Cluster Nodes Research Articles

SECURE and energy-efficient routing protocol based on micro-segmentation and batch authentication

An effective technique in designing routing algorithms for Wireless Sensor Networks (WSNs) is named clustering nodes, which augments the network lifespan. More tasks are performed by the Cluster Heads (CHs) of the clustered WSNs, thereby consuming more energy. Hence, a secure energy-efficient Routing Protocol (RP) centered on micro-segmentation and batch authentication is proposed here. The proposed method begins by initializing the WSN nodes. After that, a random key is engendered for every node centered on their Node ID. The nodes undergo a Micro-segmentation process; in addition, they are securely grouped utilizing the Supremum Distance based K-Prototype (SD-KP) algorithm. Now, a fog layer with a number of fog nodes is initialized; also, every single micro-segmented group is assigned with an optimal fog node. Then, by using the Quasi Deterministic Sequence-Black Widow Optimization (QDS-BWO) algorithm, the optimal fog nodes are selected. By using the Root Squared- Diffie Hellman (R2-DH) technique, a key agreement is created for each group centered on their Node IDs. By using the Bitwise Cyclic Shift –BLAKE 512 (BCS-BLAKE-512) algorithm, the created keys are converted into hashcode. Now, for the purpose of performing batch authentication, the hashcode of each group is sent to the trusted authority. Subsequent to verification, data sensing occurs. Lastly, by using the Elliptic Curve Cryptography (ECC) algorithm, the data is securely encrypted and it is transmitted via the routes selected utilizing the Learned Gradient Weight Initialization-based Geography and Energy Aware Routing (LGWI-GEAR) algorithm. The proposed framework's efficacy is proved by the experimental outcomes.

Data-driven allocation of smart grid-connected system based on ant colony optimization algorithm

In the era of advanced technology, integrating and distributing data are crucial in smart grid-connected systems. However, as energy loads continue to increase, practical implementation of these systems faces challenges in resource allocation and lacks efficient data collaboration. In this study, the ant colony optimization algorithm is further investigated for stochastic crossover systems and cluster nodes in intelligent path planning management. To improve the pheromone setting method in smart grid-connected systems, we propose an adaptive intelligent ant colony optimization algorithm called the Group Allocation Optimization Algorithm (GAOA). This algorithm expands the pheromone transmission rate of network nodes, establishes a multi-constrained adaptive model with data mining as the pheromone target, and analyzes the accuracy of resource allocation to import the optimal scheme for smart grid-connected systems. Through experimental results, we demonstrate that the optimized adaptive ant colony algorithm leads to effective improvements in grid-connected systems, pheromone evaluation, data throughput, convergence speed, and data load distribution. These findings provide evidence that the optimized ant colony algorithm is both feasible and effective for resource allocation in smart grid-connected systems.

An energy efficient fusing data gathering protocol in wireless sensor networks

The Internet of Things (IoT) is rapidly developing as a promising technology for today’s digital world. Wireless sensor networks (WSN) are a crucial component of IoT, and one of the major problems faced is energy constraints. The clustering protocol is an effective energy saving solution. In this paper, we propose an energy efficient fusing data gathering protocol via distributed second level clustering and deep learning-based data fusion (EEFDG) for WSN. EEFDG divides the WSN into several second-level clusters by a novel distributed cluster algorithm. The transmission distance and the number of neighbors are jointly considered to select an energy-efficient parent node in each second-level cluster. A new cluster head weight function is proposed for intra-cluster parent node rotation. To minimize redundant data transmission, a multivariate convolutional neural network and long short-term memory network (MVCLNet) are designed for data fusion. MVCLNet is deployed hierarchically into the second-level clustered structure to extract features of the multi-sensor data. The EEFDG is an effective solution to the energy limitation problem of WSN. The proposed second-level clustering and data fusion algorithm based on MVCLNet typically decreases communication distances and the amount of data, thereby minimizing communication overhead and energy consumption. Simulation results show that EEFDG significantly prolongs the network lifetime in terms of the first node dead, compared to existing protocols including MR-LEACH, BPDA, CNNDA, FDEAM, DCNN, and LPLL-LEACH, achieving improvements of 181.3%, 42.2%, 48.2%, 19.6%, 62.0%, and 30.2%, respectively. The EEFDG enables highly accurate data fusion and has superior performance in terms of remaining energy, network lifetime, and data collection.

Research on WSN Intelligent Routing Algorithm based on Bayesian Learning and Particle Swarm Optimization

Background: Wireless sensor networks have the characteristics of strong scalability, easy maintenance, and self-organization, but the energy of nodes is limited and it is difficult to replace the energy supply module. The survival time of the network has always been the key to restricting the development of wireless sensor networks. Objective: Aiming at the problems of short network lifetime and low coverage, a multi-objective optimization routing algorithm has been proposed, focusing on how to balance the communication energy consumption of each node in the network and improve the coverage area of the remaining nodes. Method: Firstly, the node region was divided into several fan ring subregions. Then, the particle swarm optimization algorithm was used to find the fan angles and radii of each fan ring subregion. Next, Bayesian learning was used to select the appropriate cluster head. Results: The simulation results showed the convergence speed of the proposed algorithm to be improved, solving the problems of cluster head election and node routing planning, improving the utilization of node energy, and verifying the effectiveness. Conclusion: The particle swarm optimization algorithm and Bayesian learning have been introduced to cluster network nodes, and a multi-objective fitness function compatible with the energy consumption and coverage of network nodes has been designed. By optimizing the selection method of convergence nodes, the network communication cost of each node can be effectively balanced, and the speed of network coverage area reduction can be effectively reduced in the later period of node communication.

One-step in situ preparation of Zr(Ti)-MOFs on titanium alloys at a low temperature of 80 °C and atmospheric pressure

Preparing metal-organic frameworks (MOFs) film in situ on the surface of the Ti-6Al-4V (TC4) titanium alloy is challenging owing to the dense and stable surface TiO2 passive film. In this study, titanium-doped zirconium-based MOFs (Zr(Ti)-MOFs) film was constructed in situ on the TC4 alloy in one step at a low temperature and atmospheric pressure, with ZrCl4 as the source of Zr metal ions. ZrCl4 attacked and dissolved the TiO2 film to form Zr-O-Ti bonds on the surface of TC4 to provide nucleation sites for further reactions. At the same time, Zr4+ and dissolved Ti4+ combined in solution to form Ti containing Zr-O cluster nodes, which aggregated and grafted to the Zr-O-Ti bonds on the substrate surface, and reacted with organic ligands (terephthalic acid) to form Zr(Ti)-MOFs film. The Ti4+ dissolved from the TC4 participate in the Zr-MOFs formation, thereby changing the high temperature and high-pressure conditions required for the traditional preparation of Zr-MOFs (UiO-66). The in situ preparation of Zr(Ti)-MOFs on TC4 was achieved at 80 °C under atmospheric pressure conditions. This study elucidated the growth mechanism of Zr(Ti)-MOFs on the TC4 surface, providing a reference for the design and In-situ preparation of multifunctional MOFs on the TC4 titanium alloy.

Resting-state cortical electroencephalogram rhythms and network in patients after chronic stroke

ObjectiveTo investigate the resting-state cortical electroencephalogram (EEG) rhythms and networks in patients with chronic stroke and examine their correlation with motor functions of the hemiplegic upper limb.MethodsResting-state EEG data from 22 chronic stroke patients were compared to EEG data from 19 age-matched and 16 younger-age healthy controls. The EEG rhythmic powers and network metrics were analyzed. Upper limb motor functions were evaluated using the Fugl–Meyer assessment-upper extremity scores and action research arm test.ResultsCompared with healthy controls, patients with chronic stroke showed hemispheric asymmetry, with increased low-frequency activity and decreased high-frequency activity. The ipsilesional hemisphere of stroke patients exhibited reduced alpha and low beta band node strength and clustering coefficient compared to the contralesional side. Low beta power and node strength in the delta band correlated with motor functions of the hemiplegic arm.ConclusionThe stroke-affected hemisphere showed low-frequency oscillations and decreased influence and functional segregation in the brain network. Low beta activity and redistribution of delta band network between hemispheres were correlated with motor functions of hemiplegic upper limb, suggesting a compensatory mechanism involving both hemispheres post-stroke.

Assessment and solutions for vulnerability of urban rail transit network based on complex network theory: A case study of Chongqing

As a typical complex network system, the operating environment of rail transit network (RTN) is complex and demanding. This study aims to accurate assess the weaknesses and vulnerability of RTN, which is crucial for ensuring its smooth operation. Taking Chongqing Rail Transit (CRT) as an example, this study developed a network topology model using the spatial L method and analyzed the network structure characteristics, along with the importance of key nodes under different indicators, based on complex network theory. Additionally, this study analyzed the geographical spatial distribution characteristics of nodes based on the topography and urban spatial structure of Chongqing. Then, this study classified the nodes in the RTN according to basic topological indicators, namely degree, betweenness centrality, network efficiency, and passenger flow volume (PFV). The results indicated six cluster of nodes, reflecting the variability in node vulnerability concerning overall influence (providing alternative paths, reducing path length), regional aggregation capacity, and transportation capacity. Finally, this study proposed targeted management strategies for different clusters of nodes and their respective geographical locations, providing necessary references for rational planning, safety protection, and sustainable construction of RTN.

TransNeT-CGP: A cluster-based comorbid gene prioritization by integrating transcriptomics and network-topological features

The local disruptions caused by the genes of one disease can influence the pathways associated with the other diseases resulting in comorbidity. For gene therapies, it is necessary to prioritize the key genes that regulate common biological mechanisms to tackle the issues caused by overlapping diseases. This work proposes a clustering-based computational approach for prioritising the comorbid genes within the overlapping disease modules by analyzing Protein-Protein Interaction networks. For this, a sub-network with gene interactions of the disease pair was extracted from the interactome. The edge weights are assigned by combining the pairwise gene expression correlation and betweenness centrality scores. Further, a weighted graph clustering algorithm is applied and dominant nodes of high-density clusters are ranked based on clustering coefficients and neighborhood connectivity. Case studies based on neurodegenerative diseases such as Amyotrophic Lateral Sclerosis- Spinal Muscular Atrophy (ALS-SMA) pair and cancers such as Ovarian Carcinoma-Invasive Ductal Breast Carcinoma (OC-IDBC) pair were conducted to examine the efficacy of the proposed method. To identify the mechanistic role of top-ranked genes, we used Functional and Pathway enrichment analysis, connectivity analysis with leave-one-out (LOO) method, analysis of associated disease-related protein complexes, and prioritization tools such as TOPPGENE and Heml2.0. From pathway analysis, it was observed that the top 10 genes obtained using the proposed method were associated with 10 pathways in ALS-SMA comorbidity and 15 in the case of OC-IDBC, while that in similar methods like SAPDSB and S2B were 4, 6 respectively for ALS-SMA and 9, 10 respectively for OC-IDBC. In both case studies, 70 % of the disease-specific benchmark protein complexes were linked to top-ranked genes of the proposed method while that of SAPDSB and S2B were 55 % and 60 % respectively. Additionally, it was found that the removal of the top 10 genes disconnect the network into 14 distinct components in the case of ALS-SMA and 9 in the case of OC-IDBC. The experimental results shows that the proposed method can be effectively used for identifying key genes in comorbidity and can offer insights about the intricate molecular relationship driving comorbid diseases.

IMN4NPD: An Integrated Molecular Networking Workflow for Natural Product Dereplication.

Molecular networking has emerged as a standard approach for natural product (NP) discovery. However, the current pipeline based on molecular networks tends to prioritize larger clusters comprising multiple nodes. To address this issue, we present the integrated molecular networking workflow for NP dereplication (IMN4NPD). This approach not only expedites the rapid dereplication of extensive clusters within the molecular network but also places specific emphasis on self-looped or pairs of nodes, which are often overlooked by the current methods. By amalgamating the outputs from various computational tools, we efficiently dereplicate compounds falling into specific categories and provide annotations for both large cluster nodes and self-looped or pair of nodes within the molecular network. Furthermore, we have incorporated several fundamentally distinct similarity algorithms, namely, Spec2Vec and MS2DeepScore, for constructing the t-SNE network. Through comparison with modified cosine similarity, we have observed that integrating additional diverse spectral similarity measures, the resulting t-SNE network enhanced the ability to dereplicate NPs. Demonstrating the use case of an ethanol extract of Plumula nelumbinis, we illustrate that an integration of multiple computational solutions with IMN4NPD aids the dereplication, especially self-looped nodes, and in the discovery of novel compounds in NPs.

Similarities and Differences Between Study Addiction and Study Engagement and Work Addiction and Work Engagement: A Network Analysis

AbstractThis study aimed to investigate the relationships between symptoms of study addiction and their relationships with dimensions of study engagement. We used two samples in which study addiction was measured with the Bergen Study Addiction Scale and study engagement was measured with the Utrecht Work Engagement Scale for Students. The samples comprised Norwegian (n = 1107) and Polish (n = 776) students. The networks featured three clusters of nodes (two clusters of the study addiction symptoms and one cluster of the study engagement dimensions). Study addiction clusters were connected with study engagement cluster through positive edges between absorption and study addiction symptoms, negative edges which vigor shared with conflict and problems, and negative edge between dedication and mood modification. Among the symptoms of study addiction, problems, conflict, and withdrawal were the most central and salience, tolerance, and mood modification were the least central. Moreover, conflict was the most predictable and relapse was the least predictable symptom of study addiction. The results from the present study are similar to those of previous studies on work addiction and support the notion that study addiction may be a precursor to work addiction. Although based on cross-sectional data, the nuanced differences between work addiction and study addiction networks are discussed as they lay a foundation for further investigation of the potential differences in transition mechanisms from healthy engagement to compulsive behavior.

Data transmission via hybrid fuzzy time slot scheduling in long-range communication

ABSTRACT Several communication protocols have been tuned for LoRaWAN since the development of LoRa technology more than 10 years ago. LPWAN technology can address issues like overlooked delivery ratio, reducing low frequency and limited data availability, thereby improving system effectiveness, and enormous packet drops due to collisions will exist. Hence the hybrid fuzzy time slot scheduling is proposed to avoid collision in the packet with secure data transmission in the LoRa communication. The fuzzy C means clustering algorithm is suggested to cluster nodes and select the Cluster head based on distance and energy usage. The packet transmission is carried out using the time slot scheduling mechanism by TDMA, which sends the packet without collision. The proposed hybrid method can improve data transmission reliability by minimising collisions, optimising time slots, improving network capacity and decreasing idle time. The outcomes reveal that the proposed hybrid method attains high bandwidth with frequency, throughput and PDR also the PLR attains low without collision of a packet received when related to other prior techniques.

An Overview of Advanced Deep Graph Node Clustering

Graph data have become increasingly important, and graph node clustering has emerged as a fundamental task in data analysis. In recent years, graph node clustering has gradually moved from traditional shallow methods to deep neural networks due to the powerful representation capabilities of deep learning. In this article, we review some representatives of the latest graph node clustering methods, which are classified into three categories depending on their principles. Extensive experiments are conducted on real-world graph datasets to evaluate the performance of these methods. Four mainstream evaluation performance metrics are used, including clustering accuracy, normalized mutual information, adjusted rand index, and F1-score. Based on the experimental results, several potential research challenges and directions in the field of deep graph node clustering are pointed out. This work is expected to facilitate researchers interested in this field to provide some insights and further promote the development of deep graph node clustering.

Performance of ML-Based Unsupervised Clustering Algorithms for WSN Node Clustering

Abstract: Wireless Sensor Networks are generally deployed in dynamically changing environment. When compared to common wired network nodes, WSN nodes must do more work. Since WSN devices are battery powered, so power management is a challenge. Clustering is one solution that has been proposed to alleviate the issue of limited power. Clustering is the most important method stabilizes the lifetime of the network. It entails the aggregation of sensor nodes into clusters and cluster head is picked out from all the clusters. Clustering is implemented in wireless sensor networks through the Machine Learning techniques. In wireless sensor networks, machine learning algorithms have an important role in cluster head formation and maintain the stability of the nodes in the cluster. Machine Learning approaches used in wireless sensor networks can be classified as Supervised learning, Unsupervised learning, and Reinforcement learning. Among these learning techniques, unsupervised learning deals with different clustering algorithms such as k-means, K-medoids, Fuzzy C-means, hierarichalbased, and SOM. This paper evaluates the performance of the variants of k-Means (kM) and Fuzzy C-means (FCM) algorithms in terms of the clustering and accuracy. This paper imparts performance analysis of different clustering algorithms in machine learning applied for wireless sensor networks. From the analysis, the Fuzzy C-Means algorithm found to be more suitable for node clustering in WSN

The relations of psychotic-like experiences (PLEs) and depressive symptoms and the bias of depressive symptoms during the clustering among Chinese adolescents: Findings from the network perspective

BackgroundThere are rare studies about the network structure of psychotic-like experiences (PLEs) and depressive symptoms among adolescents. Studies have widely acknowledged that PLEs in adolescents confer a higher risk of depressive symptoms, but the complex interactions remain inadequately understood. Our study aimed to examine the hierarchy and inter-associations of PLEs and depressive symptoms in a large adolescent sample from the network analysis perspective. MethodsA total of 5008 Chinese adolescents were enrolled in our sample. Community Assessment of Psychic Experiences-42 (CAPE-42) was applied to build the network. Centrality indexes were calculated to represent the significance of nodes in the network. Community detection was conducted to figure out the specific clustering of nodes. Demographic information was collected for the sub-network comparisons. Accuracy and stability of the network were also tested. Results“Failure”, “External control”, and “Lack of activity” were the most central nodes. The main bridge nodes linking PLEs and depressive symptoms were “Failure”, “Guilty”, and “No future”. Positive PLE “Odd looks” and negative PLE “Unable to terminate” are the two PLEs that were most relevant to depressive nodes. Community detection further demonstrated the bias of depressive nodes in the data-driven clustering. Comparative sub-network analysis suggested that age was the only demographic factor related to the current network. ConclusionIn this study of a large adolescent sample, we first demonstrated the network structure and specific clustering preference of PLEs and depressive symptoms.Our findings may enhance the understanding of the relationship between PLE and depressive symptoms.

Modeling and solving the two-level generalized median tour problem

In many real-world situations, the beneficiaries of a distribution system may be grouped into clusters, requiring a transportation structure to serve each cluster of beneficiaries efficiently. For instance, in public services, such as health, education, and emergencies, delivery tasks rely on the local authority of each district or cluster. In this context, a two-level distribution system named the Two-Level Generalized Median Tour Problem (TLGMTP) is introduced. The first level addresses product distribution using a specialized vehicle, starting and ending at a depot and visiting some clusters. In this manner, products are delivered to one or more nodes belonging to the visited clusters. The second level comprises smaller vehicles that start their trips from the nodes belonging to the first level and transport the products to one or more nodes located in a non-visited cluster, ensuring that all non-visited clusters of the first level are visited in the second level. Then, the non-visited nodes in each cluster must reach a node in the same cluster to collect their products. In this study, we present, model, and solve the TLGMTP to minimize the total transportation costs. We develop three mathematical formulations and solve them using a branch-and-cut algorithm. Exhaustive computational experiments involving tests and real-world instances are presented to show the efficiency and advantages of the proposed methodology.

Energy capable protocol for heterogeneous blue brain network

The Blue Brain has a wide range of applications, which raises a number of challenging issues. Electronics may continuously monitor their surroundings depending on the real data that their Blue Brain nodes are acquiring by employing situational intelligence based on the Blue Brain environment. The Blue Brain does more than only monitor user behavior when utilizing this technology. Blue Brain is linked to a critical prerequisite for energy-efficient communication methodologies. Through the Blue Brain network, it utilizes the heterogeneity and variety of the interconnected components. Blue Brain nodes that are outsourced and have limited energy resources must utilize less energy. IoT nodes with differing energy levels are frequently dispersed across different geographic regions. The main goal of this work is to provide an energy-efficient Blue Brain framework capable of managing cluster head (CH) selection and Blue Brain node clustering. The appropriate CHs are selected, and an energetic cutoff concept is developed to guarantee that energy is shared equally among the CHs and participating Blue Brain nodes. The proposed concept envisions three different kinds of Blue Brain nodes for a Blue Brain infrastructure: expert, intermediary, and normal Blue Brain nodes. Level 1 Blue Brain nodes are regarded as normal nodes; level 2 nodes are regarded as intermediate Blue Brain nodes; and level 3 nodes are regarded as expert Blue Brain nodes. Level 1 Blue Brain nodes use the least amount of energy. The outcomes of the simulation demonstrate that the recommended strategy outperforms other existing methods.

Uncertainty in GNN Learning Evaluations: A Comparison between Measures for Quantifying Randomness in GNN Community Detection.

(1) The enhanced capability of graph neural networks (GNNs) in unsupervised community detection of clustered nodes is attributed to their capacity to encode both the connectivity and feature information spaces of graphs. The identification of latent communities holds practical significance in various domains, from social networks to genomics. Current real-world performance benchmarks are perplexing due to the multitude of decisions influencing GNN evaluations for this task. (2) Three metrics are compared to assess the consistency of algorithm rankings in the presence of randomness. The consistency and quality of performance between the results under a hyperparameter optimisation with the default hyperparameters is evaluated. (3) The results compare hyperparameter optimisation with default hyperparameters, revealing a significant performance loss when neglecting hyperparameter investigation. A comparison of metrics indicates that ties in ranks can substantially alter the quantification of randomness. (4) Ensuring adherence to the same evaluation criteria may result in notable differences in the reported performance of methods for this task. The W randomness coefficient, based on the Wasserstein distance, is identified as providing the most robust assessment of randomness.

SWARAM: Osprey Optimization Algorithm-Based Energy-Efficient Cluster Head Selection for Wireless Sensor Network-Based Internet of Things.

The Internet of Things (IoT) has transformed various aspects of human life nowadays. In the IoT transformative paradigm, sensor nodes are enabled to connect multiple physical devices and systems over the network to collect data from remote places, namely, precision agriculture, wildlife conservation, intelligent forestry, and so on. The battery life of sensor nodes is limited, affecting the network's lifetime, and requires continuous maintenance. Energy conservation has become a severe problem of IoT. Clustering is essential in IoT to optimize energy efficiency and network longevity. In recent years, many clustering protocols have been proposed to improve network lifetime by conserving energy. However, the network experiences an energy-hole issue due to picking an inappropriate Cluster Head (CH). CH node is designated to manage and coordinate communication among nodes in a particular cluster. The redundant data transmission is avoided to conserve energy by collecting and aggregating from other nodes in clusters. CH plays a pivotal role in achieving efficient energy optimization and network performance. To address this problem, we have proposed an osprey optimization algorithm based on energy-efficient cluster head selection (SWARAM) in a wireless sensor network-based Internet of Things to pick the best CH in the cluster. The proposed SWARAM approach consists of two phases, namely, cluster formation and CH selection. The nodes are clustered using Euclidean distance before the CH node is selected using the SWARAM technique. Simulation of the proposed SWARAM algorithm is carried out in the MATLAB2019a tool. The performance of the SWARAM algorithm compared with existing EECHS-ARO, HSWO, and EECHIGWO CH selection algorithms. The suggested SWARAM improves packet delivery ratio and network lifetime by 10% and 10%, respectively. Consequently, the overall performance of the network is improved.

BigDEC: A multi-algorithm Big Data tool based on the [formula omitted]-mer spectrum method for scalable short-read error correction

Despite the significant improvements in both throughput and cost provided by modern Next-Generation Sequencing (NGS) platforms, sequencing errors in NGS datasets can still degrade the quality of downstream analysis. Although state-of-the-art correction tools can provide high accuracy to improve such analysis, they are limited to apply a single correction algorithm while also requiring long runtimes when processing large NGS datasets. Furthermore, current parallel correctors generally only provide efficient support for shared-memory systems lacking the ability to scale out across a cluster of multicore nodes, or they require the availability of specific hardware devices or features. In this paper we present a Big Data Error Correction (BigDEC) tool that overcomes all those limitations by: (1) implementing three different error correction algorithms based on the widely extended k-mer spectrum method; (2) providing scalable performance for large datasets by efficiently exploiting the capabilities of Big Data technologies on multicore clusters based on commodity hardware; (3) supporting two different Big Data processing frameworks (Spark and Flink) to provide greater flexibility to end users; (4) including an efficient, stream-based merge operation to ease downstream processing of the corrected datasets; and (5) significantly outperforming existing parallel tools, being up to 79% faster on a 16-node multicore cluster when using the same underlying correction algorithm. BigDEC is publicly available to download at https://github.com/UDC-GAC/BigDEC.

Earth system modeling on modular supercomputing architecture: coupled atmosphere–ocean simulations with ICON 2.6.6-rc

Abstract. The confrontation of complex Earth system model (ESM) codes with novel supercomputing architectures poses challenges to efficient modeling and job submission strategies. The modular setup of these models naturally fits a modular supercomputing architecture (MSA), which tightly integrates heterogeneous hardware resources into a larger and more flexible high-performance computing (HPC) system. While parts of the ESM codes can easily take advantage of the increased parallelism and communication capabilities of modern GPUs, others lag behind due to the long development cycles or are better suited to run on classical CPUs due to their communication and memory usage patterns. To better cope with these imbalances between the development of the model components, we performed benchmark campaigns on the Jülich Wizard for European Leadership Science (JUWELS) modular HPC system. We enabled the weather and climate model Icosahedral Nonhydrostatic (ICON) to run in a coupled atmosphere–ocean setup, where the ocean and the model I/O is running on the CPU Cluster, while the atmosphere is simulated simultaneously on the GPUs of JUWELS Booster (ICON-MSA). Both atmosphere and ocean are running globally with a resolution of 5 km. In our test case, an optimal configuration in terms of model performance (core hours per simulation day) was found for the combination of 84 GPU nodes on the JUWELS Booster module to simulate the atmosphere and 80 CPU nodes on the JUWELS Cluster module, of which 63 nodes were used for the ocean simulation and the remaining 17 nodes were reserved for I/O. With this configuration the waiting times of the coupler were minimized. Compared to a simulation performed on CPUs only, the MSA approach reduces energy consumption by 45 % with comparable runtimes. ICON-MSA is able to scale up to a significant portion of the JUWELS system, making best use of the available computing resources. A maximum throughput of 170 simulation days per day (SDPD) was achieved when running ICON on 335 JUWELS Booster nodes and 268 Cluster nodes.