Heterogeneous Computing System Research Articles

Control model of the phased upgrade of a heterogeneous computing system

Upgrade options for a high-performance specialized system designed to solve resourceintensive problems, such as exhaustive searches, are considered. The concept of the set of executable computational jobs of various types is introduced in order to describe the system's functional capabilities. User requirements for the upgraded system are described by the vector whose components determine the number of assigned jobs that can be executed together in one operating cycle. The method of multiparameter sliding scheduling is used to analyze the system development options. The solution of a sequence of optimization problems makes it possible to obtain system configuration options with the minimal cost, minimal energy consumption, and maximal number of executable jobs. The selection procedure of the compromise project that best takes into account various conflicting requirements is proposed.

On the role of application and resource characterizations in heterogeneous distributed computing systems

Loosely coupled applications composed of a potentially very large number (from tens of thousands to even billions) of tasks are commonly used in high-throughput computing and many-task computing paradigms. To efficiently execute large-scale computations which can exceed the capability in a single type of computing resources within expected time, we should be able to effectively integrate resources from heterogeneous distributed computing (HDC) systems such as clusters, grids, and clouds. In this paper, we quantitatively analyze the performance of three different real scientific applications consisting of many tasks on top of HDC systems based on a partnership of distributed computing clusters, grids, and clouds to understand the application and resource characteristics, and show practical issues that normal scientific users can face during the course of leveraging these systems. Our experimental study shows that the performance of a loosely coupled application can be significantly affected by the characteristics of a HDC system, along with hardware specification of a node, and their impacts on the performance can vary widely depending on the resource usage pattern of each application. We then devise a preference-based scheduling algorithm that can reflect characteristics and resource usage patterns of various loosely coupled applications running on top of HDC systems from our experimental study. Our preference-based scheduling algorithm can allocate the resources from different HDC systems to loosely coupled applications based on the preferences of the applications for the HDC systems. We evaluate the overall system performance over various preference types, using trace-based simulations, which can be determined based on different factors such as CPU specifications and application throughputs. Our simulation results demonstrate the importance of understanding the application and resource characteristics on effective scheduling of loosely coupled applications on the HDC systems.

Resource Allocation Policies for Loosely Coupled Applications in Heterogeneous Computing Systems

High-Throughput Computing (HTC) and Many-Task Computing (MTC) paradigms employ loosely coupled applications which consist of a large number, from tens of thousands to even billions, of independent tasks. To support such large-scale applications, a heterogeneous computing system composed of multiple computing platforms with different types such as supercomputers, grids, and clouds can be used. On allocating heterogeneous resources of the system to multiple users, there are three important aspects to consider: fairness among users, efficiency for maximizing the system throughput, and user satisfaction for reducing the average user response time. In this paper, we present three resource allocation policies for multi-user and multi-application workloads in a heterogeneous computing system. These three policies are a fairness policy, a greedy efficiency policy, and a fair efficiency policy. We evaluate and compare the performance of the three resource allocation policies over various settings of a heterogeneous computing system and loosely coupled applications, using simulation based on the trace from real experiments. Our simulation results show that the fair efficiency policy can provide competitive efficiency, with a balanced level of fairness and user satisfaction, compared to the other two resource allocation policies.

Priority-Based Task Scheduling in the Cloud Systems Using a Memetic Algorithm

Task scheduling is one of the major issues to achieve high performance in distributed systems such as Grid, Peer-to-Peer and cloud environment. Generally, there are two phases in heuristics-based task scheduling algorithms in heterogeneous distributed computing systems (HeDCSs). These phases are task prioritization and processor assigning respectively. Heuristic-based task scheduling algorithms may use different policies to assign priority to subtasks which produce different makespans in a heterogeneous computing system. Thus, a suitable scheduling algorithm is one that can efficiently assign a priority to tasks in order to minimize makespan. Recently, memetic algorithms (MAs) have been used as evolutionary or population-based global search approaches with local search heuristic to optimize NP-complete problems. Recent studies on MAs have discovered their success on a wide variety of real-world problems. Since the task scheduling problem is an NP-complete, in this paper, a new task scheduling algorithm on cloud environment using multiple priority queues and a memetic algorithm (MPQMA) is proposed. The proposed method uses a genetic algorithm (GA) along with hill climbing to assign a priority to each subtask while using a heuristic-based earliest finish time (EFT) approach to search for a solution for the task-to-processor mapping. The basic idea of our approach is using the advantage of MA to increase the convergence speed of the solutions. We implemented the algorithm on Azure Cloud Service by C# language where the experimental results for the set of randomly generated graphs revealed that the proposed MPQMA algorithm outperformed the existing three task scheduling algorithms in terms of makespan with fast convergence to the optimized solution.

Stochastic-based robust dynamic resource allocation for independent tasks in a heterogeneous computing system

Heterogeneous parallel and distributed computing systems frequently must operate in environments where there is uncertainty in system parameters. Robustness can be defined as the degree to which a system can function correctly in the presence of parameter values different from those assumed. In such an environment, the execution time of any given task may fluctuate substantially due to factors such as the content of data to be processed. Determining a resource allocation that is robust against this uncertainty is an important area of research. In this study, we define a stochastic robustness measure to facilitate resource allocation decisions in a dynamic environment where tasks are subject to individual hard deadlines and each task requires some input data to start execution. In this environment, the tasks that cannot meet their deadlines are dropped (i.e., discarded). We define methods to determine the stochastic completion times of tasks in the presence of the task dropping. The stochastic task completion time is used in the definition of the stochastic robustness measure. Based on this stochastic robustness measure, we design novel resource allocation techniques that work in immediate and batch modes, with the goal of maximizing the number of tasks that meet their individual deadlines. We compare the performance of our technique against several well-known approaches taken from the literature and adapted to our environment. Simulation results of this study demonstrate the suitability of our new technique in a dynamic heterogeneous computing system.

MeterPU: a generic measurement abstraction API

We present MeterPU, an easy-to-use, generic and low-overhead abstraction API for taking measurements of various metrics (time, energy) on different hardware components (e.g., CPU, DRAM, GPU) in a heterogeneous computer system, using pluggable platform-specific measurement implementations behind a common interface in C++. We show that with MeterPU, not only legacy (time) optimization frameworks, such as autotuned skeleton back-end selection, can be easily retargeted for energy optimization, but also switching between measurement metrics or techniques for arbitrary code sections now becomes trivial. We apply MeterPU to implement the first energy-tunable skeleton programming framework, based on the SkePU skeleton programming library.

Quantum-Inspired Hyper-Heuristics for Energy-Aware Scheduling on Heterogeneous Computing Systems

Power and performance tradeoff optimization is one of the most significant issues on heterogeneous multiprocessor or multicomputer systems (HMCSs) with dynamically variable voltage. In this paper, the problem is defined as energy-constrained performance optimization and performance-constrained energy optimization. Task scheduling for precedence-constrained parallel applications represented by a directed acyclic graph (DAG) in HMCSs is an NP-HARD problem. Over the last three decades, several task scheduling techniques have been developed for energy-aware scheduling. However, it is impossible for a single task scheduling technique to outperform all other techniques for all types of applications and situations. Motivated by these observations, hyper-heuristic framework is introduced. Moreover, a quantum-inspired high-level learning strategy is proposed to improve the performance of this framework. Meanwhile, a fast solution evaluation technique is designed to reduce the computational burden for each iteration step. Experimental results show that the fast solution evaluation technique can improve average algorithm search speed by 38 percent and that the proposed algorithm generally exhibits outstanding convergence performance.

Energy and Makespan Tradeoffs in Heterogeneous Computing Systems using Efficient Linear Programming Techniques

Resource management for large-scale high performance computing systems pose difficult challenges to system administrators. The extreme scale of these modern systems require task scheduling algorithms that are capable of handling at least millions of tasks and thousands of machines. These large computing systems consume vast amounts of electricity leading to high operating costs. System administrators try to simultaneously reduce operating costs and offer state-of-the-art performance; however, these are often conflicting objectives. Highly scalable algorithms are necessary to schedule tasks efficiently and to help system administrators gain insight into energy/performance trade-offs of the system. System administrators can examine this trade-off space to quantify how much a difference in the performance level will cost in electricity, or analyze how much performance can be expected within an energy budget. In this study, we design a novel linear programming based resource allocation algorithm for a heterogeneous computing system to efficiently compute high quality solutions for simultaneously minimizing energy and makespan. These solutions are used to bound the Pareto front to easily trade-off energy and performance. The new algorithms are highly scalable in both solution quality and computation time compared to existing algorithms, especially as the problem size increases.

Study of access template to graphics engine GM effect on the performance

The work objective is to study the effect of the graphical processor unit computational cores load level and memory access pattern on the memory bus bandwidth and scaling acceleration. The research subject is the problem of scalability of the parallel computing performance and acceleration. The following hypothesis is checked: while processing images for multi-core shared-memory systems, Gustafson - Barsis’s law is more crucial than the memory access template at the underloading of the GPU cores. The research methodology is a computational experiment with further analysis of the obtained results. The conclusions are as follows. The suggested hypothesis is proved. For that, a series of experiments on various heterogeneous computational systems with OpenCL standard support is conducted. The application field of the results obtained includes the development of algorithms and software for the highly parallel computer systems. The memory access template starts to place certain restrictions on the algorithm efficiency only when the load level of the computational cores is sufficient. Video cards with the private memory show more stable results in comparison to those which share memory with the central processing unit.

Harmonica: A Framework of Heterogeneous Computing Systems With Memristor-Based Neuromorphic Computing Accelerators

Following technology scaling, on-chip heterogeneous architecture emerges as a promising solution to combat the power wall of microprocessors. This work presents <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><i>Harmonica</i></b> —aframework of heterogeneous computing system enhanced by memristor-based neuromorphic computing accelerators (NCAs). In Harmonica, a conventional pipeline is augmented with a NCA which is designed to speedup artificial neural network (ANN) relevant executions by leveraging the extremely efficient mixed-signal computation capability of nanoscale memristor-based crossbar (MBC) arrays. With the help of a mixed-signal interconnection network (M-Net), the hierarchically arranged MBC arrays can accelerate the computation of a variety of ANNs. Moreover, an inline calibration scheme is proposed to ensure the computation accuracy degradation incurred by the memristor resistance shifting within an acceptable range during NCA executions. Compared to general-purpose processor, Harmonica can achieve on average 27.06 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\times$</tex-math></inline-formula> performance speedup and 25.23 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\times$</tex-math></inline-formula> energy savings when the NCA is configured with auto-associative memory (AAM) implementation. If the NCA is configured with multilayer perception (MLP) implementation, the performance speedup and energy savings can be boosted to 178.41 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\times$</tex-math></inline-formula> and 184.24 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\times$</tex-math></inline-formula> , respectively, with slightly degraded computation accuracy. Moreover, the performance and power efficiency of Harmonica are superior to the designs with either digital neural processing units (D-NPUs) or MBC arrays cooperating with a digital interconnection network. Compared to the baseline of general-purpose processor, the classification rate degradation of Harmonica in MLP or AAM is less than 8% or 4%, respectively.

Pricing Scheme for Heterogeneous Multiserver Cloud Computing System

Pricing Scheme for Heterogeneous Multiserver Cloud Computing System

Exploring the Efficacy of Branch and Bound Strategy for Scheduling Workflows on Heterogeneous Computing Systems

Computationally complex applications featuring workflows comprises of modules that can be deployed on Heterogeneous Computing Systems (HCS) for accomplishing high performance. The problem of scheduling the workflows on HCS is proven to be NP-Complete. In the present work, Branch and Bound (B&B) strategy for scheduling the workflows on HCS is proposed to attain globally optimal solutions. The primary merit of the proposed strategy is due to the estimation of the rank functions which are sharper and not complex. The proposed B&B strategy expands the most promising states first. The sharper ranks aid in converging to the solution quickly by pruning the unpromising states which do not lead to an optimal solution. Therefore, the search space is drastically reduced hence higher performance can be expected. The experimental results reveal that the proposed B&B scheme is efficient in exploring high potentials of B&B strategy in finding exact solutions. The performance analysis on a set of benchmark workflows shows that the proposed B&B strategy has generated optimal schedules for 94.37% of the cases.

Hybrid Bio-inspired Scheduling Algorithms for Batch of Tasks on Heterogeneous Computing System

Due to high operational cost, the problem of scheduling a batch of tasks (BoT) applications on heterogeneous computing system (HCS) remains a challenging problem. Accordingly, a plethora of evolutionary algorithms (EAs) and non-EAs have been proposed as solutions. Due to the ability of exploration of major solution space, EAs have been proven to be very effective in addressing the job scheduling problem. This work proposes two hybrid bio-inspired scheduling algorithms VPG and VDG featuring the combined best properties of VNS, PSO, DE and GA. The expected-time-to-compute (ETC) benchmark have been used to first present the performance of eight non-EAs viz. MCT, MinMin, MaxMin, Sufferage, HLTF, relative cost, MINMin and MINSuff in terms of makespan and energy consumption. The study is then extended to evaluate the performance of VPG, VDG and their seeded variants with GA, PSO and DE. Simulation study establishes the superior performance of VDG over peers.

CFD Problems Solving Parallel Approaches on Supercomputers

The work is devoted to developing and testing parallel algorithms, suit of computer programs for numerical solution of CFD-problems on modern supercomputers. The paper summarizes our experience in solving various practical problems of gas dynamics. These problems include research of gas flow around bodies of complex shape by viscous gas flows with radiation processes near the surface of the bodies, calculations of jet streams in the open space and flows in micro channels of technical systems. We have developed approaches combining unstructured grids, domain decomposition technique, parallel implementation via MPI technology, OpenMP, CUDA. Calculations are performed on the heterogeneous computer systems with Graphical Processing Units, classical microprocessors (Central Processing Units) and use advanced parallel approaches.

Numerical Simulation of the Perturbed KdVB Equation

The solution of nonintegrable nonlinear equations is very diffcult even numerically and practically impossible by standard analytical techniques. New view, offered by heterogeneous computational systems, gives new possibilities, but also need novel approaches for numerical realization of pertinent algorithms. We give some examples of such analysis on the base of nonlinear wave’s evolution study in multiphase media with chemical reaction.

Hybridization of Genetic Algorithm with Cuckoo Search Algorithm for Job Scheduling on Distributed Heterogeneous Computing Systems

Scheduling problems are at the heart of any Grid Computing system. Grid computing enables sharing, selection, aggregation and collections of geographically distributed heterogeneous resources to increase computational, and storage power and resource scheduling for solving NP-Hard problems. Different types of scheduling based on criteria, such as static, dynamic environment, adaptivity etc. The research already done in the field of other scheduling problems can be modified to be applied in this new scenario. Thus, we present an extensive study on the evolutionary hybridization of meta-heuristics methods for designing efficient Grid schedulers when makespan and flowtime are minimized. In this paper we present a hybridization of Genetic Algorithms (GA) and Cuckoo Search (CS) for scheduling in computational grids. We evaluated the proposed hybrid algorithm using different Grid scenarios generated by a Grid simulator. This proposed technique is effectively schedule jobs onto available resources and result showed that our scheduler outperform existing implementation in a grid environment thus resulting and also revealed their efficiency when makespan and flowtime are minimized.

Joint optimization of energy efficiency and system reliability for precedence constrained tasks in heterogeneous systems

Voltage scaling is a fundamental technique in the energy efficient computing field. Recent studies tackling this topic show degraded system reliability as frequency scales. To address this conflict, the subject of reliability aware power management (RAPM) has been extensively explored and is still under investigation. Heterogeneous Computing Systems (HCS) provide high performance potential which attracts researchers to consider these systems. Unfortunately, the existing scheduling algorithms for precedence constrained tasks with shared deadline in HCS do not adequately consider reliability conservation. In this study, we design joint optimization schemes of energy efficiency and system reliability for directed acyclic graph (DAG) by adopting the shared recovery technique, which can achieve high system reliability and noticeable energy preservation. To the best of our knowledge, this is the first time to address the problem in HCS. The extensive comparative evaluation studies for both randomly generated and some real-world applications graphs show that our scheduling algorithms are compelling in terms of enhancement of both system reliability and energy saving.

Energy‐efficient scheduling algorithms for batch‐of‐tasks (BoT) applications on heterogeneous computing systems

SummaryOne of the major design constraints of a heterogeneous computing system is optimal scheduling, that is, mapping of tasks on the processing nodes in order to optimize the QoS parameters. Because of the huge energy consumption by computing resources, negative environmental effects and reduced system reliability, energy has unavoidably been added as a new parameter to the list of QoS parameters. Energy optimization in scheduling strategies along with makespan makes it an even more challenging combinatorial optimization problem. This work proposes two energy‐aware scheduling algorithms G1 and G2 to schedule a batch‐of‐tasks, made of a collection of independent tasks, on heterogeneous processors in order to minimize the makespan and the energy consumption. The proposed algorithms schedule tasks based on weighted aggregation cost function to the appropriate processors followed by task migration phase designed to further minimize the makespan and the energy consumption. The study evaluates the performance of the proposed algorithms with some of the peers, that is, MinMin, MINSuff on account of makespan, energy consumption, flowtime, and utilization. An experimental study reveals that the proposed algorithm (G2) consistently performs better under various test conditions. Copyright © 2015 John Wiley &amp; Sons, Ltd.

A Hybrid Chemical Reaction Optimization Scheme for Task Scheduling on Heterogeneous Computing Systems

Scheduling for directed acyclic graph (DAG) tasks with the objective of minimizing makespan has become an important problem in a variety of applications on heterogeneous computing platforms, which involves making decisions about the execution order of tasks and task-to-processor mapping. Recently, the chemical reaction optimization (CRO) method has proved to be very effective in many fields. In this paper, an improved hybrid version of the CRO method called HCRO (hybrid CRO) is developed for solving the DAG-based task scheduling problem. In HCRO, the CRO method is integrated with the novel heuristic approaches, and a new selection strategy is proposed. More specifically, the following contributions are made in this paper. (1) A Gaussian random walk approach is proposed to search for optimal local candidate solutions. (2) A left or right rotating shift method based on the theory of maximum Hamming distance is used to guarantee that our HCRO algorithm can escape from local optima. (3) A novel selection strategy based on the normal distribution and a pseudo-random shuffle approach are developed to keep the molecular diversity. Moreover, an exclusive-OR (XOR) operator between two strings is introduced to reduce the chance of cloning before new molecules are generated. Both simulation and real-life experiments have been conducted in this paper to verify the effectiveness of HCRO. The results show that the HCRO algorithm schedules the DAG tasks much better than the existing algorithms in terms of makespan and speed of convergence.

Numerical simulation as an important tool in developing novel hypersonic technologies

Development of novel hypersonic technologies necessarily requires the development of methods for analyzing a motion of hypervelocity vehicles. This paper could be considered as the initial stage in developing of complex computational model for studying flows around hypervelocity vehicles of arbitrary shape. Essential part of the model is a solution to three-dimensional transport equations for mass, momentum and energy for the medium in the state of both LTE (local thermodynamic equilibrium) and non-LTE. One of the primary requirements to the developed model is the realization on the modern heterogeneous computer systems including both CPU and GPU. The paper presents the first results on numerical simulation of hypersonic flow. The first problem considered is three-dimensional flow around curved body under angle of attack. The performance of heterogeneous 4-GPU computer system is tested. The second problem highlights the capabilities of the developed model to study heat and mass transfer problems. Namely, interior heat problem is considered which takes into account ablation of thermal protection system and variation of the surface shape of the vehicle.