Dynamic Load Balancing Scheme Research Articles

A Review Load balancing algorithms in Fog Computing

With the rapid advance of the Internet of Things (IoT), technology has entered a new era. It is changing the way smart devices relate to such fields as healthcare, smart cities, and transport. However, such rapid expansion also challenges data processing, latency, and QoS. This paper aims to consider fog computing as a key solution for addressing these problems, with a special emphasis on the function of load balancing to improve the quality of service in IoT environments. In addition, we study the relationship between IoT devices and fog computing, highlighting why the latter acts as an intermediate layer that can not only reduce delays but also achieve efficient data processing by moving the computational resources closer to where they are needed. Its essence is to analyze various load balancing algorithms and their impact in fog computing environments on the performance of IoT applications. Static and dynamic load balancing strategies and algorithms have been tested in terms of their impact on throughput, energy efficiency, and overall system reliability. Ultimately, dynamic load balancing methods of this sort are better than static ones for managing load in fog computing scenarios since they are sensitive to changing workloads and changes in the system. The paper also discusses the state of the art of load balancing solutions, such as secure and sustainable techniques for Edge Data Centers (EDCs), It manages the allocation of resources for scheduling. We aim to provide a general overview of important recent developments in the literature while also pointing out limitation where improvements might be made. To this end, we set out to better understand and describe load balancing in fog computing and its importance for improving QoS. We thus hope that a better understanding of load balancing technologies can lead us towards more resilient and secure systems.

Rapid 3D modelling: Clustering method based on dynamic load balancing strategy

AbstractThree‐dimensional (3D) reconstruction is a pivotal research area within computer vision and photogrammetry, offering a valuable foundation of data for the development of smart cities. However, existing methods for constructing 3D models from unmanned aerial vehicle (UAV) images often suffer from slow processing speeds and low central processing unit (CPU)/graphics processing unit (GPU) utilization rates. Furthermore, the utilization of cluster‐based distributed computing for 3D modelling frequently results in inefficient resource allocation across nodes. To address these challenges, this paper presents a novel approach to 3D modelling in clusters, incorporating a dynamic load‐balancing strategy. The method divides the 3D reconstruction process into multiple stages to lay the groundwork for distributing tasks across multiple nodes efficiently. Instead of traditional traversal‐based communication, this approach employs gossip communication techniques to reduce the network overhead. To boost the modelling efficiency, a dynamic load‐balancing strategy is introduced that prevents nodes from becoming overloaded, thus optimizing resource usage during the modelling process and alleviating resource waste issues in multidevice clusters. The experimental results indicate that in small‐scale data environments, this approach improves CPU/GPU utilization by 35.8%/23.4% compared with single‐machine utilization. In large‐scale data environments for cluster‐based 3D modelling tests, this method enhances the average efficiency by 61.4% compared with traditional 3D modelling software while maintaining a comparable model accuracy.

A parallel p‐adaptive discontinuous Galerkin method for the Euler equations with dynamic load‐balancing on tetrahedral grids

AbstractA novel p‐adaptive discontinuous Galerkin (DG) method has been developed to solve the Euler equations on three‐dimensional tetrahedral grids. Hierarchical orthogonal basis functions are adopted for the DG spatial discretization while a third order TVD Runge‐Kutta method is used for the time integration. A vertex‐based limiter is applied to the numerical solution in order to eliminate oscillations in the high order method. An error indicator constructed from the solution of order and is used to adapt degrees of freedom in each computational element, which remarkably reduces the computational cost while still maintaining an accurate solution. The developed method is implemented with under the Charm++ parallel computing framework. Charm++ is a parallel computing framework that includes various load‐balancing strategies. Implementing the numerical solver under Charm++ system provides us with access to a suite of dynamic load balancing strategies. This can be efficiently used to alleviate the load imbalances created by p‐adaptation. A number of numerical experiments are performed to demonstrate both the numerical accuracy and parallel performance of the developed p‐adaptive DG method. It is observed that the unbalanced load distribution caused by the parallel p‐adaptive DG method can be alleviated by the dynamic load balancing from Charm++ system. Due to this, high performance gain can be achieved. For the testcases studied in the current work, the parallel performance gain ranged from 1.5× to 3.7×. Therefore, the developed p‐adaptive DG method can significantly reduce the total simulation time in comparison to the standard DG method without p‐adaptation.

DLB: A Dynamic Load Balance Strategy for Distributed Training of Deep Neural Networks

Synchronous strategies with data parallelism are widely utilized in distributed training of Deep Neural Networks (DNNs), largely owing to their easy implementation yet promising performance. In these strategies, the workers with different computational capabilities need to wait for each other because of the essential gradient or weight synchronization. This will inevitably cause the high-performance workers to waste time waiting for the weak computational workers, which in turn results in the inefficiency of the cluster. In this paper, we propose a Dynamic Load Balance (DLB) strategy for the distributed training of DNNs to tackle this issue. Specifically, the performance of each worker is evaluated first based on the performance demonstration during the previous training epochs, and then the batch size and dataset partition are adaptively adjusted in consideration of the current performance of the workers. As a result, the waiting cost among the workers will be eliminated, thereby the utilization of the clusters is highly improved. Furthermore, the essential theoretical analysis has also been provided to justify the convergence of the proposed algorithm. Extensive experiments have been conducted on the CIFAR10 and CIFAR100 benchmark datasets with four state-of-the-art DNN models. The experimental results indicate that the proposed algorithm can significantly improve the utilization of the distributed cluster. In addition, the proposed algorithm can also prevent the load imbalance of the distributed DNN training from being affected by the disturbance and can be employed flexibly in conjunction with the other synchronous distributed DNN training methods.

DSAG: A Mixed Synchronous-Asynchronous Iterative Method for Straggler-Resilient Learning

We consider straggler-resilient learning. In many previous works, e.g., in the coded computing literature, straggling is modeled as random delays that are independent and identically distributed between workers. However, in many practical scenarios, a given worker may straggle over an extended period of time. We propose a latency model that captures this behavior and is substantiated by traces collected on Microsoft Azure, Amazon Web Services (AWS), and a small local cluster. Building on this model, we propose DSAG, a mixed synchronous-asynchronous iterative optimization method, based on the stochastic average gradient (SAG) method, that combines timely and stale results. We also propose a dynamic load-balancing strategy to further reduce the impact of straggling workers. We evaluate DSAG for principal component analysis, cast as a finite-sum optimization problem, of a large genomics dataset, and for logistic regression on a cluster composed of 100 workers on AWS, and find that DSAG is up to about 50% faster than SAG, and more than twice as fast as coded computing methods, for the particular scenario that we consider.

Enabling Large Scale Simulations for Particle Accelerators

International high-energy particle physics research centers, like CERN and Fermilab, require excessive studies and simulations to plan for the upcoming upgrades of the world's largest particle accelerators, and the design of future machines given the technological challenges and tight budgetary constraints. The Beam Longitudinal Dynamics (BLonD) simulator suite incorporates the most detailed and complex physics phenomena in the field of longitudinal beam dynamics, required for providing extremely accurate predictions. Modern challenges in beam dynamics dictate for longer, larger and numerous simulation studies to draw meaningful conclusions that will drive the baseline choices for the daily operation of current machines and the design choices of future projects. These studies are extremely time consuming, and would be impractical to perform without a High-Performance Computing oriented simulator framework. In this article, at first, we design and evaluate a highly-optimized distributed version of BLonD. We combine approximate computing techniques, and leverage a dynamic load-balancing scheme to relax synchronization and improve scalability. In addition, we employ GPUs to accelerate the distributed implementation. We evaluate the highly optimized distributed beam longitudinal dynamics simulator in a supercomputing system and demonstrate speedups of more than two orders of magnitude when run on 32 GPU platforms, w.r.t. the previous state-of-art. By driving a wide range of new studies, the proposed high performance beam longitudinal dynamics simulator forms an invaluable tool for accelerator physicists.

A novel MPI-based parallel smoothed particle hydrodynamics framework with dynamic load balancing for free surface flow

This paper presents a new Smoothed Particle Hydrodynamics (SPH) parallel framework, which is designed for free surface flows and is scalable on a High Performance Computer (HPC). The framework is accomplished by adopting a Message Passing Interface (MPI) approach with a domain partitioning strategy. A regular background grid is used to partition the entire computational domain and each subdomain is labelled using an index ordering method. Adjacent subdomains can be determined by the index list, and avoid global communications in the particle distribution process. Within the local grid, the grid is divided into an internal grid as well as an interactive grid to identify the particles for which information is to be transferred. The implementation of the dynamic loading balance strategy considers two different ways of determining loading: computation particle numbers and running time. The dynamic load balance strategy repositions neighbouring subdomains based on the local load imbalance between cores. To demonstrate the framework's capacity and distinctive properties, a variety of free surface flow benchmarks are studied. Intensive numerical experiments at various scales are used to assess the performance in detail.

Challenges of Load Balancing Techniques in Cloud Environment: A Review

To highlight issues and challenges of the Load Balancing Algorithm (LBA) in the context of cloud computing. The authors of this work discussed load balancing and the load balancing problems in cloud computing, how to solve them, the basic techniques used to maintain load over the cloud, and the benefits and drawbacks of these algorithms. Furthermore, the static and dynamic load balancing strategies are compared. Load balancing methods in the cloud for data centers must avoid VM overload/low load, which is a challenge in cloud computing. Load balancing mechanisms aid in the management of workloads. The primary goal of load balancing among machines is to spread workloads and computing resources across one or more servers while lowering energy consumption and maximizing resource usage. Cloud computing is a type of computing in which scalable resources are made available on-demand and are frequently virtual, as in an online service. For a variety of reasons, including cost saving, speed, efficiency, performance, and security, cloud computing has recently become a popular choice for individuals and enterprises. Cloud computing uses the Internet to deliver a variety of services to a company's computers and devices. These resources can include data storage, databases, servers, networks, and software, among other tools and applications. In this paper, developers and researchers must design and construct a load balancer e for parallel and distributed cloud systems.

Parallel multiphysics simulation for the stabilized Optimal Transportation Meshfree (OTM) method

This paper presents a parallel implementation for the Optimal Transportation Meshfree (OTM) method on large CPU clusters. Communications are handled with the Message Passing Interface (MPI). The Recursive Coordinate Bisection (RCB) algorithm is utilized for domain decomposition and for implementing dynamic load-balancing strategy. This work involves three new concepts to reduce the computational efforts: Dynamic halo regions, Efficient data management strategies for ease of addition and deletion of nodes and material points using advanced STL container, and nearest neighborhood communication for detection of neighbors and communication. Also, Linked Cell approach has been implemented to further reduce the computational efforts. Parallel performance analysis is investigated for challenging multiphysics applications like Taylor rod impact and serrated chip formation process. Adequate scalability of parallel implementation for these applications is reported.

Research on Performance Optimization for Power Big Data Storage based on HBase

With the development of smart grid, the online monitoring of power equipment has become the focus. Due to the increasing of power data, data storage in power grid is faced with great challenge. HBase is widely used for storing and managing power data due to its excellent performance in distributed access to data. However, HBase is prone to write a hot spot problem when storing data. To address this problem, this paper proposes an optimization method for data storage performance based on a dynamic load balancing strategy, which is implemented by pre-partitioning processing, RowKey substitution, and WLC (Weighted Least Connection) algorithm adapted to HBase storage mechanism. Through the experimental analysis of insulator current data storage, the results show that the method improves the performance of HBase cluster parallel data storage.

EdgeSafe: Dynamic Load Balancing Among Edge Nodes for Provisioning Safety-as-a-Service in Vehicular IoT Applications

In this paper, we propose a dynamic load balancing scheme, EdgeSafe, for provisioning Safety-as-a-Service (Safe-aaS) (Roy et al., 2018). Typically, in a Safe-aaS infrastructure, the sensor nodes are either static or mobile in nature. With the variation in the geographical location of the vehicles, the sensor nodes attached to them attain mobility. Consequently, the distance between the mobile sensor node and the edge nodes present within their vicinity changes. As the data is time-critical, it is necessary to be primarily processed at the edge nodes. Considering road transportation as the application scenario of Safe-aaS, we perform load balancing in two stages. In the first stage, we calculate the preferred capacity ratio of the edge nodes present within the communication range of the sensor nodes. We apply Markowitz Portfolio Selection Theory in the second stage to select the appropriate edge node. The profit return and risk incurred in their selection is calculated to design the portfolio of the edge nodes. Thereafter, the utility of each edge node is computed. We formulate an optimization function to obtain the minimum value of the utility of the edge nodes. Extensive simulation results show that the proposed scheme, EdgeSafe, is capable of improving the data rate by 41.37%, 35.64%, and 21.05% respectively, compared to the existing schemes, $HO$ (Park et al. , 2018), $MLB$ (Lobinger et al. , 2010), and Honeybee (Fernando et al. , 2019).

Cloak-Reduce Load Balancing Strategy for Mapreduce

The advent of Big Data has seen the emergence of new processing and storage challenges. These challenges are often solved by distributed processing. Distributed systems are inherently dynamic and unstable, so it is realistic to expect that some resources will fail during use. Load balancing and task scheduling is an important step in determining the performance of parallel applications. Hence the need to design load balancing algorithms adapted to grid computing. In this paper, we propose a dynamic and hierarchical load balancing strategy at two levels: Intrascheduler load balancing, in order to avoid the use of the large-scale communication network, and interscheduler load balancing, for a load regulation of our whole system. The strategy allows improving the average response time of CLOAK-Reduce application tasks with minimal communication. We first focus on the three performance indicators, namely response time, process latency and running time of MapReduce tasks.

Cloak-Reduce Load Balancing Strategy for Mapreduce

The advent of Big Data has seen the emergence of new processing and storage challenges. These challenges are often solved by distributed processing. Distributed systems are inherently dynamic and unstable, so it is realistic to expect that some resources will fail during use. Load balancing and task scheduling is an important step in determining the performance of parallel applications. Hence the need to design load balancing algorithms adapted to grid computing. In this paper, we propose a dynamic and hierarchical load balancing strategy at two levels: Intrascheduler load balancing, in order to avoid the use of the large-scale communication network, and interscheduler load balancing, for a load regulation of our whole system. The strategy allows improving the average response time of CLOAK-Reduce application tasks with minimal communication. We first focus on the three performance indicators, namely response time, process latency and running time of MapReduce tasks.

Dynamic energy efficient load balancing strategy for computational grid

AbstractComputational grid is a pool of computational resources that is used by various organizations for storage purposes and execution of a large number of tasks. Load balancing and energy consumption are two core performance parameters in the computational grid. Since, performance of a computational grid is highly dependable to variations in load and energy consumption, therefore, it is important to pay attention to these aspects in grid. Therefore, we proposed an energy‐aware dynamic load balancing strategy for tasks scheduling in grid referred to as dynamic energy efficient load balancing strategy (DEEL) for computational grid. DEEL combined energy efficiency and load balancing paradigms in order to improve the utilization and reduce the energy consumption of resources. Simulations were performed by using GridSim. Simulation results were compared with published strategy IEGDC with respect to performance indices: finished gridlets, unfinished gridlets, resubmitted times, and response time. The simulation results were evaluated through performance metrics that is, (a) cost, (b) energy consumption, and (c) energy saved. As per simulation results, DEEL outperformed the published method IEGDC by reducing the execution time of finished gridlets up to 28%, minimizing the execution and resource cost up to 35%, and saving the energy consumption up to 38%.

A Proposed Adaptive Least Load Ratio Algorithm to Improve Resources Management in Software Defined Network OpenFlow Environment

So far, the existing load-balancing schemes (static and dynamic) have not taken over a concise and reasonable mechanism for systematic isolation of the resource counters monitored in the OpenFlow switch periodically. To address the aforementioned issue, the adaptive least load ratio algorithm was proposed and followed a straightforward procedure to generate the best possible decisions based on significant motivation characteristics identified in dynamic load balancing schemes like Least Connection-Based and Least Bandwidth-Based. the study was carried out using the HTTPerf tool as it provided a flexible facility for the generation of various HTTP workloads to evaluate server performance. Results revealed that, while Software defined network was under the influence of client requests load ranging from 0 up to 180 (req/sec), the proposed algorithm confirmed a faster server reply time up to 13.37%, server connection time up to 16.3% and average network throughput was improved up to 8%. Moreover, server CPU utilization time and average queue length were reduced to 2.5% and 14%. respectively. accordingly, and based on what the proposed algorithm has achieved in terms of quantitative performance parameters, it could be adopted in SDN-based data centers.

LES of a turbulent swirl flame using a mesh adaptive level-set method with dynamic load balancing

A turbulent lean premixed swirl flame is numerically investigated by solving the Navier–Stokes equations with a finite-volume large-eddy simulation (LES) method. A combined G-equation progress variable approach is applied to model the combustion process using a solution adaptive level-set solver. The finite-volume and the level-set solver are parallelized and coupled on a joint hierarchical Cartesian mesh, where each solver can individually use and adapt a subset of the mesh cells. The level-set solver adapts the mesh locally in a band close to the flame front location to automatically satisfy the high accuracy requirements of the level-set solution. The numerical results for the turbulent swirl flame are in excellent agreement with experimental findings. Since the flame shape varies during the computation, the number of cells associated with the flame shifts between the individual subdomains which changes the workload distribution on the parallel computing processes. To achieve a high parallel efficiency, a dynamic load balancing method is applied which determines a new partitioning of the grid using estimated cell weights based on measured computing times to redistribute the cells among all subdomains accordingly. The efficiency of the dynamic load balancing scheme to reduce load imbalances is demonstrated for a large-scale simulation of the investigated swirl flame on 170,000 compute cores. That is, the dynamic load balancing scheme reduces the computing time of this simulation by approximately 30%.

A HDFS dynamic load balancing strategy using improved niche PSO algorithm in cloud storage

A Hadoop distributed file system (HDFS) NameNode dynamic load balancing strategy (NDLBT) using improved niche particle swarm optimisation (PSO) is proposed, which is used to solve the load balancing problem of urban surveillance video big data. Firstly, the relationship between bandwidth consumption and the bit rate of video files, the size of data blocks and access hotspots of online video files are analysed. Then, adaptive backup is realised by dynamic multiple copies of heterogeneous nodes. The dynamic distribution of metadata is implemented to ensure the performance of metadata server clusters. Finally, we propose a new improved niche PSO algorithm to achieve load balancing scheduling. In the experimental scenario where high bandwidth consumption video files are used as service access hotspots, the proposed method, which can reduce the bandwidth peak value of bottleneck nodes in data node clusters by 20%, is superior to the original load balancing method in 90% scenarios.

A HDFS dynamic load balancing strategy using improved niche PSO algorithm in cloud storage

A HDFS dynamic load balancing strategy using improved niche PSO algorithm in cloud storage

High-Performance, Graphics Processing Unit-Accelerated Fock Build Algorithm.

We present a high-performance, GPU (graphics processing unit)-accelerated algorithm for building the Fock matrix. The algorithm is designed for efficient calculations on large molecular systems and uses a novel dynamic load balancing scheme that maximizes the GPU throughput and avoids thread divergence that could occur due to integral screening. Additionally, the code adopts a novel ERI digestion algorithm that exploits all forms of permutational symmetry, combines efficiently the evaluation of both Coulomb and exchange terms together, and eliminates explicit thread synchronization requirements. Performance results obtained using a number of large molecules reveal remarkable speedups up to 24.4× with respect to the QUICK GPU code and up to 237× with respect to the GAMESS CPU parallel code.

The Design of Marine Remote Sensing Image Fusion Model Based on Dynamic Load Balancing Strategy

Feng, X.; Wan, M., and Li, S., 2020. The design of marine remote sensing image fusion model based on dynamic load balancing strategy. In: Yang, Y.; Mi, C.; Zhao, L., and Lam, S. (eds.), Global Topics and New Trends in Coastal Research: Port, Coastal and Ocean Engineering. Journal of Coastal Research, Special Issue No. 103, pp. 943–946. Coconut Creek (Florida), ISSN 0749-0208.In this work, a model of marine remote sensing image fusion based on dynamic load balancing is proposed. The remote sensing image acquisition system based on Android intelligent device obtains the marine remote sensing image, and enhances and registers the image based on the fuzzy theory and template matching, respectively, to realize the image preprocessing. Based on the preprocessing results, considering the small number of objects in the marine remote sensing image, the target area is divided by the idea of quad tree structure. At the same time, the task allocation strategy of dynamic load balancing and the idea of parallel computing are used to propose the task allocation algorithm processing model of the cluster system for the fusion processing of the image in the target area, so as to achieve the fusion of marine remote sensing image. It can be known that the proposed model takes less time and has superior average gradient and clarity of image.