Dynamic Load Balancing Model Research Articles

Proactive and dynamic load balancing model for workload spike detection in cloud

Dynamic Load Balancing (DLB) has been an ever-green research topic since the beginning of cloud computing. DLB model distributes the available cloud resources (i.e., CPU, RAM, Storage, BW, etc.) among the VMs using a systematic approach to ensure efficient resource utilization. Overloading, an abnormal condition, occurs when the host encounters a resource shortage due to the high workload of Federated Learning. VM migration and consolidation are reliable load-balancing techniques to fix dynamic overloading issues. As part of the migration, the oversubscribed host VMs are moved to the other hosts till the load is regulated. Besides the high workload, the sudden spike in workload also leads the hosts to instant overloading. The former DLB models are efficient in handling general overloading conditions. Still, they are now suffering from limitations in taking the spike overloading due to 1) a sudden spike in workloads, 2) instant overloading, 3) recurrent migrations, and 4) SLA violations. To address these limitations in handling spike overloading, in this paper, we proposed a proactive and dynamic load balancing model (DLBM), which is designed to handle spike overloading with optimal solutions. The offered dynamic load balancing model with a spike detection algorithm helps detect workload spikes in advance to prevent the hosts from overloading and SLA violations. Experimental conducted using the Cloudsim tool and the planet lab workload dataset prove that the proposed DLBM identified the spikes in workload and efficiently controlled the recurrent migrations.

A dynamic load balancing model coupled with DAC and ISAT for a stochastic turbulent combustion model

Due to the composition-dependent stiffness of chemistry, simulations of reactive turbulent flows may present computational load imbalance among parallel processes when spatial decomposition is used for parallelisation, causing high CPU idle time and waste of computational resources. To increase computational efficiency, a dynamic load balancing (DLB) model is proposed to redistribute computational load among computing cores. The DLB model exploits a decomposition in the mixture fraction space with two dynamic adjusting decomposition strategies to realise load redistribution. The DLB model is suitable for the integration of chemistry on stochastic particles in hybrid Eulerian/Lagrangian turbulent combustion models in which the Eulerian field is conventionally decomposed statically in physical space in a way that balances the computational load for the solution of the Navier-Stokes equation but which does not generally lead to balanced load for the computation of the composition fields. Here it is tested using an OpenFOAM-based platform, mmcFoam, which is a comprehensive object-orientated C++ library for stochastic turbulent combustion modelling. Apart from direct integration (DI) for chemistry, the DLB model is also coupled with dynamic adaptive chemistry (DAC) and in situ adaptive tabulation (ISAT), which allows for extra speedup. The performance of the coupled models is validated and assessed for two laboratory flame conditions that exhibit different levels of computational load imbalance. Overall, the DLB model effectively balances the computational load distribution and increases the effective usage of computing power, shortening the simulation wall time required. Moreover, a strong scaling test is carried out using up to 512 cores. Although all approaches have sub-ideal scalability, the scalability of each with DLB is significantly better than without DLB. While DLB-ISAT has relatively poor scalability compared to the DI- and DAC-based methods, DLB-ISAT still ranks the fastest among the algorithms in all scaling trials.

Fast reactive flow simulations using analytical Jacobian and dynamic load balancing in OpenFOAM

Detailed chemistry-based computational fluid dynamics (CFD) simulations are computationally expensive due to the solution of the underlying chemical kinetics system of ordinary differential equations (ODEs). Here, we introduce a novel open-source library aiming at speeding up such reactive flow simulations using OpenFOAM, an open-source software for CFD. First, our dynamic load balancing model by Tekgül et al. [“DLBFoam: An open-source dynamic load balancing model for fast reacting flow simulations in OpenFOAM,” Comput. Phys. Commun. 267, 108073 (2021)] is utilized to mitigate the computational imbalance due to chemistry solution in multiprocessor reactive flow simulations. Then, the individual (cell-based) chemistry solutions are optimized by implementing an analytical Jacobian formulation using the open-source library pyJac, and by increasing the efficiency of the ODE solvers by utilizing the standard linear algebra package. We demonstrate the speed-up capabilities of this new library on various combustion problems. These test problems include a two-dimensional (2D) turbulent reacting shear layer and three-dimensional (3D) stratified combustion to highlight the favorable scaling aspects of the library on ignition and flame front initiation setups for dual-fuel combustion. Furthermore, two fundamental 3D demonstrations are provided on non-premixed and partially premixed flames, viz., the Engine Combustion Network Spray A and the Sandia flame D experimental configurations, which were previously considered unfeasible using OpenFOAM. The novel model offers up to two orders of magnitude speed-up for most of the investigated cases. The openly shared code along with the test case setups represent a radically new enabler for reactive flow simulations in the OpenFOAM framework.

DLBFoam: An open-source dynamic load balancing model for fast reacting flow simulations in OpenFOAM

Computational load imbalance is a well-known performance issue in multiprocessor reacting flow simulations utilizing directly integrated chemical kinetics. We introduce an open-source dynamic load balancing model named DLBFoam to address this issue within OpenFOAM, an open-source C++ library for Computational Fluid Dynamics (CFD). Due to the commonly applied operator splitting practice in reactive flow solvers, chemistry can be treated as an independent stiff ordinary differential equation (ODE) system within each computational cell. As a result of the highly non-linear characteristics of chemical kinetics, a large variation in the convergence rates of the ODE integrator may occur, leading to a high load imbalance across multiprocessor configurations. However, the independent nature of chemistry ODE systems leads to a problem that can be parallelized easily (called an embarrassingly parallel problem in the literature) during the flow solution. The presented model takes advantage of this feature and balances the chemistry load across available resources. Additionally, a reference mapping model is utilized to further speed-up the simulations. When DLBFoam it utilized with both these features enabled, a speed-up by a factor of 10 is reported for reactive flow benchmark cases. To the best of our knowledge, this model is the first open-source implementation of chemistry load balancing in the literature. Program summaryProgram Title: DLBFoamCPC Library link to program files:https://doi.org/10.17632/bb9zjfzcmm.1Developer's repository link:https://github.com/blttkgl/DLBFoamLicensing provisions: GPLv3Programming language: C++Nature of problem: Solution of chemical kinetics in parallel reacting flow solvers raises a computational imbalance across multiprocessor architectures. DLBFoam balances the load distribution evenly, providing significant speed-up in reacting CFD applications.Solution method: The dynamic load balancing is implemented by distributing the point-wise chemistry problems from most loaded processes to less loaded ones using MPI communication protocol.Additional comments including restrictions and unusual features: The present model is designed to work with the standard chemistry model class available in OpenFOAM (versions 7 and 8). For the time being, the model does not support derived combustion models such as “TDAC” and covers gas-phase reaction kinetics only. In addition, the boundary surface chemistry problems are neglected by the model.

Capacity based deadline aware dynamic load balancing (CPDALB) model in cloud computing environment

ABSTRACT Cloud computing is a promising infrastructure in which many activities are carried out in both the commercial and academic spheres. Non-uniform utilization of virtual machines (VMs) in cloud creates an imbalance of the load, resulting in the violation of Service Level Agreements (SLAs) and tends to decreases the rate of revenue. There is a huge concern in scaling the load balancing algorithm due to the heterogeneous nature of the cloud. In this paper, a Capacity based Deadline Aware Dynamic Load Balancing (CPDALB) algorithm is proposed to address the problem. CPDALB is a novel heuristic based load balancing approach which focuses on the selection of VM for allocation of tasks to ensure customer satisfaction in terms of meeting deadline constraints and cost of running applications. CPDALB uses the concept of deploying requests to VMs based on their processing capacities to minimize load imbalance with satisfying the deadline. Other QoS parameters viz. turnaround time, response time and utilization are also computed for performance evaluation. An extensive simulation study has been carried out for the comparison of CPDALB with other standard scheduling algorithms given so far in the literature. The experimental results reveal that the proposed algorithm outperforms its competitors on account of almost all parameters under study.

Fault tolerance based load balancing approach for web resources

ABSTRACTGrid computing is a prominent tool for assembling and incorporating groups of heterogeneous resources scattered around the world, connected through a network. Since load balancing is a challenging issue in a distributed environment, Grid computing provides an architecture, which is well suited, low-cost, and consistent. For task execution on suitable resources, grid computing provides efficient load balancing and fault tolerance approaches. But, due to the dynamic nature of grid resources, sometimes tasks can’t be completed within given constraints (deadline, cost). To solve these issues, this paper proposes a fault tolerance–based load balancing approach by considering the dynamic nature of resources. First, significant contributions in the field of load balancing are analyzed based on several parameters. Second, a fault tolerance dynamic load balancing model is proposed for task execution based on resource load and fault index value. For fault tolerance, checkpoints are set at various determined intervals to resume tasks at the next possible instance that avoids unnecessary placement of checkpoints. Third, the proposed model is validated and provides improved performance in terms of response time, makespan, and throughput. The model provides up to 12% and 9% performance improvement depending on method of measurement in comparison to other state-of-the-art methods.

A Fuzzy Neural Network Based Dynamic Data Allocation Model on Heterogeneous Multi-GPUs for Large-scale Computations

The parallel computation capabilities of modern GPU (Graphics Processing Unit) processors have attracted increasing attention from researchers and engineers who have been conducting high computational throughput studies. However, current single GPU based engineering solutions are often struggle to fulfill their real-time requirements. Thus, the multi-GPU-based approach has become a popular and cost-effective choice for tackling the demands. In those cases, the computational load balancing over multiple GPU “nodes” is often the key and bottleneck that affect the quality and performance of the runtime system. The existing load balancing approaches are mainly based on the assumption that all GPU nodes in the same computer framework are of equal computational performance, which are often not the case due to cluster design and other legacy issues. This paper presents a novel dynamic load balancing (DLB) model for rapid data division and allocation on heterogeneous GPU nodes based on an innovative fuzzy neural network (FNN). In this research, a 5-state parameter feedback mechanism defining the overall cluster and node performances is proposed. The corresponding FNN-based DLB model will be capable of monitoring and predicting individual node performance under different workload scenarios. A real-time adaptive scheduler has been devised to reorganize the data inputs to each node when necessary to maintain their runtime computational performances. The devised model has been implemented on two dimensional (2D) discrete wavelet transform (DWT) tasks for evaluation. Experiment results show that this DLB model has enabled a high computational throughput while ensuring real-time and precision requirements from complex computational tasks.

LB-networks: A model for dynamic load balancing in queueing networks

Dynamic load balancing plays an important role in the management of data centres where a fair workload distribution may be hard to achieve. Dynamic load balancing algorithms are usually clustered into two families: receiver-initiated and sender-initiated. In the former approach, an idle node transfers some of the jobs of a highly loaded node to its queue, whereas in the latter it is the overloaded node that decides to send somewhere else some of the jobs that have been assigned to itself. The literature has proved the receiver-initiated strategies to be more effective than the sender-initiated ones under heavy load. In this paper, we give for the first time a closed form expression of the stationary distribution of a class of queueing networks where an arbitrary number of nodes implement a receiver-initiated dynamic load balancing algorithm. The result can be applied to Markovian queueing networks with probabilistic routing and exponentially distributed service times.

An Improved Dynamic Load-Balancing Model

With the rapid development of big data and the national big data strategy is put forward, the Web server cluster facing more complex and severe challenges. The traditional load balancing algorithm has obvious limitations. This paper proposes a dynamic load-balancing model based on the SSAWF (Strong Suspend And Weak Forecast) mechanism. This model uses strong suspend mechanism and cubic exponential smoothing prediction method based on AHP algorithm for dynamic load balancing scheduling. Results of the experiments show that the improved model has more positive influence on read/ write performance of the cluster under abnormal system transient performance, high concurrency and high system load interaction, that's to say the load balancing effect is better than the traditional load balancing.

Agent-based Dynamic Adaptive Cluster Load Balancing Model

In a cluster environment, the identification and the classification of the requested task is a prerequisite for choosing the appropriate load balancing algorithm and improving the processing capacity of the network. As the existing load balancing strategies cannot relate tasks and load balancing algorithm intelligently, this paper present an Agent-based dynamic adaptive load balancing model. Through the identification and classification of the requested task, the model selects corresponding load balancing algorithm to reach the load balancing in cluster. This model has a distinct advantage over other load balancing schedule strategies on response time. DOI : http://dx.doi.org/10.11591/telkomnika.v12i3.4568 Full Text: PDF

A Soft Computing Approach to Dynamic Load Balancing in 3GPP LTE

of high-speed data services. These services must be guaranteed under varying radio propagation conditions, to stochastically distributed mobile users. A necessity for determining and regulating the traffic load of eNodeBs naturally ensues. Load balancing is a self-optimization operation of self-organizing networks (SON). It aims at ensuring an equitable distribution of users in the network. This translates into better user satisfaction and a more efficient use of network resources. Several methods for load balancing have been proposed. Most of the algorithms are based on hard (traditional) computing which does not utilize the tolerance for precision of load balancing. This paper proposes the use of soft computing, precisely adaptive Neuro-fuzzy inference system (ANFIS) model for dynamic QoS aware load balancing in 3GPP LTE. The use of ANFIS offers learning capability of neural network and knowledge representation of fuzzy logic for a load balancing solution that is cost effective and closer to human intuition

Agent scheduling model for adaptive dynamic load balancing in agent-based distributed simulations

Agent-based distributed simulations are confronted with load imbalance problem, which significantly affects simulation performance. Dynamic load balancing can be effective in decreasing simulation execution time and improving simulation performance. The characteristics of multi-agent systems and time synchronization mechanisms make the traditional dynamic load balancing approaches not suitable for dynamic load balancing in agent-based distributed simulations. In this paper, an adaptive dynamic load balancing model in agent-based distributed simulations is proposed. Due to the complexity and huge time consuming for solving the model, a distributed approximate optimized scheduling algorithm with partial information (DAOSAPI) is proposed. It integrates the distributed mode, approximate optimization and agent set scheduling approach. Finally, experiments are conducted to verify the efficiency of the proposed algorithm and the simulation performance under dynamic agent scheduling. The experiments indicate that DAOSPI has the advantage of short execution time in large-scale agent scheduling, and the distributed simulation performance under this dynamic agent scheduling outperforms that under static random agent distribution.

On a Virtual Shared Memory Cluster System with Virtual Machines

In this paper, an architecture with a load balancing model and a fault tolerant model for virtual shared memory clusters is proposed. The centralized dynamic load balancing model uses manager worker concept with a virtualized compute server. The virtual server can be expanded on the fly for load balancing by adding virtual machines temporarily. The fault tolerant model controlled by a virtual machine monitor describes the checkpointing based recovery methods. The performance evaluation results show that the proposed system achieves significant speedup in terms of execution time and checkpoint time .

An Improved Dynamic Load Balancing Algorithm for a Distributed System in LAN

A dynamic time delay model for load balancing was proposed in distributed heterogeneous system by Hayat. But how to compute the load-transfer delay and the gain coefficient were not gived, so it is imposible to quantitatively analyse the performance of a distributed system. In this paper, an improved dynamic load balancing model with load-transfer delay is designed for LAN, and the gain coefficient is defined by a number of simulation test, so to meet requirements of LAN. At last, the simulation test is presented. The conclusion is that the improved algorithm not only saves the resources, but also realizes a reliable and efficient operation of the system. So the improved algorithm can be used to the distributed system in LAN.

A DYNAMIC LOAD BALANCING MODEL FOR CONCURRENTLY CONNECTED USERS IN U-HEALTHCARE MONITORING SYSTEMS

U-healthcare systems are based on a ubiquitous and wireless computing and communication environment. They are comprised of the U-healthcare management center, electronic medical records (EMR) system, and the associated services for users and patients. The U-healthcare management center performs continuous monitoring and provides support services in multiple areas, requiring careful allocation of the limited service resources to provide customized healthcare services for users of the mobile distributed system. When the number of locally connected users increases rapidly, a mobile allocation and distribution server can be imbalanced by the load on service resources, resulting in delayed services. This study proposes a dynamic load balancing model for reducing the load of users on service resources and supporting efficient response services in a mobile distributed system. The proposed dynamic load balancing model clusters the system resources of servers dynamically, according to each users' movement and time. The dynamic clustering of system resources uses wFCM (weighted Fuzzy C-Means), which changes the cluster center by transforming existing FCM (Fuzzy C-Means) from a fixed weight to a dynamic one. Using wFCM, the load balance can be maintained, based on the usage rate of service resources, such as CPU, memory, and network. In addition, the balance between QoS (Quality of Service) requests and network response times can be maintained by adding an abstraction layer between application services and network infrastructure. Therefore, when the proposed model is applied to a U-healthcare monitoring system, the system can perform near real-time monitoring of service users in the mobile distributed environment, and effectively address emergent situations. This study evaluates the response time of the implemented model in relation to the number of concurrently connected users, and confirms that the proposed model is faster in response and service processing than existing WLC (Weighted Least-Connection Scheduling) and FCM (Fuzzy C-Means).

A repartitioning hypergraph model for dynamic load balancing

In parallel adaptive applications, the computational structure of the applications changes over time, leading to load imbalances even though the initial load distributions were balanced. To restore balance and to keep communication volume low in further iterations of the applications, dynamic load balancing (repartitioning) of the changed computational structure is required. Repartitioning differs from static load balancing (partitioning) due to the additional requirement of minimizing migration cost to move data from an existing partition to a new partition. In this paper, we present a novel repartitioning hypergraph model for dynamic load balancing that accounts for both communication volume in the application and migration cost to move data, in order to minimize the overall cost. The use of a hypergraph-based model allows us to accurately model communication costs rather than approximate them with graph-based models. We show that the new model can be realized using hypergraph partitioning with fixed vertices and describe our parallel multilevel implementation within the Zoltan load balancing toolkit. To the best of our knowledge, this is the first implementation for dynamic load balancing based on hypergraph partitioning. To demonstrate the effectiveness of our approach, we conducted experiments on a Linux cluster with 1024 processors. The results show that, in terms of reducing total cost, our new model compares favorably to the graph-based dynamic load balancing approaches, and multilevel approaches improve the repartitioning quality significantly.

A Dynamic Load Balancing Model for a Distributed System

A communication protocol designed for fault tolerance in distributed real time systems is implemented and a dynamic load balancing model is designed and implemented over this protocol. The protocol consists of cluster based, hierarchical rings which use synchronous communication. The rings are synchronous. At the lowest level in the hierarchy, there are clusters that consist of computing processors, called nodes. The higher level consists of the cluster representatives that manage the clusters of the lower level. There can be two or more levels in the hierarchy. Ring protocols in each cluster can work in parallel. Also, a fault tolerance mechanism is integrated to the protocol. The dynamic distributed load balancing module designed over the protocol aims to maximize the overall performance of the whole system by distributing the load submitted to the system efficiently and transparently among the nodes. While performing operations to achieve this goal, the module also considers the real-time constraints of the system

A general matrix iterative model for dynamic load balancing

Many applications require some form of load balancing in order to obtain good performance on parallel computing systems. A number of load balancing algorithms have been proposed, and implemented on real systems. However, application developers are in need of a performance tool that compares and evaluates various balancing algorithms, and helps select the best algorithm for a given application and computing platform. In this paper, we propose a matrix iterative model that can represent a wide range of dynamic load balancing algorithms. The model captures the change in task distribution over the processors due to both the application imbalance and load balancing algorithm. We also define performance measures that capture the costs of executing the parallel application and load balancing algorithm on a given computing platform. The model and the performance measures are used to compare and evaluate various load balancing algorithms. In this paper, the model is parameterized to represent three load balancing algorithms — the random, diffusion and redistribution algorithms. Model predictions and experimental measurements are compared for the parallel N-body simulation application. This application is representative of a wide class of scientific applications where time-varying workload distributions are present. The model predictions are within 20% of measured values. The performance of the three algorithms are compared, and optimal algorithm parameters are derived showing how the model can be used to both select an algorithm and set its parameters.