Load Imbalance Research Articles

A wholistic optimization of containerized workflow scheduling and deployment in the cloud–edge environment

Digital Twin in Industry 4.0 utilizes Internet of Things (IoT) to collect real-life data and combine it with simulation models for product design and development. The simulation process can be executed as a workflow, consisting of tasks with precedence constraints. In a container-based workflow execution system, each task in the workflow is executed in a container within a virtual machine (VM). In this paper, a three-step scheduling model is proposed to combine scheduling of container-based workflows and the deployment of containers on a cloud–edge environment. In the first step, virtual CPU (vCPU) is allocated for each container to enable vCPU sharing among different containers. Next, two-step resource deployment is used to schedule the containers onto VM, and VM onto the physical machines in either edge or cloud environment. Multiple objectives are considered, including minimizing makespan, load imbalance, and energy consumption, from the perspective of cloud–edge resources as well as containerized workflows. To obtain a set of non-dominated solutions, three evolution strategies are designed and combined with two multi-objective algorithm frameworks — co-evolution strategy (CES), basic non-co-evolution strategy (B-NCS), and hybrid non-co-evolution strategy (H-NCS). Simulation results demonstrate that the proposed model outperforms the existing two-step scheduling model and H-NCS performs better than other strategies.

An effective multi-objective task scheduling and resource optimization in cloud environment using hybridized metaheuristic algorithm

In today’s world, cloud computing plays a significant role in the development of an effective computing paradigm that adds more benefits to the modern Internet of Things (IoT) frameworks. However, cloud resources are considered to be dynamic and the demands necessitated for resource allocation for a certain task are different. These diverse factors may cause load and power imbalance which also affect the resource utilization and task scheduling in the cloud-based IoT environment. Recently, a bio-inspired algorithm can work effectually to solve task scheduling problems in the cloud-based IoT system. Therefore, this work focuses on efficient task scheduling and resource allocation through a novel Hybrid Bio-Inspired algorithm with the hybridized of Improvised Particle Swarm Optimization and Ant Colony Optimization. The vital objective of hybridizing these two approaches is to determine the nearest multiple sources to attain discrete and continuous solutions. Here, the task has been allocated to the virtual machine through a particle swarm and continuous resource management can be carried out by an ant colony. The performance of the proposed approach has been evaluated using the CloudSim simulator. The simulation results manifest that the proposed Hybridized algorithm efficiently scheduling the task in the cloud-based IoT environment with a lesser average response time of 2.18 sec and average waiting time of 3.6 sec as compared with existing state-of-the-art algorithms.

River Routing in the INM RAS-MSU Land Surface Model: Numerical Scheme and Parallel Implementation on Hybrid Supercomputers

The land surface model (LSM) is a necessary compartment of any numerical weather forecast system or the Earth system model. This paper presents a new version of the INM RAS-MSU land surface model where the river hydrodynamic and thermodynamic scheme is embedded into the parallel execution framework using MPI and OpenMP. Numerical experiments have been performed for the East European domain with resolution 0.5°× 0.5°. The soil model parallel efficiency at 1–144 MPI cores was 0.52–0.79 and limited by the presence of ocean area, and by imbalance of computational load between soil columns. The acceleration of the river model at MPI level was defined by the size of the largest river basin in the domain. At the OpenMP level, the potential for acceleration of large river basin simulation is shown to be close to number of threads used, based on fractal properties of the river networks. This acceleration was hindered in our numerical experiments by the reduced river orders at the coarse land surface model resolution, so that the optimal speedup for the Volga river basin was 2.5–3 times attained at 4–6 threads. This performance is projected to improve with refinement of the LSM spatial resolution.

Providing renewable cooling in an office building with a Ground–Source heat pump system hybridized with natural ventilation & personal comfort systems

A ground–source heat pump system (GSHP) can be an alternative to air–source heat pumps (ASHP), as exchanging heat with the ground instead of with the air can result in higher coefficients of performance (COP). However, unbalanced loads can significantly shift the temperature profile in the ground, decreasing the GSHP’s overall performance. Hybrid systems can reduce the imbalanced loads and improve performance. This study proposed the use of personal comfort systems (PCS) and natural ventilation (NV) to decrease the thermal load imbalance in a cooling–dominated building.Building energy simulations were used to assess the performance of eight renewable heating and cooling solutions in three Portuguese cities (Porto, Lisbon and Faro): (1) ASHP, (2) GSHP, (3) ASHP–PCS, (4) GSHP–PCS, (5) ASHP–NV, (6) GSHP–NV, (7) ASHP–PCS–NV, (8) GSHP–PCS–NV. The simulation results show that without hybridization, the use of the GSHP resulted in an increase in ground temperature of more than 20 °C, over 50 years, which decreased its COP and rendered it unable to supply the full cooling load, and may not be acceptable in environmental terms. The use of either PCS or NV for hybrid cooling decreased the thermal cooling load and increased the energy efficiency of the GSHP system relative to the equivalent ASHP–hybrid. However, full coverage of the cooling load was still not always possible, as ground heat build–up was still high. Hybridization with both NV and PCS allowed full coverage of the thermal load by the GSHP. Both the cooling load and final energy needs decreased by over 90%, and the increase in ground temperature was limited to about 6 °C.

Efficient microscopy image analysis on CPU-GPU systems with cost-aware irregular data partitioning

The analysis of high resolution whole slide tissue images is a computationally expensive task, which adversely impacts effective use of pathology imaging data in research. We propose runtime solutions to enable efficient execution of pathology image analysis applications on modern distributed memory hybrid platforms equipped with both CPUs and GPUs. Hybrid systems offer significant computation capacity, but taking advantage of this computing power is complex. An application developer may have to implement multiple versions of data processing codes targeted for different computing devices. The developer also has to tackle the challenges of efficiently distributing computational load among the nodes of a distributed memory machine and among computing devices on a node. This is particularly difficult in analysis of high resolution images because of irregular computing costs of processing different image regions. In order to address these problems, we have leveraged a high-level image processing language (Halide) and integrated it into our runtime system called Region Templates (RT). The language simplifies the application development while generating code for multiple devices, such as CPU and GPU. The integration with RT allows for efficient multiple node hybrid execution. We also developed a novel cost-aware data partitioning (CADP) strategy that considers the workload irregularity to minimize load imbalance. Our experimental evaluation shows significant performance improvements on hybrid CPU-GPU machines, as compared with using a single processor (CPU or GPU), as well as on multi-GPU systems. CADP resulted in 1.7× better performance than other workload partitioning approaches (e.g., KD-Trees) on a hybrid machine and was up to 2.24× faster in multi-node settings.

GraphBLAST: A High-Performance Linear Algebra-based Graph Framework on the GPU

High-performance implementations of graph algorithms are challenging to implement on new parallel hardware such as GPUs because of three challenges: (1) the difficulty of coming up with graph building blocks, (2) load imbalance on parallel hardware, and (3) graph problems having low arithmetic intensity. To address some of these challenges, GraphBLAS is an innovative, on-going effort by the graph analytics community to propose building blocks based on sparse linear algebra, which allow graph algorithms to be expressed in a performant, succinct, composable, and portable manner. In this paper, we examine the performance challenges of a linear-algebra-based approach to building graph frameworks and describe new design principles for overcoming these bottlenecks. Among the new design principles is exploiting input sparsity , which allows users to write graph algorithms without specifying push and pull direction. Exploiting output sparsity allows users to tell the backend which values of the output in a single vectorized computation they do not want computed. Load-balancing is an important feature for balancing work amongst parallel workers. We describe the important load-balancing features for handling graphs with different characteristics. The design principles described in this paper have been implemented in “GraphBLAST”, the first high-performance linear algebra-based graph framework on NVIDIA GPUs that is open-source. The results show that on a single GPU, GraphBLAST has on average at least an order of magnitude speedup over previous GraphBLAS implementations SuiteSparse and GBTL, comparable performance to the fastest GPU hardwired primitives and shared-memory graph frameworks Ligra and Gunrock, and better performance than any other GPU graph framework, while offering a simpler and more concise programming model.

Comparative Analysis of Three Structures of Second-Order Generalized Integrator and Its Application to Phase-Locked Loop of Linear Kalman Filter

The present work explores the power quality problems of a microgrid in the aviation field, such as frequency offset or waveform distortion, caused by voltage imbalance and nonlinear load, to ensure the efficient and stable flight of aircraft, unmanned aerial vehicles, and other aircraft. There are many problems such as excessive harmonics, voltage imbalance, and direct current (DC) component in aviation variable frequency power supply voltage in microgrid under different conditions. Therefore, three kinds of second-order generalized integrators (SOGIs) with different structures are combined with linear Kalman filter phase-locked loops (PLLs). Besides, intelligent sensors are utilized for signal processing. Finally, simulation experiments are conducted to compare three SOGIs. The results show that the system is stable when the parameters k1 = 738.9553, k2 = 1092108.98405, k3 = 1477.91, and Ts = 0.000125 s and k ranges in [0.5, 3]. The angular frequency of PLL output is very low under the problems of many harmonics, three-phase voltage imbalance, and DC component ω g . The angular frequency ω g output by the PLL finally changes linearly with a change rate of 400 πrad/s2, so that the output phase angle θ g reaches a stable state. Thus, the proposed steady-state linear Kalman filter phase locking can accurately phase-lock the aviation variable frequency power supply. It provides an important reference for the power supply module in the microgrid to select the appropriate second-order generalized integrator to realize the accurate phase locking of the phase-locked loop under different conditions.

A Comprehensive Virtual Synchronous Generator Control Strategy for Harmonic and Imbalance Voltage Suppression of Multi-Inverter Parallel Microgrid

To reduce the impact of the imbalance of mixed non-linear loads on an inverter voltage output in the microgrid, we improve the disadvantage of the lack of damping and inertia for traditional droop control. This paper proposes a comprehensive virtual synchronous generator (VSG) control strategy for harmonic suppression and imbalance suppression of a multi-inverter parallel microgrid. On one hand, an improved VSG control strategy is proposed to increase the damping and inertia of distributed generations (DGs) in the microgrid, and secondary control is introduced to improve system stability. On the other hand, the frequency division suppression control strategy is used to eliminate the influence of harmonics, and the negative sequence component is compensated to eliminate the influence of imbalance. Then small-signal analysis is used for analysis of the stability of the strategy. Finally, we verify the comprehensive control strategy proposed in this paper through experiments. The experimental results suggest a significant improvement on the voltage, frequency, power optimization, handling of non-linear load and capacity distribution precision, as well as providing inertia support for the system.

Electric Energy Management for Plug-in Electric Vehicles Charging in the Distribution System by a dual cascade scheduling algorithm

This paper presents an algorithm for plug-in electric vehicles (PEVs) charging in the three-phase distribution system for residential houses. It aims to prevent violent voltage level deviation and increasing losses on the three-phase distribution system due to uncontrolled charging and allocate power to each plug-in electric vehicle. The algorithm is comprised of two processes. The first process is power limitation and limited power of load imbalance by if-else rules, while the second process is power allocation to each PEV by the dual cascade scheduling algorithm which is the integration of tasking scheduling algorithms. A 100 kVA distribution transformer and 30 houses are defined in the simulation situation. Also, the available PEVs in single-phase, two-phase, and three-phase systems are assigned for verification of the proposed algorithm. Root-mean-square deviation (RMSD) referred to the satisfaction of PEV owners, total PEVs charged energy, and the average percentage of achieved charging time, as the result indicators. The results show the proposed algorithm can provide good results without rejected PEVs charging. Furthermore, this paper also displays the analysis of voltage level, percentage of voltage unbalance factor, and loss in the distribution system. In the future, coordination with home appliances to gain a high load margin or electric energy cost control will be improved in the proposed algorithm

PostMan: Rapidly Mitigating Bursty Traffic via On-Demand Offloading of Packet Processing

Unexpected bursty traffic brought by certain sudden events, such as news in the spotlight on a social network or discounted items on sale, can cause severe load imbalance in backend services. Migrating hot data - the standard approach to achieve load balance - meets a challenge when handling such unexpected load imbalance, because migrating data will slow down the server that is already under heavy pressure. This article proposes PostMan, an alternative approach to rapidly mitigate load imbalance for services processing small requests. Motivated by the observation that processing large packets incurs far less CPU overhead than processing small ones, PostMan deploys a number of middleboxes called helpers to assemble small packets into large ones for the heavily-loaded server. This approach essentially offloads the overhead of packet processing from the heavily-loaded server to helpers. To minimize the overhead, PostMan activates helpers on demand, only when bursty traffic is detected. The heavily-loaded server determines when clients connect/disconnect to/from helpers based on the real-time load statistics. To tolerate helper failures, PostMan can migrate connections across helpers and can ensure packet ordering despite such migration. Driven by real-world workloads, our evaluation shows that, with the help of PostMan, a Memcached server can mitigate bursty traffic within hundreds of milliseconds, while migrating data takes tens of seconds and increases the latency during migration.

Software-Defined Internet of Multimedia Things: Energy-Efficient and Load-Balanced Resource Management

Internet of Multimedia Things (IoMT) is becoming attractive day by day and provides more services to the Internet users. The ever-increasing multimedia applications and services led to an outburst in IoMT. The multimedia things connected to IoMT are also increasing; millions of devices and high volume of traffic. This large traffic is directed toward the servers of the service provider in the IoMT cloud through the network switches. As a result, the IoMT infrastructure is facing the crisis of the resource management of switches and servers from two aspects: 1) load imbalance and 2) energy loss. This article proves that the problem of optimal resource management of IoMT networks with the energy and load constraints simultaneously is an NP-hard problem that has a high time complexity. The problem has decomposed. Then, we propose a modular system of energy and load control in IoMT using the concepts of network softwarization and virtual resources. The proposed controller first dynamically adjusts the resources through accurate determination of the IoMT network size. It then distributes the load between the IoMT servers as well as routes the traffic between switches to the desired server. The Open vSwitch, Floodlight Controller, and Kaa Servers are, respectively, used for implementing the switches, controller, and servers of IoMT in the test platform. The results show that the proposed system both minimizes the number of IoMT active servers and switches and distributes the load between them. As a result, the parameters for evaluating the quality of service and quality of experience, such as throughput, multimedia delay, R factor, and mean opinion score improved.

Non-singular Fast Terminal Sliding Mode Control of Wind Turbine Based on Combined Feedback of Power and Load

A nonsingular fast terminal sliding mode control of wind turbine based on combined power and load feedback is proposed to deal with the problem of power instability and load imbalance of wind turbine when the wind turbine operates above the rated wind speed. The control strategy uses three independent nonsingular fast terminal sliding mode controllers to reduce the Tilt moment, Yaw moment and stable power on the hub fixed coordinate system respectively. Finally, the 2MW wind turbine is simulated by MATLAB and Bladed software, and the control strategies of three PI controllers are compared. The results show that the control strategy not only has a good effect on stabilizing the output power, but also has a prominent performance in reducing the unbalanced load of the wind turbine.

Hybrid parallelization of molecular dynamics simulations to reduce load imbalance

The most widely used technique to allow for parallel simulations in molecular dynamics is spatial domain decomposition, where the physical geometry is divided into boxes, one per processor. This technique can inherently produce computational load imbalance when either the spatial distribution of particles or the computational cost per particle is not uniform. This paper shows the benefits of using a hybrid MPI+OpenMP model to deal with this load imbalance. We consider LAMMPS (Large-scale Atomic/Molecular Massively Parallel Simulator), a prototypical molecular dynamics simulator that provides its own balancing mechanism and an OpenMP implementation for many of its modules, allowing for a hybrid setup. In this work, we extend the current OpenMP implementation of LAMMPS and optimize it and evaluate three different setups: MPI-only, MPI with the LAMMPS balance mechanism, and hybrid setup using our improved OpenMP version. This comparison is made using the five standard benchmarks included in the LAMMPS distribution plus two additional test cases. Results show that the hybrid approach can deal with load balancing problems better and more effectively (50% improvement versus MPI-only for a highly imbalanced test case) than the LAMMPS balance mechanism (only 43% improvement) and improve simulations with issues other than load imbalance.

Joint User-Centric Clustering and Multi-cell Radio Resource Management in Coordinated Multipoint Joint Transmission

Coordinated multipoint joint transmission (JT-CoMP) is a promising solution to address inter-cell interference in dense future wireless networks due its strength in converting interfering signals into useful signals, thereby enhancing capacity especially at the cell edge. However, allowing all user equipments (UEs) to operate using the JT-CoMP mode reduces the availability of radio resources. This paper develops an efficient algorithm that can identify which UEs will benefit from operating in a JT-CoMP mode and how to efficiently allocate radio resources from multiple base stations. Joint user-centric JT-CoMP clustering and multi-cell resource management is used in two steps where user-centric clusters are constructed as a first step and according to the clustering results obtained, resources are assigned. This paper also provides a new user-centric clustering approach that allows a user to utilize the JT-CoMP technique only if JT-CoMP boosts its rate above a certain threshold level. A multi-cell resource allocation scheme that can address the resource mismatching problem between cooperative BSs that happens due to load imbalance is proposed. Simulation results show that the proposed user-centric clustering algorithm outperforms the traditional power level difference scheme in terms of the system’s overall throughput as well as the throughput of cell-edge users. Also, results show that the performance of JT-CoMP is mainly affected by the user-centric approach and the amount of physical radio resources assigned to CoMP UEs.

Analytic performance model for parallel overlapping memory‐bound kernels

AbstractComplex applications running on multicore processors show a rich performance phenomenology. The growing number of cores per ccNUMA domain complicates performance analysis of memory‐bound code since system noise, load imbalance, or task‐based programming models can lead to thread desynchronization. Hence, the simplifying assumption that all cores execute the same loop can not be upheld. Motivated by observations on plain and modified versions of the HPCG benchmark, we construct a performance model of execution of memory‐bound loop kernels. It can predict the memory bandwidth share per kernel on a memory contention domain depending on the number of active cores and which other workload the kernel is paired with. The only code features required are the single‐thread memory request fraction per kernel, which is directly related to the single‐thread memory bandwidth, and its saturated bandwidth. The former can either be measured directly or predicted using the Execution‐Cache‐Memory performance model. The computational intensity of the kernels and the detailed structure of the code is of no significance. We validate our model on Intel Broadwell, Intel Cascade Lake, and AMD Rome processors pairing various streaming and stencil kernels. The error in predicting the bandwidth share per kernel is less than 8%.

Fast and efficient parallel breadth-first search with power-law graph transformation

In the big data era, graph computing is widely used to exploit the hidden value in real-world graphs in various scenarios such as social networks, knowledge graphs, web searching, and recommendation systems. However, the random memory accesses result in inefficient use of cache and the irregular degree distribution leads to substantial load imbalance. Breadth-First Search (BFS) is frequently utilized as a kernel for many important and complex graph algorithms. In this paper, we describe a preprocessing approach using Reverse Cuthill-Mckee (RCM) algorithm to improve data locality and demonstrate how to achieve an efficient load balancing for BFS. Computations on RCM-reordered graph data are also accelerated with SIMD executions. We evaluate the performance of the graph preprocessing approach on Kronecker graphs of the Graph500 benchmark and real-world graphs. Our BFS implementation on RCM-reordered graph data achieves 326.48 MTEPS/W (mega TEPS per watt) on an ARMv8 system, ranking 2nd on the Green Graph500 list in June 2020 (the 1st rank uses GPU acceleration).

Tree-structured data placement scheme with cluster-aided top-down transmission in erasure-coded distributed storage systems

In erasure-coded distributed storage systems, the rapid completion of data placement process is very critical to maintain system performance, where the process is defined as to insert coded blocks into a set of redundant storage nodes. However, the existing data placement schemes have some shortcomings, such as load imbalance and long transmission latency. To solve these problems, this paper proposes a tree-structured data placement scheme with cluster-aided top-down transmission, which improves the practicality and the efficiency of the data insertion. Specifically, the storage nodes are clustered by a location-based K-mean method at first, and then selecting a master node as the source node to encode file and distribute coded blocks in each cluster by designing a master node’s selection algorithm based on evolutionary game theory. Since multiple master nodes are used to receive data files from data core, the load imbalance problem caused by single node receiving data files is mitigated. In addition, to improve the data transmission efficiency, we construct a tree-structured network to organize the connections of the storage nodes in cluster, where the master node is the root node. Meanwhile, a data placement scheme with top-down transmission by combining an improved metaheuristic algorithm is proposed to manage the pipelined data transmission along the tree-structured network. Experimental results show that the proposed scheme can effectively improve the survive life of storage network and reduce the data insertion time. Furthermore, the proposed scheme has better performance than the optimal tree-structured scheme based on other optimization algorithms.

RAFL: A hybrid metaheuristic based resource allocation framework for load balancing in cloud computing environment

This paper proposes a hybrid metaheuristic based resource allocation framework named RAFL for load balancing in cloud computing environment. The objective is to proactively minimize the load imbalance across active physical machines and among their considered resource capacities (e.g., CPU and RAM). This prevents overloading/underloading of active physical machines and utilizes their considered resource capacities in a balanced manner. In the proposed framework, a phasor particle swarm optimization and dragonfly algorithm based hybrid optimization algorithm named PPSO-DA is used to generate an optimal resource allocation plan for balancing the load. Simulation experiments are performed using CloudSim simulator to measure the metrics of load imbalance across active physical machines and among their considered resource capacities. Results show that the proposed PPSO-DA algorithm outperforms phasor particle swarm optimization, dragonfly algorithm, comprehensive learning particle swarm optimization, memory based hybrid dragonfly algorithm, sine cosine algorithm, and elephant herding optimization, in finding an optimal resource allocation for balancing the load. The statistical analysis and benchmark testing also validates the relative superiority of PPSO-DA.

Improving system performance in Homogeneous Multicore Systems

Allocation of parallel load in multicore systems has become a challenging task for high performance computing system. There are several parameters to evaluate the performance of a scheduling algorithm such as task imbalance and execution time. This paper proposes a task scheduling approach that targets multiple cores connected through appropriate interconnection network. The proposed approach utilizes the computing resources effectively by assigning the tasks dynamically among different cores of the system in a realistic time. Each task has its own timeline and multiple sequence of tasks are mapped on different cores of the system. In particular, performance is evaluated on n x n Mesh, DMesh, ZMesh and Torus networks. The load imbalance and execution times are considered as metrics to evaluate the performance of the proposed algorithm. Simulation results are obtained and compared with well-known minimum distance scheduling algorithm which shows reduction in execution time while maintaining the load imbalance. An improvement of 20-30% is obtained in load imbalance for considered multicore systems with improved execution time. The simulation study reveals that the proposed algorithm is best suited to take architectural benefits for mesh-based multicore systems.

Система управления выравниванием нагрузки в многодвигательном электроприводе скребкового конвейера

Скребковый конвейер относится к основным средствам доставки угля от очистного комбайна до транспортных средств конвейерного штрека при подземной разработке угольных месторождений. В мощных скребковых конвейерах, в условиях ограничения оборудования по габаритам, используется многодвигательный электропривод, включающий от двух до четырех электродвигателей. Дисбаланс нагрузки между двигателями в таком электроприводе является основной проблемой. С целью решения этой задачи проанализированы существующие способы выравнивания нагрузки в многодвигательном электроприводе и предложена стратегия частотно-регулируемого привода с векторным управлением нескольких двигателей, основанная на принципе «ведущий-ведомый». При этом главный двигатель управляется с помощью регулятора скорости, а ведомый - с использованием стратегии заданного крутящего момента. Данная схема управления реализована в среде MatLab/Simulink. Результаты моделирования показали, что режим управления по принципу «ведущий-ведомый» может решить проблему выравнивания нагрузки скребкового конвейера во всех режимах работы, тем самым увеличив срок его службы, снизив износ оборудования, а значит, повысив эффективность производства шахты.