A simulation approach to the comparative analysis of various job scheduling methods

One of the main challenges in Grid systems is designing an adaptive, scalable, and model-independent method for job scheduling to achieve a desirable degree of load balancing and system efficiency. Centralized job scheduling methods have some drawbacks, such as single point of failure and lack of scalability. Moreover, decentralized methods require a coordination mechanism with limited communications. In this paper, we propose a multi-agent approach to job scheduling in Grid, named Centralized Learning Distributed Scheduling (CLDS), by utilizing the reinforcement learning framework. The CLDS is a model free approach that uses the information of jobs and their completion time to estimate the efficiency of resources. In this method, there are a learner agent and several scheduler agents that perform the task of learning and job scheduling with the use of a coordination strategy that maintains the communication cost at a limited level. We evaluated the efficiency of the CLDS method by designing and performing a set of experiments on a simulated Grid system under different system scales and loads. The results show that the CLDS can effectively balance the load of system even in large scale and heavy loaded Grids, while maintains its adaptive performance and scalability.

The use of parallel and distributed systems has become very common in the last decade. Dividing data is one of the challenges in these types of systems. Divisible load theory (DLT) is one of the proposed methods for scheduling data distribution in parallel or distributed systems. Many researches have been done in this field, but scheduling a multi-installment heterogeneous system with two-level hierarchical topology in which communication mode is blocking has not been addressed. In this paper, we find the proper size of task for each sub tree. Finally, in the experiments section, we show that the proposed methods work correctly and give us the best scheduling.

Deadline-aware job scheduling problems have been attracting attention in the application domains of scientific workflows and commercial data-center wide area networks. Satisfying all of the demands before the corresponding deadlines requires optimally spreading traffic along future times. In this paper, we propose a novel deadline-aware job scheduling method that leverages a deep reinforcement learning policy gradient algorithm. We obtain the following findings through several experiments on evaluating the proposed method. The EDF-trained initial policy with the proposed method outperforms a random initial policy. We evaluate how the training period influence the performance of the proposed method and the relationship between the training period length and the success ratio. We also demonstrate the proposed method outperformed with other job-scheduling methods, i.e., round-robin, first-come-first-served, and earliest-due-first, with the gaps between these methods and the proposed method expanding as the traffic load increased.

Multi-channel multi-core CPU parallel and CPU+GPU heterogeneous parallel are effective means to enhance Network Analysis Application computing. To solve the job scheduling problem of Network Analysis Application with multi-level dispatching, multi-user, multi-task in CPU/GPU heterogeneous cluster environment, a multi-user concurrent job scheduling method for Network Analysis Application in CPU/GPU cluster is proposed. According to the characteristics of Network Analysis Application, it is suggested that state estimation and dispatcher power flow should be accelerated in parallel with CPU, while static security analysis, perturbation power flow calculation and interruption capacity scanning should be accelerated in parallel with CPU+GPU. The job scheduling method proposed in this paper can satisfy the high concurrent requests of multi-level Dispatching & Control Center in the isomorphic and heterogeneous computing environments.

Fog computing is an extended version of cloud computing, which is used to serve huge data in the Internet of Things scenario. The development of fog computing improves the equipment’s computational power and delay by using modern information technology. Additionally, this fog computing improves calculation, communique, and storage services towards the edge of a network. The applications of fog computing are smart grid, street line parking automation system, smart health-care monitoring system, mobile big data analytics, etc. Additionally, the fog computing is used to solve the issues faced by the rapid development of IoT. The acceptable quality of experience is difficult to obtain among the concentrated applications and resource constrained gadgets. The aforementioned issues is resolved by using the task scheduling in fog computing. This paper discusses the classification of job scheduling and conventional methods that are used for job scheduling in fog computing along with its advantages and limitations. This detailed analysis will help researchers to obtain a better solution for the current concerns faced in fog computing in big data processing.

A distributed network of computer systems is operational at the medical faculty of Leiden. Data acquisition, interactive preprocessing and primary data storage is performed on small local systems. For program development, data storage and interactive dataprocessing centralised multiproject computers are used. The computers are interconnected using Decnet. Batch processing is done on mainframes via remote job entry.

The Intra-Laboratory Network (ILN) is a general-purpose computer communications network located at Argonne National Laboratory. This network provides communications between computers connected to the network. In addition to its general-purpose capability, the ILN also provides a remote job entry (RJE) system that allows users to move files from the user's computer system to ANL's Central Computing Facility (CCF), move files from the CCF to the user's computer system as batch job retrievals, check on the status of CCF jobs, cancel jobs, etc. This document describes the use of the ILN RJE system for users who have TI990 computer systems running the TX990/TXDS operating system. The ILN RJE system consists of three parts. Closest to the user is the ILN RJE utility program that runs on the user's host. Between the host and the CCF is the ILN which contains, in addition to the network control, an ILN RJE facility that supports the utility program and interfaces to the CCF. The third part is the CCF, which is accessed via the CCF front-end Varian V73 communications processors. The procedures for running the ILN RJE utility program are described first. Next, the commands that can be used to submit and retrieve jobs, check on job status, etc., are discussed. Each command description includes the command syntax, a brief description of the function, and one or more examples. Then techniques for transferring files between a host and either CCF or another host are outlined. Appendixes summarize the ILN RJE system commands, the possible error messages and the ASCII/EBCDIC conversion table that is used in the ILN. (RWR)

During the last decade, the mode of operation of many centralized computer installations has been significantly changed by the widespread use of remote job entry (RJE) facilities. The increasing usage of remote work stations presents the analyst a significant problem in sizing remotes to efficiently handle the flow of jobs. Currently, few, if any, tools exist to aid the analyst in this area.

A basic satellite system consists of a minicomputer, with card reader and line printer, emulating the Remote Job Entry (RJE) workstation of a mainframe manufacturer. The addition of a disk stroe increases the range of facilities which can be offered by such a system. Apart from a number of subsidiary advantages, this leads to a system which can interface a community of computer users to its mainframe computing facilities, and hence to a user image which is stable over some long period of time. The paper outlines how an advanced satellite system was simply assembled from four basic components. These components were written in different languages by different institutions, but the use of a standard message passing programming interface provided a sound framework for the detection and elimination of errors.

The job scheduling system is a core component of high-performance computer systems, and the goal of job scheduling is to meet the performance requirements under the premise, and get the lowest total power consumption of all tasks. In this paper, we design and realize a cluster job power scheduling strategy based on genetic algorithm. This strategy uses energy efficiency ratio of the job queue as job scheduling parameters. Experiments show the scheduling policy can raise energy efficiency on the basis of guarantee the response time and job throughput. The performance of this strategy is also better in savings energy consumption than traditional job scheduling strategy.

We describe a novel method for scheduling highspeed network switches. The targeted architecture is an input-buffered switch with a non-blocking switch fabric. The input buffers are organized as virtual output queues to avoid head-of-line blocking. The task of the scheduler is to decide when the input ports can forward packets from the virtual output queues to the corresponding output ports. Our Least Choice First (LCF) scheduling method selects the input and output ports to be matched by prioritizing the input ports according to the number of virtual output queues that contain packets: The fewer virtual output queues with packets, the higher- the scheduling priority, of the input port. This way the number- of switch connections and, with it, switch throughput is maximized. Fairness is provided through the addition of a round-robin algorithm. We present two alternative implementations: A central implementation intended for narrow, switches and a distributed implementation based on an iterative algorithm intended for wide switches. The simulation results show that the LCF scheduler outperforms other scheduling methods such as the parallel iterative matcher, iSLIP, and the wave front arbiter.

Fourier series is a powcrful tool for estimating frequency spectrum of a signal. The frequencies contained in the basis sinusoids of Fourier series are discrete, whereas, the frequency of Fourier transform are continuous. However, to compute Fourier transform, we have to calculate too much numerical integraions which cost very long CPU times. In this paper, we developed a fast computing algorithm to compute a Fourier transform where a fast computing algorithm FFT for a Fourier series was utilized.

Aims: We show the use of principal component analysis (PCA) and Fourier decomposition (FD) method as tools for variable star diagnostics and compare their relative performance in studying the changes in the light curve structures of pulsating Cepheids and in the classification of variable stars. Methods: We have calculated the Fourier parameters of 17,606 light curves of a variety of variables, e.g., RR Lyraes, Cepheids, Mira Variables and extrinsic variables for our analysis. We have also performed PCA on the same database of light curves. The inputs to the PCA are the 100 values of the magnitudes for each of these 17,606 light curves in the database interpolated between phase 0 to 1. Unlike some previous studies, Fourier coefficients are not used as input to the PCA. Results: We show that in general, the first few principal components (PCs) are enough to reconstruct the original light curves compared to FD method where 2 to 3 times more number of parameters are required to satisfactorily reconstruct the light curves. The computation of required number of Fourier parameters on the average needs 20 times more CPU time than the computation of required number of PCs. Therefore, PCA does have some advantage over the FD method in analysing the variable stars in a larger database. However in some cases, particularly in finding the resonances in Fundamental mode (FU) Cepheids, the PCA results show no distinct advantages over the FD method. We also demonstrate that the PCA technique can be used to classify variables into different variability classes in an automated, unsupervised way, a feature that has immense potential for larger databases of the future.

An operating system is software that is designed to manage computer hardware and software resources. However, this management requires applying an ample number of techniques and algorithms which are called synchronization and scheduling. The scheduling algorithms are used to arrange the way that the CPU is assigned to the processes, while synchronization is utilized to indicate how to work with multi-processes at the same time. Therefore, they are related to each other. CPU scheduling is a vital phenomenon of an operating system. At present, numerous CPU scheduling algorithms exist as First Come First Serve) FCFS(, Shortest Job First (SJF), Shortest Remaining Time First (SRTF), Priority Scheduling, and Round Robin (RR). In this paper, a survey of the current synchronization and scheduling algorithms have been presented. An overview of each technique with the main algorithms have been described in detail with the advantages and the issues of each algorithm. Furthermore, this paper has dug deep into the real-time operating system scheduling issues, which is the current trend in operating system researches.

In this paper, an axial flux coil assisted stator excitation BLDCM (CASEM) is designed. Because of the high magnetic saturation and strong coupling, the mathematical model of the motor has strong nonlinearity, and it is difficult to establish an accurate flux linkage model. Therefore, a three-dimensional flux linkage modeling method based on Fourier series is proposed in this paper, which considers multi-phase coupling and non-linear problems. Finally, a simulation model and an experimental platform are established to verify the correctness and effectiveness of the modeling method, which lays a foundation for further research on motor control to reduce torque ripple.