Scheduling jobs across geo-distributed datacenters

With the rapid development of digital technology, from the application of traditional databases and scientific computing to the emerging cloud computing services, the analysis and processing of massive data has become the focus of society. Providing low-cost, scalable, and configurable shared cloud services to users on cloud service platforms is a new hotspot for the development of major cloud service providers. Job scheduling plays an important role in improving the overall system performance of cloud service capabilities. Simple job scheduling strategies (such as Fair and FIFO scheduling) do not consider job size and may degrade performance when jobs of different sizes arrive. This paper proposes the MQWAG (Multi-queue Load-Sensitive Greedy Scheduling Algorithm) job scheduling algorithm to reorder multi-queue jobs so that short jobs are executed preferentially in multiple queues. In our experiments, our algorithm shortened the average job completion time by about 26% compared with other algorithms.

In this paper, we study the problem of job dispatching and scheduling, where each job consists of a set of tasks. Each task is processed by a set of machines simultaneously. We consider two important performance metrics, the average job completion time (JCT), and the number of deadline-aware jobs that meet their deadlines. The goal is to minimize the former and maximize the latter. We first propose OneJ to minimize the job completion time (JCT) when there is exactly one single job in the system. Then, we propose an online algorithm called MultiJ, taking OneJ as a subroutine, to minimize the average JCT, and prove it has a good competitive ratio. We then derive another online algorithm QuickJ to maximize the number of jobs that can meet their deadlines. We show that QuickJ is competitive via a worst case analysis. We also conjecture that the competitive ratio of QuickJ is likely to be the best one that any deterministic algorithm can achieve. We also shed light on several important merits of MultiJ and QuickJ, such as no severe coordination overhead, scalability, work conservation, and no job starvation.

Scheduling is an important technique to improve the productivity of workflow applications in a grid system. Serving workflow applications added overhead on the scheduling system since the scheduler should choose the best zero‐dependent job that improves the performance of a grid system. This paper proposes two dependent job scheduling algorithms based on analyzing the effect of categorical and continuous variables of the job, which are used in the calculation of job weight. To find the job weight, twostep clustering is used with 10 groups and the ranking equation. In addition, to verify the ability to apply the proposed algorithms in a real environment, weighted least squares estimation is used. The results showed that the prediction rate is equal to 99.88%, which indicates that the proposed algorithms could be applied in a real grid system with low overhead. Through simulations and after testing the proposed algorithms, the average results showed that the Dependent Job Scheduling (DJS) algorithm outperformed the previous algorithms, in total execution time and an average waiting time with an improvement value of 1.18 and 1.92 times, respectively. While DJS algorithm with weighting factor (DJSJP) outperformed the previous algorithms in total execution time only with an improvement value equal to 1.17 times. The overall results indicated that the proposed algorithms are efficient to be used in a grid system, besides four bits are sufficient to improve the performance of the job scheduling system.

Cloud Computing provides a Computing environment where different resources, infrastructures, development platforms and software are delivered as a service to customers virtually on pay per time basis. Low cost, scalability, reliability, utility-based computing are important aspects of cloud computing. Job scheduling is an essential and most important part in any cloud environment. With increasing number of users, Job scheduling becomes a strenuous task. Ordering the jobs by scheduler while maintaining the balance between quality of services (QoS), efficiency and fairness of jobs is quite challenging. Scheduling algorithms are implemented considering parameters such as throughput, resource utilization, latency, cost, priority, computational time, physical distances, performance, bandwidth, resource availability. Though there are different scheduling algorithms available in cloud computing, a very less comparative study has been done on performance of various scheduling algorithms with respect to above mentioned parameters. This paper aims at a comparative study of various types of job scheduling algorithms that provide efficient cloud services.

Grid Computing provides sharing of geographically distributed resources among large scale complex applications. Due to dynamic nature of resources in grid, there is a need of highly efficient job scheduling and resource management policies in grid. In this paper, a novel Grid Resource Scheduler (GRS) is proposed to effectively utilize the available resources in Grid. Proposed GRS contributes, an optimal job scheduling algorithm on Job Rank-Backfilling policy and a resource matching algorithm based on ranking of resources with best fit allocation model. Performance of GRS is measured by considering a web based BLAST algorithm -- a bioinformatics application -- . GRS aims in reducing; Make span of the job workflow and execution time of varied size jobs. The experimental results prove that proposed job scheduling and resource matching algorithms performs better when evaluated against widely used FCFS job scheduling and biggest free resource matching strategies.

Two notable characteristics of modern GPU-accelerated HPC clusters are: (1) they increasingly run deep learning (DL) model-training workloads, and (2) they consist of multiple generations of GPUs, i.e., they are heterogeneous. However, existing works in GPU cluster scheduling for DL workloads have not addressed the GPU multi-generation problem. We propose DASH, a GPU cluster scheduler designed to optimally make a match between different DL workloads and GPU types in a multi-generational GPU environment. By lever-aging execution characteristics of co-scheduled DL workloads, DASH can improve the average job runtime by 17% and the average job completion time by 14 % compared to the traditional heterogeneity-unaware job scheduler.

Job scheduling in grid computing is a very important problem. To utilize grids efficiently, we need a good job scheduling algorithm to assign jobs to resources in grids. The main scope of this article is to propose a new Ant Colony Optimization (ACO) algorithm for balanced job scheduling in the Grid environment. To achieve the above goal, we will indicate a way to balance the entire system load while minimizing the makespan of a given set of jobs. Based on the experimental results, the proposed algorithm confidently demonstrates its practicability and competitiveness compared with other job scheduling algorithms.

<para xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> This paper deals with vehicle-scheduling problem (VSP) in an automatic material-handling environment in 300-mm semiconductor wafer manufacturing. We adopt Petri nets (PNs) modeling techniques to model the complicated coupling dynamics among transport jobs and overhead hoist transport (OHT) vehicles in a 300-mm OHT loop. The congestion phenomenon among OHT vehicles is captured. With help of the PN models, we formulate the OHT VSP as an integer programming problem whose objective is to schedule OHT vehicles to transport jobs such that average job completion time is minimized. Instead of solving for the optimal solution, we develop a solution methodology to generate a feasible schedule efficiently. A Lagrangian relaxation step is first taken to decompose the PN-based, integer programming problem into individual job-scheduling subproblems. To reduce computation efforts in solving each subproblem optimally, we develop an approximation method to solve each job subproblem by utilizing a reduced PN model of the job. Lagrangian multipliers are then optimized by a surrogate subgradient method. A heuristic algorithm is developed to adjust the dual solution to a feasible schedule. Numerical results demonstrate that our solution methodology can generate good schedules within a reasonable amount of computation time for realistic problems. Compared to a popular vehicle-dispatching rule, our approach can achieve in average 32% improvements on the average delivery time in our realistic test cases. </para>

In the field of Big Data analytics the steps taken to collect and analyze the large amount of data to get useful information is the major use of cloud services today. Traditionally the large volume of data was gathered, stored and analyzed in only one data center. As the volume of data has started increasing at a tremendous rate, using the single data centers will lead to improper and less efficient handling of data from the point of performance, where the single data center cannot handle the huge volume of datasets. So, for the better performance and availability, deploying the multiple data centers around the world geographically by the large cloud services providers can resolve this challenge [9]. There are many approaches but widely used approach for the analytics of the geographical distributed data is the centralized approaches. In centralized approach, the data is gathered and stored locally in the main datacenter but this approach may leads to worse performance because this approach may consume a significant amount of bandwidth. To achieve the optimal performance numbers of mechanisms have been used where the data analytics has been performed over the geo-distributed data center.

Job scheduling is a fundamental issue in achieving high performance on the Grids. In this paper, a hierarchical job scheduling architecture has been proposed for the Grids. The stochastic high-level Petri net (SHLPN) model of the proposed architecture and the refined model are presented. The job scheduling algorithms for an incoming job to be distributed to a waiting queue of any processor in any node of the Grid architecture is integrated by the node-selection algorithm at the global scheduler, the job-dispatching algorithm at the local scheduler, and job-scheduling algorithm at the processor scheduler. The integrated job scheduling policy is investigated using SHLPN performance analysis methods. The performance metrics concerned include the delay time of jobs in the Grid node and the response time perceived by the users. The numerical example shows the efficiency of the performance evaluation technique.

Grid Computing is a widespread distributed computing environment that consists of heterogeneous resources and it provides huge computational power for the large scale distributed application. In this job scheduling is an essential objective to prioritize the jobs based on the job characteristics. It is a big challenge to design and implement an efficient Job Scheduling algorithm and the Job Scheduling model. This paper proposes the priority based Job Scheduling algorithm with Nonpreemptive Priority (NPRP) Job Scheduling model. The priority based job scheduling algorithm is the combination of Backfill algorithm and Earliest Deadline based algorithm. The proposed model has been integrated with Care Resource Broker (CRB) which supports virtualized grid environment. The proposed model has been integrated with CRB helps to make the intelligent scheduling decisions for prioritizing the jobs and also the proposed model minimizes the turnaround time, waiting time and response time of the jobs available in the queue and the user's quality of service has been maximized in the heterogenous and virtualized grid environment.

Decentralized distributed systems, such as grids, clouds or networks of sensors, have been widely investigated recently. An important nature of such systems is the heterogeneity of their resources; in order to archive the availability, scalability and flexibility. As a consequence, managing the systems to meet requirements is obviously a nontrivial work. The issue is even more challenging in term of job scheduling when the task dependency within each job exists. In this paper, we address such problem of job scheduling, so called workflow-based job scheduling, in the decentralized distributed systems with heterogeneous resources. As such problem is proven to be an NP-complete problem, an efficient heuristic algorithm to address this problem is proposed. The algorithm is based on an observation that the heterogeneity of the resources can affect the execution time of the scheduling. We compare the effectiveness and efficiency of the proposed algorithm with a baseline algorithm. The result shows that our algorithm is highly effective and efficient for the scheduling problem in the decentralized distributed system with heterogeneous resources environment both in terms of the solution quality and the execution time respectively.KeywordsDecentralized algorithmHeterogeneous resourcesWorkflow-based job scheduling

Cloud computing is widely used in academia and industry. Users can provide jobs to cloud service providers that provide pay-as-you-go methods by cloud for calculation and obtaining results. Cloud service providers make reasonable scheduling for different jobs of users and in the case of successful operation, the cost of energy consumption becomes a key factor. Prior works proposed various algorithms based on reinforcement learning on job scheduling. However, due to the lack of online training mode and refined container decision-making, this paper makes some improvements and extensions. This paper presents the algorithm ACT4JS, an actor-critic with transformer for three stage job scheduling algorithm, that finely manages data centers, server nodes, and application containers. The job scheduling algorithm is designed to select the best long-term decision by learning from cloud computing resource model that accepts single-task multi-instance jobs and directed acyclic graph jobs with dependencies. To learn key features and loss optimization, use the transformer network structure and the proximal policy optimization method. To minimize energy consumption costs for large scale job scheduling and running, designs a reward function to reflect optimized energy consumption targets and independent simulation experiments. To achieve low energy consumption cost and low task reject rate, use training methods such as experience replay and advantage estimate. Compared with the mainstream baseline algorithm, the results show that the ACT4JS algorithm is better than the baseline in the online training model. In addition, the effect of ACT4JS is more effective in the long-term experimental.

It has become routine for large volumes of data to be generated, stored, and processed across geographically distributed datacenters. To run a single data analytic job on such geo-distributed data, recent research proposed to distribute its tasks across datacenters, considering both data locality and network bandwidth across datacenters. Yet, it remains an open problem in the more general case, where multiple analytic jobs need to fairly share the resources at these geo-distributed datacenters. In this paper, we focus on the problem of assigning tasks belonging to multiple jobs across datacenters, with the specific objective of achieving max-min fairness across jobs sharing these datacenters, in terms of their job completion times. We formulate this problem as a lexicographical minimization problem, which is challenging to solve in practice due to its inherent multi-objective and discrete nature. To address these challenges, we iteratively solve its single-objective subproblems, which can be transformed to equivalent linear programming (LP) problems to be efficiently solved, thanks to their favorable properties. As a highlight of this paper, we have designed and implemented our proposed solution as a fair job scheduler based on Apache Spark, a modern data processing framework. With extensive evaluations of our real-world implementation on Amazon EC2 and large-scale simulations, we have shown convincing evidence that max-min fairness has been achieved and the worst job completion time has been significantly improved using our new job scheduler.

Objectives: In this paper various task scheduling algorithm along with load distribution techniques investigated to ensure efficient mapping of tasks to resources and for coherent resource utilization in heterogeneous environment. Statistical Analysis: A detailed comparative approach is conferred for different job scheduling and load balancing algorithm available in the literature. Considered parameters for comparisons are scheduling approaches, techniques, findings, benefits, pros and cons. This study facilitates to identify the efficient scheduling and load balancing approach which optimizes resource use, minimizes completion time as well as task waiting time and balanced load among resources. Findings: Make span time and flow time are two important parameters to compute and minimize job execution time. This paper explores an intense survey of minimization of task execution time with minimum make span time and flow time. Apart from this a joint review on job scheduling, load balancing and minimizing job completion time using Expected Time to Compute (ETC) matrix is presented. At the end of this paper a research model is proposed which is a novel and hybrid approach with a combination of the efficiency of scheduling and load balancing. Application: The Load Balancing (LB) is an important feature of the grid environment. The efficient LB algorithm assures cost effective resource usage by the provisioning of resources to users on demand schedule in pay-as-you-say-manner.