Performance Of Parallel Systems Research Articles

The realization of parallel airflow after flow rectification by the perforated plate under complex inflows

The parallel push-pull ventilation system uses uniform, low turbulence, and wide parallel airflow to achieve low diffusion transport of pollutants and reduce the system's energy consumption. Complex inflows caused by local connectors (fans, elbows, and reducers) in front of the air supply device seriously affect the air supply distribution. Currently, most existing studies focus on the performance of parallel push-pull ventilation systems, while lacking theoretical support and parameter guidance to rectify these complex inflows into parallel airflow. This study investigates parallel airflow after rectifying various complex inflows using perforated plates. Within the study context, three typical complex inflows models (fan, elbow, and reducer inflows) are established using dominant numerical and supporting experimental methods. In addition, the rectification effects of inflow velocity, porosity, and hole size of the perforated plate on various complex inflows are compared. The results show that the perforated plate's porosity and hole size should be reasonably matched under different complex inflows to achieve a uniform and low turbulence parallel flow. The perforated plate should have a porosity between 40% and 55% and a dimensionless hole size between 0.014 and 0.031 for elbow and reducer inflows. Swirling flow causes a reduction in the perforated plate's rectification capacity for fan inflow, requiring lower porosity and hole size, while also meaning that the resistance increases significantly. Overall, the study results are anticipated to aid in achieving supply airflow with parallel airflow.

Omega multistage interconnection network manages double-pattern traffic with a regulator and high-speed forwarding method

This study proposes a flexible and effective solution for developing high-performance multistage interconnection networks to maximize the performance of parallel computer systems, cloud computing infrastructure, grids, etc. An omega-type multistage interconnection network, consisting of regulated switchboxes, is used as a testbed to handle flexible two-class load patterns. The wormhole routing method and a special forwarding technique controlled by a global regulator are adopted to alleviate internal ‘tree saturation’ caused by periodic hotspot traffic combined with uniform traffic. Simulation experiments prove that this concept reduces packet latency, and an additional layer inserted in the final stage further improves the network’s architecture.

The Tiny-Tasks Granularity Trade-Off: Balancing Overhead Versus Performance in Parallel Systems

Models of parallel processing systems typically assume that one has <inline-formula><tex-math notation="LaTeX">$l$</tex-math></inline-formula> workers and jobs are split into an equal number of <inline-formula><tex-math notation="LaTeX">$k=l$</tex-math></inline-formula> tasks. Splitting jobs into <inline-formula><tex-math notation="LaTeX">$k &gt; l$</tex-math></inline-formula> smaller tasks, i.e. using “tiny tasks”, can yield performance and stability improvements because it reduces the variance in the amount of work assigned to each worker, but as <inline-formula><tex-math notation="LaTeX">$k$</tex-math></inline-formula> increases, the overhead involved in scheduling and managing the tasks begins to overtake the performance benefit. We perform extensive experiments on the effects of task granularity on an Apache Spark cluster, and based on these, develop a four-parameter model for task and job overhead that, in simulation, produces sojourn time distributions that match those of the real system. We also present analytical results which illustrate how using tiny tasks improves the stability region of split-merge systems, and analytical bounds on the sojourn and waiting time distributions of both split-merge and single-queue fork-join systems with tiny tasks. Finally we combine the overhead model with the analytical models to produce an analytical approximation to the sojourn and waiting time distributions of systems with tiny tasks which include overhead. We also perform analogous tiny-tasks experiments on a hybrid multi-processor shared memory system based on MPI and OpenMP which has no load-balancing between nodes. Though no longer strict analytical bounds, our analytical approximations with overhead match both the Spark and MPI/OpenMP experimental results very well.

Comprehensive performance of parallel feed MEE-MVC evaporation system

Abstract The mathematical model based on conservation of energy, mass and salinity is established to analyse the comprehensive performance of parallel feed multi-effect evaporation system with mechanical vapor compression (MEE-MVC) system. It is found that the efficiency of the MVC evaporation system gradually improves as the number of evaporator effect increases. However, the optimization trend of specific power consumption (SPC) and coefficient of performance (COP) are declining when the number of evaporator effect exceeds 3, which makes the marginal effect of the efforts made to improve the system gradually weakened. Then, the three-effect MVC evaporation system is taken as the main research subject to investigate comprehensive performance of the system. It was found that the SPC of the system increases with the increase of the total heat transfer temperature difference; on the other hand, the COP decreased significantly with the increase of the total heat transfer temperature difference. Furthermore, increasing the evaporation temperature in the last evaporator as well as compressor efficiency can optimize the energy efficiency of the system. In addition, the compressor efficiency has a greater influence on the SPC when the total heat transfer temperature difference is larger, whereas the compressor efficiency has a greater influence on the COP when the total heat transfer temperature difference is smaller.

Dynamic Frequency Performance Analysis and Improvement for Parallel VSG Systems Considering Virtual Inertia and Damping Coefficient

Parallel virtual synchronous generator (VSG) systems are emerging as an attractive solution for inertia support and frequency stability enhancement of the power grid. However, the dynamic frequency characteristic is more complicated due to the coupled interaction of line impedance, virtual inertia, and damping coefficient in parallel VSG systems. The problems for frequency stability, dynamic performance, and frequency/power oscillation have raised much more attention. In this article, a systematic analysis of the influence of virtual inertia and damping coefficient on the dynamic frequency characteristic among parallel VSG systems was presented, and a quantitative assessment for dynamic frequency performance based on the 2-norm <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\boldsymbol {\mathcal{ H}}_{2}$ </tex-math></inline-formula> was proposed. Furthermore, a coordinated control for virtual inertia and damping was proposed to improve frequency stability and achieve a better dynamic performance of parallel VSG systems. The effectiveness and superiority of the proposed method were verified by simulation and experimental results.

Design-time performance modeling of compositional parallel programs

Performance models are powerful instruments for understanding the performance of parallel systems and uncovering their bottlenecks. Already during system design, performance models can help ponder alternative development options. However, creating a performance model – whether theoretically or empirically – for an entire application that does not exist yet is challenging. In this paper, we propose to generate performance models of full programs from performance models of their components using formal composition operators derived from parallel design patterns. As long as the design of the overall system follows such a pattern, its performance model can be predicted with reasonable accuracy without an actual implementation. We demonstrate our approach with design patterns of varying complexity, including pipeline, task pool, and eventually MapReduce, which is representative of a broad class of data-analytics applications.

Performance optimization for parallel systems with shared DWM via retiming, loop scheduling, and data placement

Abstract Domain Wall Memory (DWM) as an ideal candidate for replacing traditional memories especially in parallel systems, has many desirable characteristics such as low leakage power, high density and low access latency. However, due to the tape-like architecture of DWM, shift operations have vital impact on performance. Considering data-intensive applications with massive loops and arrays, increasing parallelism of loops, appropriate loop scheduling and data placement on DWM will significantly improve the performance of parallel systems. This paper explores optimizing performance of parallel systems through retiming, loop scheduling and data placement especially when the data are arrays. It proposes Integer Linear Programming (ILP) formulation and Scheduling While Placing (SWP) algorithm to generate optimal or nearly optimal loop scheduling and data placement with minimum execution time. The experimental results show that SWP and ILP can effectively reduce execution time when compared with greedy List Scheduling First Access First Place (LF) algorithm. Besides, this paper proposes Threshold Retiming Repetition (TRR) algorithm to combine retiming technique with SWP or ILP to improve performance. The experimental results show that SWP+TRR and ILP+TRR can further reduce the execution time when compared to results without retiming.

Measuring the impact of burst buffers on data-intensive scientific workflows

Science applications frequently produce and consume large volumes of data, but delivering this data to and from compute resources can be challenging, as parallel file system performance is not keeping up with compute and memory performance. To mitigate this I/O bottleneck, some systems have deployed burst buffers, but their impact on performance for real-world scientific workflow applications is still not clear. In this paper, we examine the impact of burst buffers through the remote-shared, allocatable burst buffers on the Cori system at NERSC. By running two data-intensive workflows, a high-throughput genome analysis workflow, and a subset of the SCEC high-performance CyberShake workflow, a production seismic hazard analysis workflow, we find that using burst buffers offers read and write improvements of an order of magnitude, and these improvements lead to increased job performance, and thereby increased overall workflow performance, even for long-running CPU-bound jobs.

Provisioning and Performance Evaluation of Parallel Systems with Output Synchronization

Parallel server frameworks are widely deployed in modern large-data processing applications. Intuitively, splitting and parallel processing of the workload provides accelerated application response times and scaling flexibility. Examples of such frameworks include MapReduce, Hadoop, and Spark. For many applications, the dynamics of such systems are naturally captured by a Fork-Join (FJ) queuing model, where incoming jobs are split into tasks each of which is mapped to exactly one server. When all the tasks that belong to one job are executed, the job is reassembled and leaves the system. We consider this behavior at the output as a synchronization constraint. In this article, we study the performance of such parallel systems for different server properties, e.g., work-conservingness, phase-type behavior, and as suggested by recent evidence, for bursty input job arrivals. We establish a Large Deviations Principle for the steady-state job waiting times in an FJ system based on Markov-additive processes. Building on that, we present a performance analysis framework for FJ systems and provide computable bounds on the tail probabilities of the steady-state waiting times. We validate our bounds using estimates obtained through simulations. In addition, we define and analyze provisioning, a flexible division of jobs into tasks, in FJ systems. Finally, we use this framework together with real-world traces to show the benefits of an adaptive provisioning system that adjusts the service within an FJ system based on the arrival intensity.

Analysis of vibration isolation performance of parallel air spring system for precision equipment transportation

Precision equipment is usually accompanied with vibrations during road or railway transportation. Sometimes the vibration exceeds the given limit, leading to the damage of the equipment. It is necessary to control the vibration during the transportation. However, it is still difficult to adjust the parameters of a designed vibration isolation system for the transportation of different precision equipment under various road conditions. Aiming at satisfying the vibration isolation requirements of different precision equipment, this paper proposes a parallel air spring vibration isolation system based on the principle of limiting lateral deflection. According to the measured parameters, a rigid-body dynamics simulation model of parallel air spring vibration isolation system is established. Then its feasibility is verified, and the optimal parameters of the vibration isolation system are obtained by a simulation. Finally, the vibration isolation system is built and installed in the equipment to carry out the real vehicle transportation test. The test results show that the transportation vibration isolation system based on the parallel air spring structure has not only excellent vibration isolation efficiency but also acceptable lateral stability. The research results in this paper can provide a reference for the design of the vibration isolation system for the large precision equipment transportation.

An Approximation Method of Parallel Processing System with Big Data

Big data has been amassed and observed at bursting dimensions. The immense capacity of unstructured data from digital society has intensified at an unimaginable bound, motivating processing power and data pre-processing. The conventional simulations to manipulate gigantic data can be problematic as they cannot be practically used. The decision-maker is experiencing these problems from all activities in the workplace community, enterprise, and scientific experiments. The troubles arise from the process of data collection, the way to prepare a suitable storage space, and the improvement of processing speed. These burdens are growing for big data processing. Besides, there are some points such as data conflict, severance, confidentiality, extensibility, and time-series. Parallel processing system is a key to quantify metrics, and to investigate time-series. Recently developed simulation, so called Massive Online Analysis, gives results from multiple datasets and the results align with ones from the approximation. Cost-effective and speedup metrics are examined to confirm the performance of parallel processing system.

Experimental study of single linear compressor driving two pulse tube refrigerators with adjusting valves

Parallel system in which multi pulse tube cold fingers (PTCFs) driven by one linear compressor can achieve low temperature at different positions, which would reduce the cost and weight of cryocooler. In order to achieve low temperature at different positions, two adjusting valves are installed on the connecting tubes between linear compressor and two PTCFs. By actively adjusting the mass distribution of helium gas in different PTCFs, the cooling capacity and cooling temperature of the two PTCFs may be adjusted. Experimental results show that in the case of greater input electric power, the cooling performance of parallel system is better than that of single system (one compressor driving one PTCF). The adjusting valves can effectively change the cooling performances of two PTCFs in parallel system. However, the adjusting valves will bring large pressure drop, which results in deterioration of refrigeration performance of parallel system.

HYBRID BAT AND CUCKOO BASED META-HEURISTIC ALGORITHM FOR PARALLEL JOB SCHEDULING IN MULTI-CLUSTER ENVIRONMENT

Scheduling in multi cluster environment is NP Hard but yet it is a fundamental to satisfying and improving the performance of parallel processing system. In this paper the parallel job scheduling problem is being addressed in multicluster computing environment using hybrid BAT+Cuckoo Algorithm. Scheduling policy is considered in which BAT algorithm is used to allocate the available resources and Cuckoo algorithm is used to look for constraint resources. Constraint resources may be located within distinct clusters.A mating pool in which resources from distinct clusters are maintained. In case resources are limited, Cuckoo algorithm look for available resources from each cluster and combine them within the mating pool. Parameters considered for evaluation includes Makespan and Flowtime which are substantially reduce using the proposed literature.

A coarse‐grained page cache aware multivariate analytical model for the storage performance of a parallel file system

SummaryHuge requirements of emerging Big Data applications combined with the performance impairment of current I/O subsystems pose a great challenge on data storage, management, and access performance. In order to design efficient storage systems, a clear understanding about how factors in the I/O path affect the performance of a data‐intensive application is of utmost importance. This paper reports our ongoing research toward addressing this issue, presenting a coarse‐grained page cache aware multivariate analytical model for the performance of write operations in a parallel file system. The proposed model was developed to reflect the performance behavior observed in an extensive experimental effort, in which the impact of 14 parameters in the response time and throughput of the OrangeFS was investigated. More than one million experiments were carried out using four distinct computing infrastructures, providing a detailed performance characterization. Additionally, a thorough evaluation of the proposed model, covering more than 14 000 scenarios, is reported, discussing both qualitative and quantitative aspects. Evaluation results indicate that the model succeeded in representing the behavior of the parallel file system performance, achieving a Mean Absolute Percentage Error of 39.94%.

Performance Analysis and Relay Location Research of OFDM Free-Space Optical Communication Systems Under Moderate and Strong Turbulence

In this paper, based on the Gamma–Gamma channel model for describing moderate and strong atmospheric turbulence, we study the relay location of serial decode-and-forward relay systems and parallel decode-and-forward relay systems in free-space optical (FSO) communication. According to the orthogonal frequency-division multiplexing modulation (OFDM) and coherent detection demodulation technology, we develop a novel statistical fading channel model for relay FSO systems by incorporating the atmospheric turbulence, pointing errors, and path loss effects. Based on this channel model, we derive the closed-form expression of the outage probability in the FSO serial relay system and parallel relay system, using the Meijer G-function. The serial decode-and-forward relay system with different relay locations and parallel decode-and-forward relay system, which consider different number of links and different relay locations, are simulation analyzed under moderate and strong atmospheric turbulence. The performance of serial relay systems and parallel relay systems in free-space optical communication can be improved by optimizing the relay location.

Optimal loading of series parallel systems with arbitrary element time-to-failure and time-to-repair distributions

Many real-life systems have series parallel structures, where some components or subsystems work in parallel while some others have to work in series or consecutively. This paper models dynamic performance of multi-state series parallel systems with repairable elements that can function at different load levels. Performance (productivity) and time-to-failure distribution of an operating system element depend on its load level. An element, upon failure, can be repaired with repair time obeying a known distribution. The entire system must satisfy a random demand during a fixed mission time or must complete a fixed amount of work. A discrete numerical algorithm is first proposed for evaluating instantaneous availability of a system element with a particular load level, which further defines stochastic process of the element's performance. A universal generating function technique is then used for assessing system performance metrics including expected system performance, expected probability of meeting system demand, expected amount of unsupplied demand over a particular mission time and expected time needed to perform a given amount of work for the considered system. The proposed methodology is applicable to arbitrary types of failure time and repair time distributions. Another original contribution of this work is formulating and solving elements loading optimization problems, which choose elements load levels to achieve one of the following objectives: maximum system expected performance, maximum expected probability of meeting demand during a time horizon, minimum total unsupplied demand during the mission, or minimum completion time for a given amount of work. As demonstrated through a case study of a power station coal transportation system, optimization results can provide effective guidance on optimal operational load of multi-state series parallel system elements.

The design and implementation of parallel switch system simulation platform based on Java

During the study of switching fabric and scheduling algorithm, simulation has become an important mean to analyze and evaluate the performance of a design scheme. But the simulation software about switching fabric and scheduling algorithm has defects in inheritance and scalability exist at present. In order to evaluate the performance of parallel switch system intuitively and accurately, the simulation platform PPS-CICQ-SIM has been designed and developed, which adopted object-oriented technology for analysis, design and implement. The simulation platform overcomes the problems of the existing simulation software, and provides an efficient mean to evaluate the performance of switch system. At the end of this paper, the simulation result of typical parallel switch system scheduling algorithms such as iSLIP, OQ were given, which verified the effectiveness of the simulation platform PPS-CICQ-SIM.

Elastic-RAID: A New Architecture for Improved Availability of Parity-Based RAIDs by Elastic Mirroring

In this paper, we propose Elastic-RAID, a new RAID architecture to achieve high performance and high reliability for large-scale distributed and parallel storage systems. The key idea behind Elastic-RAID is to smartly utilize the free space existing in parity-based disk arrays to store additional mirroring data. This additional mirroring data redundancy, when strategically and judiciously activated and exploited in a RAID system, enables improved system I/O performance, fault tolerance and recovery. Depending on the amount of free space available and whether the emphasis is on performance or reliability, the elasticity in Elastic-RAID is manifested in how each design objective is achieved. For the performance objective, Elastic-RAID improves small-write performance by writing original and mirroring data synchronously and leaving the costly parity update in the background at a later idle/lightly-loaded time. For the reliability objective, at least two concurrent disk failures can be tolerated when Elastic-RAID is employed in a RAID5 system that has 50 percent or more free space. Higher reliability is provided for important data when free space is less than 50 percent. To achieve the design goal of elasticity, we introduce a novel data layout and addressing scheme. Our extensive trace-driven evaluations on an Elastic-RAID prototype in the typical configurations of RAID5 show that Elastic-RAID boosts the small-write performance in the normal operational state by at least 40 percent, improves the user I/O performance in the reconstruction state by at least 30 percent and shortens the recovery time by at least 40 percent.

Load Balancing Approaches: Recent Computing Trends

This paper presents thorough survey of work addressing on load balancing in recent computing trends. There are many issues whose solutions lead to the need for load balancing. The objective of load balancing is to increase the performance of parallel and distributed system by distributing the load among the processors. Load balancing is a major factor for achieving high performance. It affects the execution time significantly by expediting it. Load imbalance is a wellknown problem in the areas involving parallelism. However, offering load balancing is a difficult and challenging task. Various algorithms have been proposed for load balancing. These algorithms have distinguished features and each uses different mechanisms. Various Load balancing algorithms like biased sampling, honey bee, active clustering, and join idle queue have been studied.

Reliability analysis of multilayer multistage interconnection networks

Multistage interconnection networks (MINs) play a key role in the performance of parallel computers and multiprocessor systems. A non-negligible demand on today's modern systems is to deliver multicast traffic. Therefore, design of efficient MINs that meets the routing requirement is vital. One of the main ideas to cope with this problem is the use of replicated MINs. However, one of the major concerns about these networks is the problem of unnecessary layer replication in the first stages, which recently proposes a new idea called multilayer MINs. Previous analyzes demonstrated that this new idea could lead to cost-effective topologies that had a very close performance to the replicated MINs in terms of throughput. Also, these analyzes indicate that these networks outperform replicated MINs in terms of delay. However, another critical parameter to prove the performance of most systems is reliability. Therefore, in this paper, we will focus on two essential parameters of cost and reliability to achieve both objectives; first, evaluating the performance of multilayer MINs in terms of reliability, second, to find the best topology among the multilayer MINs introduced in previous works in terms of cost-effectiveness (mean time to failure/cost ratio).