EDGE LEVEL COMPLEX EVENT PROCESSING AND OPTIMISATION BASED FOG LEVEL LOAD BALANCING MECHANISM IN SOFTWARE DEFINED NETWORK

Journal of Theoretical and Applied Information Technology 31st May 2025. Vol.103. No.10 © Little Lion Scientific ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195 4310 EDGE LEVEL COMPLEX EVENT PROCESSING AND OPTIMISATION BASED FOG LEVEL LOAD BALANCING MECHANISM IN SOFTWARE DEFINED NETWORK SARKARSINHA HARSINHA RAJPUT*1, DR. MANOJ EKNATH PATIL2 *1Ph.D. Research Scholar, Department of Computer Engineering, SSBT’s College of Engineering & Technology, Jalgaon, Maharashtra, 425001, India. 2Associate Professor, Department of Computer Engineering, SSBT’s College of Engineering & Technology, Jalgaon, Maharashtra, 425001, India. Email: *[email protected], [email protected] ABSTRACT This study uses an efficient transmission model based on the Hybrid Meta-heuristic Model to enhance data transfer by reducing time complexity. Initially, data is moved into the Complex Event Processing (CEP), which is positioned between the fog layer and the IoT layer. In edge IoT devices, complex event processing comprises real-time analysis, correlation, and interpretation of continuous data streams generated by sensors and edge devices. It seeks to identify significant trends or intricate occurrences in various data streams in order to facilitate quick decisions or immediate reactions. After CEP, a multi-tier priority queue-based model is used to attain priority-aware task scheduling. After the arrival of all the tasks, each task is sorted into slots based on its category. High-priority tasks are completed first due to their preference over lower-priority slots. A software-defined network's optimal resource utilization and task response time are guaranteed by an effective load-balancing method called Hybrid Pigeon Cat Search Optimization Algorithms (HPC_SOA). Arranging tasks based on their availability, capacity, proximity, and energy efficiency may optimize the fog nodes' resource utilization and energy usage. In the evaluation, the proposed approach has consumed 22051 Kw/h of energy. Keywords: Complex Event Processing, Priority Queue Approach, Pigeon Optimization, Cat Search Optimization, Software-Defined Network. 1. INTRODUCTION Large amounts of data are being produced as industrial processes become more digital, and the complexity of the data is also growing. The employees must be backed by technology in order to extract meaningful information from the massive amounts of created data, as the complexity of the data exceeds human comprehension [1]. Automated techniques for analyzing different types of data processing were needed for human interpretation. Due to the necessity of handling a growing volume of data and promptly responding to data triggers, real-time analytics is becoming progressively vital for enterprises and social applications [2]. The realtime schemes should have the following crucial features for application in industrial fields: low latency, high availability and Horizontal scalability [3]. Intelligent data handling and analysis are required to achieve near-real-time manufacturing and logistics process monitoring and control [4]. This led to the development of Complex Event Processing (CEP), which enables near-realtime processing of massive data streams [5, 6]. CEP refers to the tactics, instruments, and processes used to handle events immediately. CEP's primary purpose is to detect complex event patterns in data streams from sensors and other sources [7]. The basic goal of CEP systems is to create rule forms for similar occurrences based on semantic and spatial correlations. CEP engines rely exclusively on these rules, the vast majority of which are defined by subject matter experts [8, 9]. The complexity of defining and deriving rules is influenced by the industrial procedure and scenario. As a result, one of the limitations that complicates CEP integration and implementation is expert-based rule formulation [10]. Journal of Theoretical and Applied Information Technology 31st May 2025. Vol.103. No.10 © Little Lion Scientific ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195 4311 The network structure has become increasingly important in recent years due to the rapid growth of traffic rules and network quality, which has caused the network supplies to move quickly [11]. Due to their rigidity, traditional network topologies continue to face challenges in adapting to the dynamic nature of contemporary networks and keeping pace with evolving requirements [12, 13]. Developers created SoftwareDefined Networking (SDN) to meet the requirement for adaptable networks [14]. SDN enables a more adaptive, scalable, and cost-effective network architecture by separating the control and data planes [15]. The control planes' centralized SDN controller is in charge of packet routing [16]. The data plane is a layer of infrastructure that includes networked forwarding devices such as SDN switches [17]. For SDN-related technologies to be applied effectively, the networking elements must include software in physical architecture [18]. It is the perfect environment for load-balancing implementation because the controller gives information about the network resources that may be employed for load optimization [19]. To avoid network overload, load balancing (LB) is a technique that uses many resources to handle a single operation. In general, LB aims to optimize network traffic without reducing response time and throughput. Load balancing techniques are reputably imprecise in modern networks, but in SDN, they are distinguished by their precision and excellent efficiency [20]. 1.1 Motivation The rapid expansion of IoT applications and the increasing demand for real-time data processing have underscored the importance of efficient load balancing and CEP in edge and fog computing environments. Recent literature from 2024 highlights several challenges and advancements in this domain, justifying the need for continued research and innovation. Fog computing serves as an intermediary layer between the cloud and IoT devices, aiming to reduce latency and improve response times. However, the limited resources of fog nodes pose significant challenges in meeting the demands of resource-intensive applications. Efficient load balancing is crucial to prevent overloads and ensure optimal resource utilization. The dynamic nature of IoT applications, characterized by fluctuating workloads and diverse device capabilities, necessitates adaptive scheduling mechanisms. Traditional static load balancing approaches are inadequate in such environments, leading to suboptimal performance and increased latency. The incorporation of AI techniques, such as deep reinforcement learning and hybrid optimization algorithms, has shown promise in enhancing load balancing strategies. These approaches enable systems to learn and adapt to changing conditions, improving overall efficiency and response times. As fog computing environments handle sensitive data, ensuring secure data transmission and processing is paramount. Additionally, energy efficiency remains a critical concern, especially for battery-powered edge devices. Recent studies have proposed models that address both security and energy consumption, highlighting the need for holistic solutions. Given these challenges, there is a pressing need for advanced load balancing and CEP mechanisms that can operate efficiently in dynamic, resourceconstrained, and heterogeneous fog computing environments. The development of hybrid optimization algorithms, such as the proposed Hybrid Pigeon Cat Search Optimization Algorithm (HPC_SOA), aims to address these issues by combining the strengths of multiple AI techniques. Such approaches can lead to improved resource allocation, reduced latency, enhanced security, and better energy efficiency, ultimately supporting the growing demands of IoT applications. By building upon the insights from recent literature, this study seeks to contribute to the advancement of load balancing and CEP strategies in fog computing, ensuring that future systems are more robust, adaptive, and capable of meeting the evolving needs of real-time data processing. The main contributions of the paper are given as follows. To introduce an edge-level complex event processing and optimization-based fog level load balancing mechanism in SDN To present a Complex Event Processing (CEP) that entails the real-time analysis, correlation and interpretation of continuous streams of data produced by sensors and edge devices. To deploy a Multi-tier priority queue-based model for attaining priority-aware task scheduling. To implement the Hybrid Pigeon Cat Search Optimization Algorithms (HPC_SOA) method to build an efficient load balancing mechanism. In real-time IoT systems, not all tasks are equally critical. CEP enables early identification and categorization of events into simple or complex tasks based on contextual parameters like soil moisture or Journal of Theoretical and Applied Information Technology 31st May 2025. Vol.103. No.10 © Little Lion Scientific ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195 4312 temperature. To handle dynamic data streams and maintain processing speed, the CEP module was designed with rule-based logic focusing on frequent event patterns relevant to industrial use cases (e.g., agriculture, monitoring). Once tasks are classified, scheduling must ensure that time-sensitive tasks are prioritized without starving less critical ones. A multi-tier non-preemptive queue ensures fairness and deadline sensitivity. Utilization factor and task characteristics (e.g., arrival rate, frequency) were used to assign tasks to queues. Task aging and queue rotation were used to prevent starvation and ensure throughput. Traditional optimization methods struggle with either convergence speed or solution quality. HPC_SOA combines the exploration ability of Pigeon Optimization with the local refinement of Cat Search to balance tasks across fog nodes efficiently. Management of Factors: Fitness functions were carefully formulated to incorporate energy usage, response time, proximity, and resource availability. The hybrid algorithm dynamically adjusted parameters (like search range and memory pool) to adapt to workload fluctuations. Most prior works only address specific components (e.g., controller load balancing, or event filtering). To validate practicality, experiments were conducted on a realworld dataset rather than synthetic data, offering realistic variance in workloads and resource constraints. Parameters such as CPU usage, migration time, energy, and response time were closely tracked across different optimization techniques (WOA, SMO, GACO) to ensure a fair comparative analysis. The paper is organized from Section 1 as the introduction of SDN, and the literature review is given in Section 2. In section 3, the proposed methodology is described, and its evaluation results are given in section 4. The conclusion and future scope are provided in the final section 5. 2. RELATED WORK This section provides a detailed overview of techniques and drawbacks in the existing works on designing an edge-level complex event processing. For Monitoring the Real-Time IoT, Lan et al. [21] suggested a Universal Edge-Based Complex Event Processing Mechanism. To simplify event modelling, give a defined hierarchical complex event model made up of raw, simple, as well as complex events. The paradigm supports complex time as well as space semantics, allowing programmers to construct flexible, complicated events. A CEP system design was established on the network edge, positioned between cloud applications and terminal sensing devices. The CEP rule logic scripts can be linked to the complicated event description in order to quickly identify any potential anomalous events. The CEP was tiny enough for individual usage and operates on a mobile device. For Intelligent Agriculture, da Costa Bezerra et al. [22] introduced a Handling Complicated Events in Internet of Things Systems Based on Fog. The new sensor nodes in a network are connected to a Fog node based on the type of data as well as Euclidean distance among them, and utilizing geolocation and context-aware algorithms. This can run simulations in several situations with various network designs and densities to assess the concept. There were some issues on a potential security basis. Xu et al. [23] suggested Achieving Controller Load Balance in Distributed SoftwareDefined Networks via Dynamic Switch Migration. To achieve load balancing between SDN controllers with minimal migration costs, the balanced controller (BalCon) as well as BalConPlus SDN switch migration strategies should be suggested. BalCon works well in situations when the network does not need switch requests to be processed in a serial fashion. BalConPlus migrates a small number of switches with minimum calculation overhead, significantly lowering the load imbalance between SDN controllers. A software-defined network controller adaptive load balancing technique was presented by Priyadarsini et al. [24]. In this study, several switches are migrated from source controllers to target controllers using the self-adaptive load balancing (SALB) technique, which dynamically balances load among different controllers. The ability to disperse load under high load conditions using the estimated distance among target controllers and switches. This previous study has significant limitations due to its complexity and cost. For Several Controllers in SoftwareDefined Networking, Li et al. [25] introduced a Fuzzy Satisfaction-Based Load Balancing Method. First, the balancing judgment matrix as well as switch selection degree were introduced to select the transfer of domain as well as migrating switches for controller load monitoring. Second, the migration cost and load balancing rate were considered to be the key load balancing factors. Third, the model was Journal of Theoretical and Applied Information Technology 31st May 2025. Vol.103. No.10 © Little Lion Scientific ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195 4313 solved in a short amount of time by choosing the immigration domain using the enhanced ant colony algorithm. The main limitation of the work was handling accurate data was impossible. The abstract view of existing works and their disadvantages are presented in Table 1. Table 1: Analysis of existing works and its drawbacks Author name and Reference Approach used Performance Disadvantages Lan et al. [21] Universal Edge-Based Complex Event Processing Mechanism Throughput achieved by the approach is 5.88 Mbps Relatively high cost da Costa Bezerra et al. [22] Euclidean distance and context-aware technique Attain 98% accuracy There were some issues on a potential security basis. Xu et al. [23] BalCon and BalConPlus migration strategy Have 60% of CPU load Dramatically reducing the load imbalance among SDN controllers. Priyadarsini et al. [24] SALB Provide 8.2Mbps of throughput Complexity and cost were major limitations in this existing work. Li et al. [25] Fuzzy Satisfaction-Based Load Balancing Method The average response time of controllers by about 0.33 s Handling accurate data was impossible. From the analysis of existing models, these models have demerits, like relatively high cost in implementation. Some models have issues on a potential security basis, and some have dramatically reduced the load imbalance among SDN controllers. Moreover, complexity and cost were major limitations in this existing work. On the other hand, handling accurate data was impossible. Hence, the proposed model has been implement to overcome the existing flaws. 2.1 Research Gap and Problem Statement Despite the significant advancements in edge computing, fog computing, and Software Defined Networking (SDN), several limitations continue to persist in the existing load balancing and event processing techniques. Most existing works focus either on real-time event processing or on load balancing in SDN-based systems, but seldom integrate both into a unified framework. This disjointed approach leads to inefficiencies in task scheduling and decision-making at the fog and edge layers. Techniques like BalCon, SALB, or fuzzy satisfaction models tend to rely on predefined rules or lack adaptability, making them ineffective in highly dynamic environments with fluctuating workloads and heterogeneous resources. Although heuristic methods like Ant Colony or Genetic Algorithms are used, they often suffer from convergence issues and lack global optimization capabilities. There is a lack of comprehensive hybrid meta-heuristic models that intelligently combine exploration and exploitation for optimized task distribution. Existing models do not fully utilize contextbased task categorization for priority-aware scheduling. This leads to increased task migration, energy consumption, and delayed response time. Based on the identified gaps, the central research question addressed by this study is "How can an integrated, hybrid optimization-based model combining Complex Event Processing and priorityaware scheduling enhance task allocation and load balancing in a fog-enabled SDN environment to reduce energy consumption, migration time, and response delay?" This work addresses the gap by proposing a novel HPC_SOA integrated with CEP and a Multi-tier Priority Queue Scheduling system. This design ensures efficient task categorization, reduced time complexity, intelligent load balancing, and optimal resource utilization in dynamic IoT environments. Journal of Theoretical and Applied Information Technology 31st May 2025. Vol.103. No.10 © Little Lion Scientific ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195 4314 3. PROPOSED METHODOLOGY The rapid development of technology, along with cheaper costs and more connection capacity, have contributed to the IoT's enormous rise in popularity. Infrastructure and data expansion must be matched by a software architecture that facilitates their exploitation. It is challenging to find a software solution that fully utilizes contextual and situational information processing despite numerous proposals focusing on the edge and fog levels. This study uses an efficient transmission model based on the Hybrid Meta-heuristic Model to enhance data transfer while reducing time complexity. The schematic representation of the process flow of the suggested approach is shown in Figure 1. A publicly accessible dataset was used to acquire the data. Initially, the data is moved into the CEP, which is positioned between the fog layer and the IoT layer. Alert will be displayed after every sample of the dataset is processed. In edge IoT devices, complex event processing entails the realtime analysis, correlation and interpretation of continuous streams of data produced by sensors and edge devices. It seeks to identify significant trends or intricate occurrences in various data streams in order to facilitate quick decisions or immediate reactions. After CEP, a multi-tier priority queuebased model is used to attain priority-aware task scheduling. Upon arrival, a task from an application is sorted into slots based on its category. Highpriority tasks are finished first because they are in higher-priority slots than lower-priority slots. Tasks in lower slots are executed only if upper slots are empty. Task execution follows a nonpre-emptive approach, ensuring that the current task is completed before the next one is selected. The major goal of the load balancing mechanism on a fog server in a software-defined network is to improve the fog computing system's performance, efficiency, and reliability. The system model of the proposed Fog Level Load Balancing Mechanism is illustrated in Figure 2. This system supports a range of IoT applications that need improved security, high scalability and low latency. A software-defined network's optimal resource utilization and task response time are guaranteed by the HPC_SOA approach. Arranging tasks based on their availability, capacity, proximity, and energy efficiency may optimize the fog nodes' resource utilization and energy usage. A detailed description of the approaches is given in the further sections. Multi-tier priority queue based model Task scheduling Hybrid Pigeon Cat Search Optimization Algorithms (HPC_SOA). Load balancing Fog layer Bitbrains dataset Dataset Complex Event Processing (CEP) Simple tasks Complex task IoT layer Cloud storage Urgent task Very Urgent task Moderate or non-urgent SDN Controller Figure 1: Schematic Representation Of The Process Flow Of The Proposed Model Journal of Theoretical and Applied Information Technology 31st May 2025. Vol.103. No.10 © Little Lion Scientific ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195 4315 Edge tier Fog tier Cloud tier SDN controller Load balancing Fog server manager Fog nodes Fog region: Mixed resources Fog region: storage resources Fog region: computing resources Fog gateway Fog region: Mixed resources Fog gateway Bitbrains dataset Figure 2: Proposed System Architecture For Load Balancing In Fog Layer 3.1 Complex Event Processing The tasks are divided into simple and complex events using the rules of Complex event processing (CEP). Complex event processing [26] is a computing technique that analyses data streams in real-time to identify and understand the events. It is event-driven, meaning it is triggered by the receipt of event data. Consider an agriculture-related event, which is classified as simple or complicated based on complex event processing. The production goal in the agricultural industry is to maximize crop yield at the lowest possible cost. Making decisions in this situation is complex since several variables influence the entire process, with the primary goal of conserving water for irrigation. To meet the requirements of the atmosphere, a growing plant must extract water from the soil. Evapotranspiration reduces the amount of water stored in the soil; precipitation or irrigation takes its place. Therefore, it is important to characterize the soil features responsible for water retention. Moisture and matric potential are intimately tied to the soil's potential, which increases with humidity. A crop's ability to store enough water during its development is aided by irrigation. When managing irrigation, its matric potential is utilized to choose the kind, quantity, and timing of irrigation. The soil matric potential  containing the highest concentration of crop roots. Another method for figuring out when the optimal moisture for irrigation, or critical moisture  is where irrigation should take place. This metric represents the soil water content at which crop yields begin to decline with the potential for a decrease in evapotranspiration.  Journal of Theoretical and Applied Information Technology 31st May 2025. Vol.103. No.10 © Little Lion Scientific ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195 4316 required for the soil tier which is the most shallow feature. Based on the given concepts, the rules for the process are formed. Normally, there are two types of rules in the CEP such as monitoring rules as well as critical state rules. Monitoring rules: The foundation of the first set of events in the monitoring rule is provided by the primitives of filtration, aggregation, as well as projection. Using a portion of its resources, the projection generates complex events that are handled by the fog tier. Characteristics that are produced by filtering and aggregating. The following incidents made up the first set's rule: A low matric potential alert is sent by EV1 (Simple) A high matric potential alert is sent by EV2 (Simple) A daily matric potential sending alert plus sensor ID is sent by EV3 (Simple). The second level of events is made up of both simple and complex events that are built on the Fog tier's aggregation, enrichment, and composition primitives. The gathering of information needed for computation is referred to as enrichment. Composition refers to the computations made using the data that were examined. The following incidents make up this rule: EV5 (complex): daily irrigation water requirement (IWN) EV6 (complex): daily irrigation frequency verification EV4 (complex): matric potential variation. Critical state rules: The critical state rule consists of events based on the primitives of enrichment, composition, negation, and sequencing performed at the Fog tier. Implementing a sliding time window with an event identification and a timestamp integrated will provide the input control required to use the sequence primitive. The following are the events that make up this rule: EV7 (Complex): maximum water deficit EV8 (Complex): essential soil moisture (perfect for irrigation). According to the given events, the proposed rules of the CEP technique for separating simple and complex events are stated in the pseudocode provided in Table 2. Table 2: Pseudocode Of CEP Approach Start Initialize the task counts and task Apply the CEP rules For task i If soil humidity 50  Update as a complex task Else if wind speed 9< (Km/h) Update as a complex task Else if 1.0<field Irrigation Update as a complex task Else if 93< (Deg)direction Wind Update as a complex task Else if 50<(C) atureAir temper Update as a complex task Else if 101.3<Pressure Update as a complex task Else Update as a simple task End If Stop The CEP of tasks are done in the edge layers, and the complex tasks are sent to the fog layer and then scheduled. 3.2 Multi-Tier Priority Queue-Based Model for Priority Aware Task Scheduling After separating the complex task using CEP, the priority task scheduling [27] will take place to schedule the task by its importance. Initially, consider the scheduled system m SSSS ,...,, 21  in which the Poisson distribution method is used to distribute inter-arrival jobs between two successive tasks. These distribution is given as     !/Pr yeyYYF y     here e represents the Euler's No, 0   and Y represents mean arrival rate. According to the suggested sensor network design, each gateway device assigns tasks to four categories: very urgent, urgent, moderate, and non-urgent. Very Urgent is given high priority since it aims to assist jobs with short deadlines that must be finished fast. Soil humidity, air humidity, and air Journal of Theoretical and Applied Information Technology 31st May 2025. Vol.103. No.10 © Little Lion Scientific ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195 4317 temperature are considered to be the most urgent duties from complicated events, with the remaining tasks being urgent. Furthermore, basic events are categorized as moderate or non-urgent occurrences. To avoid long offloading and downloading periods in an SDN cloud, workloads are sent to neighbouring edge devices or fog layers. The architecture of the Multi-tier priority queue-based model for Priorityaware task scheduling is provided in Figure 3. Task from complex event processing 1 2 1 3 4 2 11 22 3 4 1122 34 Cloud Edge node FCFS Global queue Priority assignment Task classifier Very urgent task Urgent task Non-urgent task Moderate task S1 S2 Figure 3: Architecture Of Multi-Tier Priority Queue-Based Model For Priority-Aware Task Scheduling Sometimes, a more dependable computing device is designated especially for these tasks. An urgent task's structure is essentially equal to that of a very urgent task; however, an urgent task necessitates a little extra waiting time. Moderate tasks have a moderate priority level and a flexible time limit. These tasks negotiate total latency, which means that more time may be provided to complete the task processing in order to meet the deadline, and these jobs are processed by the cloud server. The lowest priority is given to the non-urgent tasks, which are expected to allow time-consuming jobs with no deadline. These tasks require more processing power without a priority on reducing latency. These kinds of tasks are usually offloaded to a centralized cloud server. In order to facilitate additional decision-making and minimize scheduling delays, the gateway devices create nonpreemptive four-level feedback queues. These queues are said to be Very Urgent (QVU), Urgent (QU), Moderate (QM), as well as Non-Urgent (QNU). Work is done on the jobs and queue using a priority-aware task scheduling methodology. The task will be removed from any queue if the waiting time reaches a certain level in order to prevent starvation. When the extremely urgent task queue is empty, the task from the urgent task will be moved to the very urgent task queue. This scheduling algorithm's main job is to use the Utilization Factor j Uf to classify the incoming tasks generated by sensor devices and then assign those jobs to the appropriate queues according to priority. Each task j T in T is assigned a task identifier from a collection of emergency information, such as the task's execution deadline ed j T and frequency of arrival f j T. Utilization factors j Uf govern the priority of each incoming job and are expressed as f j ed jj TTUf /. Then, the tasks are classified as Very Urgent vu j T, Urgent u j T, Moderate m j Tor Non-Urgent nu j T , depending on Journal of Theoretical and Applied Information Technology 31st May 2025. Vol.103. No.10 © Little Lion Scientific ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195 4318 this usage factor. Sensor device-generated tasks will be prioritized in the global queue according to the Gateway's First Come First Served (FCFS) order (i.e., Raspberry Pi). The work will be given an additional priority depending on its utilization factor, and subsequent tasks will be scheduled into the appropriate queues in accordance with their Priority Aware Scheduling Policy. To prevent starvation, a task will be removed from a queue if its waiting time in that queue exceeds a certain threshold. Two pairs of queues were added to the queue model to provide some variety. One set of queues handled urgent work, which was subsequently offloaded to the fog server, and the other pair handled non-urgent activities, which were handled directly by the cloud's offload server. Priority assignment and queue allocation will occur within the gate device. The task priority is defined by the given definition. Definition: From the sensor device, the task j T would arise and be defined as a very urgent task if 4 1  j Uf or moderate if 2 1 3 1 j Uf or urgent if 3 1 4 1 j Uf or non-urgent if 2 1  j Uf . Explanation: Assume six sets of tasks that are said to be 654321 ,,,.,, TTTTTTTj  which has set as  4,4 1 T,  6,4 2 T,   8,3 3 T,  9,2 4 T,   2,1 5 T,   5,4 6 T. According to the utilization factor norms, Task j T utilization is estimated using the expressions as 1  Uf . Similarly, 6.0  Uf , 3.0  Uf , 2.0  Uf , 5.0  Uf and 8.0  Uf . Task 4 Tis classified as a very urgent task since task priority is determined by utilization level. According to the definition, tasks 1 T,2 T and 6 Tare categorized as non-urgent, 3 Tas urgent, and 3 Tas moderate. 3.2.1 Scheduling task In order to create a multilevel feedback Queue model, a Multi-tier priority queue-based model takes into account four queues. The most urgent activities are sent to the Very High Priority queue, which is represented by 1qu . Similar to 1qu , urgent jobs are routed to 2qu which is a lowerpriority queue. Non-urgent tasks were routed to the Low Priority Queue 3qu , while moderate tasks were directed to the queue 4qu with moderate. ()deq is also used to establish a feasible schedule order for data reception based on priority. On scheduler 1: Rule 1: When   u qu2and u jj uuQt jTquWr  ,', task j Tneed higher priority than given by pushing in vu j T. Replace the task to a very urgent queue from an urgent queue based on FCFS, which is expressed as vuu jquTdeq )( . Rule 2: If   vu qu1and   vu qu , )( u j Tdeq from urgent queue to load directly on fog server based on FCFS. Rule 3: If   vu qu1and   u qu , according to the FCFS approach, the task is removed from a very urgent task using uvu jquTdeq )( to load at fog server by the gateway. At scheduler 2: Rule 1: If   nu qu4and nu jj unuQt jTquWr  ,', task j Tneed higher priority, that is given by pushing in m j T. Replace the task to moderate queue from non-urgent queue based on FCFS, which is expressed as mnu jquTdeq )( . Here, Qt j Wr denotes the time quantum Rule 2: If   m qu3and   nu qu , after that, offload j T to the cloud using Priority nu j Tto dequeue it in FCFS sequence. Rule 3: If   m qu3and   nu qu , offload j T to the cloud using Priority m j Tto de-queue it in FCFS sequence. The algorithm using a Multi-tier priority queuebased model for Priority-aware Task Scheduling is presented in Table 3. Journal of Theoretical and Applied Information Technology 31st May 2025. Vol.103. No.10 © Little Lion Scientific ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195 4325 responsiveness by identifying complex patterns from data streams, while the priority queue ensured timely execution of urgent tasks without causing starvation. HPC_SOA further optimized load balancing by adaptively distributing tasks across fog nodes based on real-time fitness evaluations considering energy, proximity, and resource availability. This study significantly advances the current understanding of edge-level task processing and fog-level load balancing in SDNs. While previous works have individually addressed either Complex Event Processing or load balancing using static or mono-strategy optimizations, this research offers a holistic and adaptive framework that combines both. By integrating CEP for intelligent task categorization, a multi-tier queue for dynamic priority scheduling, and a HPC_SOA for optimal load distribution, the proposed model bridges critical gaps identified in the literature namely, the lack of real-time responsiveness, energy-aware processing, and end-to-end task orchestration. The simulation results on a real-world dataset further substantiate its superior performance in reducing response time, energy consumption, and migration overhead compared to existing models. Thus, the study not only provides a technically sound model but also sets a new direction for future research in scalable, intelligent, and adaptive fog computing within SDNenabled IoT environments. REFERENCES [1] Munirathinam, S. (2020). Industry 4.0: Industrial internet of things (IIOT), In Advances in computers. Elsevier 117(1), 129-164. [2] Tian, Z., Shi, W., Wang, Y., Zhu, C., Du, X., Su, S., Sun, Y., Guizani, N. (2019). Real-time lateral movement detection based on evidence reasoning network for edge computing environment, IEEE Transactions on Industrial Informatics 15(7), 4285-94. [3] Qi, Q., Tao, F. (2019). A smart manufacturing service system based on edge computing, fog computing, and cloud computing, IEEE access 7, 86769-77. [4] Salaht, F.A., Desprez, F., Lebre, A. (2020). An overview of service placement problem in fog and edge computing, ACM Computing Surveys (CSUR) 53(3), 1-35. [5] Du, X., Cardie, C. (2020). Event extraction by answering (almost) natural questions, arXiv preprint arXiv:2004.13625. [6] Cortés-Ciriano, I., Lee, J.J., Xi, R., Jain, D., Jung, Y.L., Yang, L., Gordenin, D., Klimczak, L.J., Zhang, C.Z., Pellman, D.S. (2020). Comprehensive analysis of chromothripsis in 2,658 human cancers using whole-genome sequencing, Nature genetics 52(3), 331-41. [7] Subramaniyaswamy, V., Manogaran, G., Logesh, R., Vijayakumar, V., Chilamkurti, N., Malathi, D., Senthilselvan, N. (2019). An ontology-driven personalized food recommendation in IoT-based healthcare system, The Journal of Supercomputing 75, 3184-216. [8] Eskandari, M., Janjua, Z.H., Vecchio, M., Antonelli, F. (2020). Passban IDS: An intelligent anomaly-based intrusion detection system for IoT edge devices, IEEE Internet of Things Journal 7(8), 6882-97. [9] Manzoor, S., Mazhar, F., Binaris, A., Hassan, M.U., Rasab, F., Mohamed, H.G. (2023). An Adaptive Symmetrical Load Balancing Scheme for Next Generation Wireless Networks, Symmetry 15(7), 1316. [10] Moravejosharieh, A.H., Palmeira, F.C. (2019). Load-balancing In Software-Defined Networking: An Investigation on Influential System Parameters, in 2019 29th International Telecommunication Networks and Applications Conference (ITNAC), IEEE 1-6. [11] Lu, J., Zhang, Z., Hu, T., Yi, P., Lan, J. (2019). A survey of controller placement problem in software-defined networking, IEEE Access 7, 24290-307. [12] Urrea, C., Benítez, D. (2021). Software-defined networking solutions, architecture and controllers for the industrial internet of things: A review, Sensors 21(19), 6585. [13] Mokhtar, H., Di, X., Zhou, Y., Hassan, A., Ma, Z., Musa, S. (2021). Multiple-level threshold load balancing in distributed SDN controllers, Computer Networks 198, 108369. [14] Gawade, Y., Harale, A., Ekhe, D., Anjum, S., Lokhande, P. (2019). Video streaming over software defined networking with server load balancing, Research Journal of Engineering Technology and Management (ISSN: 25820028) 2(01). [15] Babbar, H., Rani, S. (2019). Emerging prospects and trends in software defined networking, Journal of Computational and Theoretical Nanoscience 16(10), 4236-41. [16] Hongvanthong, S., Chunlin, L. (2022). A novel four‐tier software‐defined network architecture for scalable secure routing and load balancing, International Journal of Communication Systems 35(1), e5020. Journal of Theoretical and Applied Information Technology 31st May 2025. Vol.103. No.10 © Little Lion Scientific ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195 4326 [17] Li, S., Xin, Z., Xu, X., Zhang, Z. (2020). Load Balancing Algorithm of SDN Controller Based on Dynamic Threshold, In 2023 3rd AsiaPacific Conference on Communications Technology and Computer Science (ACCTCS), IEEE 517-520. [18] Sakthivel, M. (2021). An Analysis of Load Balancing Algorithm Using Software-Defined Network, Turkish Journal of Computer and Mathematics Education (TURCOMAT) 12(9), 578-86. [19] Albowarab, M., Zakaria, N., Abidin, Z. (2019). Load balancing algorithms in software defined network, International Journal of Technology and Engineering (IJRTE) 7. [20] Kofi, E.O., Ahene, E. (2023). Enhanced network load balancing technique for efficient performance in software defined network, Plos one 18(4), e0284176. [21] Lan, L., Shi, R., Wang, B., Zhang, L., Jiang, N. (2019). A universal complex event processing mechanism based on edge computing for internet of things real-time monitoring, IEEE Access 7, 101865-78. [22] da Costa Bezerra, S.F., Filho, A.S., Delicato, F.C., da Rocha, A.R. (2021). Processing complex events in fog-based internet of things systems for smart agriculture, Sensors 21(21), 7226. [23] Xu, Y., Cello, M., Wang, I.C., Walid, A., Wilfong, G., Wen, C.H., Marchese, M., Chao, H.J. (2019). Dynamic switch migration in distributed software-defined networks to achieve controller load balance, IEEE Journal on Selected Areas in Communications 37(3), 515-29. [24] Priyadarsini, M., Mukherjee, J.C., Bera, P., Kumar, S., Jakaria, A.H., Rahman, M.A. (2019). An adaptive load balancing scheme for software-defined network controllers, Computer Networks 164, 106918. [25] Li, G., Cui, W., Liu, S., Zhao, W. (2022). A Load Balancing Strategy Based on Fuzzy Satisfaction Among Multiple Controllers in Software-Defined Networking, Journal of Network and Systems Management 30(3), 49. [26] Kulshrestha, U., Durbha, S. (2020). Edge analytics and complex event processing for real time air pollution monitoring and control, In IGARSS 2020-2020 IEEE International Geoscience and Remote Sensing Symposium, IEEE 893-896. [27] Liao, J.X., Wu, X.W. (2020). Resource allocation and task scheduling scheme in priority-based hierarchical edge computing system, In 2020 19th international symposium on distributed computing and applications for business engineering and science (DCABES), IEEE 46-49. [28] Cui, Z., Zhang, J., Wang, Y., Cao, Y., Cai, X., Zhang, W., and Chen, J. (2019). A pigeoninspired optimization algorithm for manyobjective optimization problems. Science China, Information Sciences 62(7), 70212. [29] Dehghani, M., Hubálovský, Š., Trojovský, P. (2021). Cat and mouse based optimizer: A new nature-inspired optimization algorithm, Sensors 21(15), 5214. [30] http://gwa.ewi.tudelft.nl/datasets/gwa-t-12bitbrains