to reorganize, the reconfiguration judgment is activated when the initial device finishes a job. Only when the RA wants to decide to reconfigure does it do so. With the help of the reconfiguration judgment system, human reconfiguration is reduced in frequency. In any case, the RA will need a lot of practice events before it figures out how to make reconfigurations less frequent.

@@ Line 112: / Line 112: @@
 In this section, we will examine the problem formulation and the reasons behind creating the simulation in further detail.
-===3.1. Resource optimisation problem statement===
+===3.1 Resource optimisation problem statement===
 With the (IoT), a wide diversity of tasks is possible, from monitoring sleep to keeping tabs on daily activities. Wearable and handheld devices are getting smarter and can link to social media accounts and monitor data, which improves people's lives. However, there are a lot of data transport and storage issues that come along with all this activity. A big data environment is being planned in the hopes of an interconnected world with a hundred-fold increase in user data-rate and connected devices, a tenfold increase in battery life for massive machine communiqué [32]. Problems with scalability, latency, and compatibility grow dramatically with the addition of even a single node leading to underappreciated services. Within the context of big data, there are two primary types of resource optimization problems: allocation and scheduling. There have been several attempts to optimize the Internet of Things (IoT), but most of these studies are either domain-specific or place too much emphasis on resource allocation at the expense of resource scheduling. Examples of domain-specific literature include studies devoted to healthcare job scheduling and management [33–34], transportation task offloading and scheduling [35], and industrial automations [36]. It is safe to state that there has not been complete adoption and implementation of a universal IoT framework that can be used in all IoT scenarios. With the correct strategy, optimizing resources can make docile to user expectations, which is a big step toward easing the complexity of popular IoT systems. In response to these issues, this study presents a new resource optimization technique that may be used in different Internet of Things settings.
-===3.2. Motivation and lapses===
+===3.2 Motivation and lapses===
 The Internet of Things (IoT) and value-added services generate and consume vast amounts of data; as a result, multidimensional optimization problems, such as how to choose the best service configuration in real-time and how to provide an efficient scheduling scheme for edge services, demand substantial research and development efforts [37]. Other issues with the dynamic resource allocation (DRA) method include the high IoT-based cloud systems and the resource under-utilization problem. Because of their limited storage and power capabilities, the edge nodes exacerbate the security flaws introduced by the decentralized nature of the Internet of Things' topology. An effective resource optimization strategy that gives equal weight to scheduling and resource allocation is required to avoid these mistakes. There are a number of critical areas that need a breakthrough, including the algorithms' ability to satisfy users' demanding requirements (QoE besides QoS) and the longevity of edge nodes, particularly sensor nodes.

@@ Line 98: / Line 98: @@
 The drawbacks and research gap of the existing techniques are mentioned in the upcoming section.
-===2.1. Ideology from the related works===
+===2.1 Ideology from the related works===
 What follows is an analysis of the problems found in the recently published research on bio-inspired and current metaheuristic load balancing systems:

@@ Line 22: / Line 22: @@
 ==1. Introduction==
-===1.1. Background of big data applications===
+===1.1 Background of big data applications===
 Tools, technologies, and architectures with increased efficiency, flexibility, and resilience have emerged as a result of the Big Data age. Complex architectures with built-in scalability and optimisation capabilities are necessary for big data applications. The environments in which big data applications are deployed must be updated and upgraded on a regular maximise their scalability and flexibility [1]. Cloud-based services are being used by organisations to improve performance and reduce overall costs. Because it is lightweight, containerisation, a cloud-based technology, is becoming more and more popular. One of the most popular and widely used container-based virtualizations is Docker, which is an open-source project that makes it easy to create, operate, and deploy applications [2].
@@ Line 38: / Line 38: @@
 Currently, real-time traffic data analysis, weather data analysis, and medical data storage are three areas where cloud computing technology is making significant progress. Cloud computing is inexpensive, and the machines in the cluster setup don't need to meet any complex specifications. Integration with conventional data mining techniques can enable more effective administration and analysis of power monitoring data thanks to cloud computing's enormous scale and quick computation speed. In conclusion, because of its capacity to delve deeper into the shifting law of the load curve and successfully identify years [15].
-===1.2. Issues on energy consumption===
+===1.2 Issues on energy consumption===
 The majority of monitoring systems were created for specific types of equipment in the early stages of condition monitoring technology development, and each scheme was dispersed and isolated. This was an info island where there was no data exchange or interaction, making it difficult to manage and thoroughly analyse monitoring data. Furthermore, it is challenging to share the hardware network, computer power, and storage—of various monitoring systems, which wastes IT resources. As a result, an integrated management system that was constructed in the main control room has surfaced and is capable of processing different monitoring data that are gathered by various monitoring devices centrally [16]. The monitoring device's present limitations, however, are that it can only transfer the streamlined, monitoring centre, and the frequency of data gathering is low. As a result, the monitoring centre will eventually gather an incredible quantity of data, and the information processing capacity of the current monitoring system will not be able to handle the demands of processing and storing such a large amount of data. It is clear that the serial processing approach has long been inadequate to handle the demands of processing massive volumes of data. Various computing challenges faced in scientific research and engineering practice have historically been attributed to the typical parallel computing paradigm based on high-performance processors.
@@ Line 44: / Line 44: @@
 As a result, the monitoring device processes expert data locally and feeds it to the present monitoring system. Before being uploaded, the monitoring device, for instance, has to analyse the partial discharge waveform data from high-voltage electrical equipment into the sum of discharges, matching discharge phase [17]. Uploading "familiar data" rather than "raw data" can save money on storage at monitoring centres and network transmission expenses. Even yet, a monitoring centre that combines data from several monitoring device specifications still has a difficult time diagnosing target device failure and doing a thorough condition assessment.
-===1.3. Issues on resource allocation===
+===1.3 Issues on resource allocation===
 Various formulations of the resource allocation problem (RAP) have been suggested in line with various issue scenarios; the RAP may describe all real-world circumstances. Internet of Things (IoT) resource allocation problems (RAPs) are large-scale and multi-faceted, and deterministic algorithms are unable to solve them because they are nondeterministic polynomial (NP)-complete. While genetic algorithms (GA) and other NP algorithms have been researched for their ability to identify near-optimal solutions, they have a propensity to generate a significant number of infeasible keys while searching [18]. Due to GA's shortcomings, a particle swarm optimisation (PSO) metaheuristic clustering method was suggested for nonlinear MORAP. This method aims to find the Pareto-optimal keys, which are solutions that are not overshadowed by other solutions; in other words, solutions that improve one preference criterion without compromising another. To tackle scheduling challenges, a Pareto optimum solution based on time is employed to convert decisions. The suggested Pareto optimisation methods have proven to be effective in producing optimal solutions. There is an urgent need to optimise energy usage in order to extend the lifespan of the network, as the Internet of Things (IoT) is primarily used for environmental monitoring, data collecting, and processing. Due to their limited battery life, sensor and actuator nodes in the Internet of Things (IoT) may be prematurely destroyed if their resources are not used efficiently [19]. As a result, EC paradigms that revolve around edge nodes have grown in popularity as a way to address IoT's MORAP. This change in strategy has prompted a lot of research on RA as a way to handle traffic pressure and the difficulties of IoT and cloud computing. Scheduling service resources, guaranteeing quality of service (QoS), and merging multiple services are some of the well-established issues that come with the edge computing paradigm, which is related to cloud computing [20].
@@ Line 52: / Line 52: @@
 Implementing communication protocols to plan data transfer is an excellent method to reduce or eliminate data collision altogether [25]. When it comes to multitasking, the simplest and oldest scheduling protocol or algorithm is the round robin (RR) static algorithm. It provides for the equitable distribution of time slots across resources and servers. The cyclic queue, which is constrained by a time slice, also called quantum time, performs each job in turn. A real-time pre-emptive method, round robin controls a node's access at each transmission instance according to a specified circular sequence and reacts to real-time occurrences. On the other side, resource-based (RB) algorithms prioritise the allocation of resources from highest to lowest using a greedy approach and a heuristic technique. So, to maximise throughput while minimising power consumption, it repeatedly chooses the most demanding request or job and assigns it to a suitable and available server for processing [26]. For diverse requests and jobs, the RB dynamic algorithm works well by looking at resource performance records over time to determine which one is the best fit, which improves performance overall.
-===1.4. Contribution of the research work===
+===1.4 Contribution of the research work===
 The research presented here pertains to load-balancing large data applications running in containerised systems such as Docker. Based on the Docker Swarm and ZOA architecture, this paper proposes a container scheduling technique for large data applications. This article explains how to manage the workload and service discovery of large data applications using the Docker Swarm concept. In order to decrease energy costs and end-to-end latency, various activities are evaluated and processed using an efficient scheduling model. To improve the optimisation algorithm's data utilisation and prevent it from being mired in local optimisation, a regional exploratory search method is employed..
-===1.5. Organization of the Work===
+===1.5 Organization of the Work===
 Section 1: Includes the background of big data, issues of energy consumption, introduction of resource allocation problem and contribution of the research work.

@@ Line 20: / Line 20: @@
 '''Keywords''': Big data application, Internet of Things, cloud storage, zebra optimization algorithm, task scheduling, resource utilization
-==1.Introduction==
+==1. Introduction==
 ===1.1. Background of big data applications===

@@ Line 284: / Line 284: @@
 To decide whether to reorganize, the reconfiguration judgment is activated when the initial device finishes a job. Only when the RA wants to decide to reconfigure does it do so. With the help of the reconfiguration judgment system, human reconfiguration is reduced in frequency. In any case, the RA will need a lot of practice events before it figures out how to make reconfigurations less frequent.
-Let <math display="inline">t'= 0,\, 1,\, 2,\, \cdot \, \cdot \, \cdot \, ,\, N</math> characterize the existing manufacture mode and <math display="inline">{w}_{t'}</math> signify its buffer. The present scheme time is designated by <math display="inline">t'</math>, <math display="inline">{r}^{c}</math>. The x’s current area tardy cost is represented by <math display="inline">{\beta }_{x}</math>, where the moment in time is defined by current cost. The Eq. (10) represents the current cost.
+Let <math display="inline">t'= 0,\, 1,\, 2,\, \cdot \, \cdot \, \cdot \, ,\, N</math> characterize the existing manufacture mode and <math display="inline">{w}_{t'}</math> signify its buffer. The present scheme time is designated by <math display="inline">t'</math>, <math display="inline">{r}^{c}</math>. The <math display="inline">x</math>’s current area tardy cost is represented by <math display="inline">{\beta }_{x}</math>, where the moment in time is defined by current cost. The Eq. (10) represents the current cost
 {| class="formulaSCP" style="width: 100%; text-align: left;"

@@ Line 380: / Line 380: @@
 The details of the projected cloud environment are detailed in the simulated environment, and the performance matrices cover each key matrix. Present scheduling and load-balancing methods are contrasted with the suggested method in the comparative analysis section. The jobs utilized in this project range from one hundred to five hundred. [[#img-2|Figures 2]] to [[#img-6|6]] demonstrate comparisons between the suggested model and many current methodologies, including the Butterfly optimization Algorithm (BOA), the Stray Lion optimization Algorithm (SLOA), and the mother optimization Algorithm (MOA).
 <div id='img-2'></div>
@@ Line 389: / Line 391: @@
 |}
 <div id='img-3'></div>