Job Completion Time Research Articles

Ant Colonized and Taguchi Parallel Scheduliing with Sequence Independent Setup Time

The Job Shop Parallel Machine Scheduling (JSPMS) is a hybrid production system, and hence has received significant attention in the past few years. The JSPMS problem is a rationalization of the traditional job shop scheduling problem in computer science and operation research that permits to process operations on single machine out of a set of possible parallel machines. To maximize the job completion rate and minimize job completion time, a hybrid production system is necessary. With this objective, a novel meta-heuristic method is designed. This paper develops a scheduling method called, Ant Colonized and Taguchi Parallel Operation Scheduling (ACTPOS), for JSPMS, aimed to minimize job completion time and maximize job completion rate. The design of AC-TPOS method involves two different models, namely, Ant Colonized Parallel Machine Selection (ACPMS) model and Taguchi Parallel Operation Scheduling (TPOS) model. In ACPMS model, optimal selection of machine is done via operation being processed by parallel machines using local pheromone updating rule concentrating on the makespan time. In addition, the processing time and sequence-independent setup time are considered. Next, in TPOS model, optimal scheduling of operation is performed using Taguchi method concentrating on the makespan rate. Finally, the test results first show that our algorithm outperforms existing methods in terms of job completion rate, job completion time and computational complexity involved in scheduling operations.

A three-agent scheduling problem for minimizing the flow time on two machines

This study introduces a two-machine three-agent scheduling problem. We aim to minimize the total tardiness of jobs from agent 1 subject to that the maximum completion time of jobs from agent 2 cannot exceed a given limit and that two maintenance activities from agent 3 must be conducted within two maintenance windows. Due to the NP-hardness of this problem, a genetic algorithm (named GA+) is proposed to obtain approximate solutions. On the other hand, a branch-and-bound algorithm (named B&B) is developed to generate the optimal solutions. When the problem size is small, we use B&B to verify the solution quality of GA+. When the number of jobs is large, a relative deviation is proposed to show the gap between GA+ and another ordinary genetic algorithm. Experimental results show that the proposed genetic algorithm can generate approximate solutions by consuming reasonable execution time.

Improvement of job completion time in data-intensive cloud computing applications

Task stragglers in MapReduce jobs dramatically impede job execution of data-intensive computing in cloud data centers. This impedance is due to the uneven distribution of input data, heterogeneous data nodes, resource contention situations, and network configurations. Data skew of intermediate data in MapReduce job causes delay failures due to the violation of job completion time. Data-intensive computing frameworks, such as MapReduce or Hadoop YARN, employ HashPartitioner. This partitioner may cause intermediate data skew, which results in straggler reducers. In this paper, we strive to make Hadoop YARN more efficient in cloud environments. We present, a new partitioning scheme, called balanced data clusters partitioner (BDCP), to handle straggler Reduce tasks based on sampling of input data and feedback information about the current processing task. Our extensive experimental results show that BDCP can outperform the default Hadoop HashPartitioner and Range partitioner. BDCP can assist in straggler mitigation during reduce phase and minimize the job completion time in MapReduce jobs within data-intensive cloud computing.

On the exact solution of a large class of parallel machine scheduling problems

This work deals with a very generic class of scheduling problems with identical/uniform/unrelated parallel machine environment. It considers well-known attributes such as release dates, deadlines, or sequence-dependent setup times and accepts any objective function defined over job completion times. Non-regular objectives are also supported. We introduce a branch-cut-and-price algorithm for such problems that makes use of non-robust cuts, i.e., cuts which change the structure of the pricing problem. This is the first time that such cuts are employed for machine scheduling problems. The algorithm also embeds other important techniques such as strong branching, reduced cost fixing and dual stabilization. Computational experiments over literature benchmarks showed that the proposed algorithm is indeed effective and could solve many instances to optimality for the first time.

Solving Permutation Flow Shop Scheduling Problem with Sequence-Independent Setup Time

In this paper, we study the resolution of a permutation flow shop problem with sequence-independent setup time. The objective is to minimize the maximum of job completion time, also called the makespan. In this contribution, we propose three methods of resolution, a mixed-integer linear programming (MILP) model; two heuristics, the first based on Johnson’s rule and the second based on the NEH algorithm; and finally two metaheuristics, the iterative local search algorithm and the iterated greedy algorithm. A set of test problems is simulated numerically to validate the effectiveness of our resolution approaches. For relatively small-size problems, it has been revealed that the adapted NEH heuristic has the best performance than that of the Johnson-based heuristic. For the relatively medium and large problems, the comparative study between the two metaheuristics based on the exploration of the neighborhood shows that the iterated greedy algorithm records the best performances.

Ant colony optimization algorithm for total weighted completion time minimization on non-identical batch machines

This paper aims at the problem of scheduling a set of jobs with arbitrary sizes on parallel batch processing machines with arbitrary capacities. The optimization objective is to minimize the total weighted completion time of jobs. After describing the studied problem, we analyze its complexity and present a lower bound of the problem. A heuristic is provided to solve the problem firstly. Then, with the proposed first job selection strategy based on the weights of jobs, two algorithms based on the ant system and the max-min ant system, respectively, are designed to address the problem. Through extensive experiments, the performance of the proposed algorithms is compared with several state-of-the-art algorithms. The comparative results verify effectiveness and efficiency of the proposed algorithms.

Joint Optimization of MapReduce Scheduling and Network Policy in Hierarchical Data Centers

As large-scale data analytic becomes norm in various industries, using MapReduce frameworks to analyze ever-increasing volumes of data will keep growing. In turn, this trend drives up the intention to move MapReduce into multi-tenant clouds. However, the application performance of MapReduce can be significantly affected by the time-varying network bandwidth in a shared cluster. Although many recent studies improve MapReduce performance by dynamic scheduling to reduce the shuffle traffic, most of them do not consider the impact by widely existing hierarchical network architectures in data centers. In this paper, we propose and design a Hierarchical topology (Hit) aware MapReduce scheduler to minimize overall data traffic cost and hence to reduce job execution time. We first formulate the problem as a Topology Aware Assignment (TAA) optimization problem while considering dynamic computing and communication resources in the cloud with hierarchical network architecture. We further develop a synergistic strategy to solve the TAA problem by using the stable matching theory, which ensures the preference of both individual tasks and hosting machines. Finally, we implement the proposed scheduler as a pluggable module on Hadoop YARN and evaluate its performance by testbed experiments and simulations. The testbed experimental results show Hit-scheduler can improve job completion time by 28% and 11% compared to Capacity Scheduler and Probabilistic Network-Aware scheduler, respectively. Our simulations further demonstrate that Hit-scheduler can reduce the traffic cost by 38% at most and the average shuffle flow traffic time by 32% compared to Capacity scheduler. In this manuscript, we have extended Hit-scheduler to a decentralized heuristic scheme to perform the policy-aware allocation in data center environments. Many existing centralized approximation approaches are too complex and infeasible to implement over a data center, which typically include large amounts of servers, containers, switches and traffic flows. In the extension, we have designed a decentralized heuristic scheme to perform the Policy-Aware Task (PAT) allocation by using existing centralize algorithm to approximately maximize the total gained utility. Finally, the simulation based experimental results show that the proposed PAT policy reduces the communication cost by 33.6% compared with the default scheduler in data centers.

Interference-aware parallelization for deep learning workload in GPU cluster

With the widespread use of GPUs for performing deep learning applications, the issue of efficient execution of multiple deep learning jobs in a GPU cluster has attracted great attention. It becomes more difficult to achieve efficient workloads parallelization since modern GPUs support concurrent execution of multiple jobs. However, traditional coarse-grained scheduling methods without taking into account interference caused by resource contention among co-executing jobs and characteristics of deep learning jobs can lead to unbalanced use of computing resource and further cause the degradation of jobs performance in the GPU cluster. In this paper, we propose a two-stage workload parallelization approach for deep learning training workloads. We firstly propose two interference-aware prediction models including the Interference-Aware Similarity Prediction (IASP) model based on deep collaborative filtering and the Interference-Aware Performance Prediction (IAPP) model based on deep neural network. Our parallelization approach includes both the cluster-level workload parallelization strategy and the node-level workload parallelization strategy. Specifically, the Cluster-Level Workload Parallelization (CLWP) strategy assigns deep learning jobs to appropriate worker node according to the proposed IASP model, and the Node-Level Workload Parallelization (NLWP) strategy places deep learning tasks to appropriate GPUs according to the proposed IAPP model and the communication costs among tasks. We evaluate our deep learning workload parallelization strategy on a prototype platform with other widely used methods. The experimental results show that the proposed strategy can averagely improve the GPU utilization by 18% and reduce the job completion time by around 22%.

Analyzing Software Rejuvenation Techniques in a Virtualized System: Service Provider and User Views

Virtualization technology has promoted the fast development and deployment of cloud computing, and is now becoming an enabler of Internet of Everything. Virtual machine monitor (VMM), playing a critical role in a virtualized system, is software and hence it suffers from software aging after a long continuous running as well as software crashes due to elusive faults. Software rejuvenation techniques can be adopted to reduce the impact of software aging. Although there existed analytical model-based approaches for evaluating software rejuvenation techniques, none analyzed both application service (AS) availability and job completion time in a virtualized system with live virtual machine (VM) migration. This paper aims to quantitatively analyze software rejuvenation techniques from service provider and user views in a virtualized system deploying VMM reboot and live VM migration techniques for rejuvenation, under the condition that all the aging time, failure time, VMM fixing time and live VM migration time follow general distributions. We construct an analytical model by using a semi-Markov process (SMP) and derive formulas for calculating AS availability and job completion time. By analytical experiments, we can obtain the optimal migration trigger intervals for achieving the approximate maximum AS availability and the approximate minimum job completion time, and then service providers can make decisions for maximizing the benefits of service providers and users by adjusting parameter values.

Coordinative Scheduling of Computation and Communication in Data-parallel Systems

For many data-parallel computing systems like Spark, a job usually consists of multiple computation stages and inter-stage communication (i.e., coflows). Many efforts have been done to schedule coflows and jobs independently. The simple combination of coflow scheduling and job scheduling, however, would prolong the average job completion time (JCT) due to the conflict. For this reason, we propose a new abstraction of scheduling unit, named coBranch, which takes the dependency between computation stages and coflows into consideration, to schedule coflows and jobs jointly. Besides, mainstream coflow schedulers are order-preserving, i.e., all coflows of a high-priority job are prioritized than those of a low-priority job. We observe that the order-preserving constraint incurs low inter-job parallelism. To overcome the problem, we employ an urgency-based mechanism to schedule coBranches, which aims to decrease the average JCT by enhancing the inter-job parallelism. We implement the urgency-based coBranch Scheduling (BS) method on Apache Spark, conduct prototype-based experiments, and evaluate the performance of our method against the shortest-job-first critical-path method and the FIFO method. Results show that our method achieves around 10 and 15 percent reduction in the average JCT, respectively. Large-scale simulations based on the Google trace show that our method performs better and reduces JCT by 23 and 35 percent, respectively.

Hawkeye: Adaptive Straggler Identification on Heterogeneous Spark Cluster With Reinforcement Learning

It is a common sense that people harbors the belief that stragglers exert huge influence upon the performance conducted by the analysis systems of big data for the reason of poor performance made by some computing nodes, data skew and so on. Accordingly, stragglers have been billed as an indispensable bottleneck in Map-Reduce framework processing. However, existing studies on stragglers identification are targeting coarse-grained detection, schedule level optimization and off-line log based cause analysis. Accuracy identifying the stragglers in time for each job, however, is an extremely tough because (1) Auite a number of root causes for stragglers in data analytics frameworks;(2) The number of key parameters affecting stragglers identification; and (3) The different clusters configurations, and their impact on the stragglers detection, vary among different job types and sizes. Either existing solutions adopt a “tweak-and-pray” manual tuning approach, which is complex, time-consuming and error-prone, or only most of them fix theirs eyes upon coarse-grained straggler detection. In this paper, we systematically conduct the exploration on the fundamental problem of automatic, adaptive straggler identification on big data analytics platform. Under the inspiration of the recent triumphs over implementing Reinforcement Learning (RL) techniques for solving complex online optimal problems, we conducted investigation that Reinforcement learning are reasonably employed to adaptively opt the optimal parameters to identify stragglers free of the intervention of human beings. Specifically, we propose Hawkeye, a general adaptive speculative execution system which identifies stragglers by reinforcement are learning to launch speculative tasks on heterogeneous cluster at runtime. In accordance with the experimental conclusions, Hawkeye manages to cut down the job completion time over the distinct type applications. An instance is that it reveals as many as nearly 37% decrease average job completion time based on an improvement of 23% on the preciseness of the present resolutions to the heterogeneous cluster.

Joint Workload Distribution and Capacity Augmentation in Hybrid Datacenter Networks

In hybrid datacenter networks, wired connections are augmented with wireless links to facilitate data transfers between racks. The usage of mmWave/FSO wireless links enables dynamic bandwidth/capacity allocation with extremely small reconfiguration delay. Also, on-demand workload distribution, where the workload of a job is divided into multiple tasks that can be distributed/routed to different racks to be processed in parallel, allows better utilization of computational resources in data centers. In prior work, the dynamic wireless capacity augmentation and workload distribution decisions were mostly made independently and in a heuristic manner for serving distributed and parallel computing jobs. In this paper, we propose a novel analytical framework and algorithms to jointly optimize both the wireless capacity augmentation and the workload distribution, to minimize the job completion time. We consider workload that is not amenable to pipelining, fully amenable to pipelining, and partially amenable to pipelining. With extensive simulation studies, we show that the gain (in terms of the reduction in the job completion time) can be very substantial when allowing such joint optimization.

LSTPD: Least Slack Time-Based Preemptive Deadline Constraint Scheduler for Hadoop Clusters

Big data refers to numerous forms of complex and large datasets which need distinctive computational platforms in order to be analyzed. Hadoop is one of the popular frameworks for analytics of big data. In Hadoop, a big job is split into multiple small tasks and then they are distributed to worker nodes in a parallel way using MapReduce to speed up computational processes. In this aspect, it is important how to improve throughput performance. MapReduce jobs require quick responses from the worker nodes to complete them under their deadlines. The existing scheduling schemes for Hadoop such as FIFO, fair, and capacity schedulers cannot guarantee the quick response requirement satisfying a prior deadline. Thus, Hadoop system needs to improve response time and completion time for the heterogeneous MapReduce jobs. In this paper, we propose an efficient preemptive deadline constraint scheduler based on least slack time and data locality. In order for better allocation of tasks and load balancing, we first analyze the task scheduling behaviors of the Hadoop platform. Based on that, we propose a novel preemptive approach which considers the remaining execution time of the job being executed in deciding preemption. The experimental results show that the proposed scheme significantly reduces the job execution time and queue waiting time, compared to existing schemes.

Branch and Bound algorithm for identical parallel machine scheduling problem to maximize system availability

In the majority of production scheduling studies, the objective is to minimise a criterion which is generally, function of completion times of production jobs. However, for some manufacturing systems, the reliability/availability of machines can be the most important performance criteria towards decision makers. In this paper, we deal with a production scheduling problem on identical parallel machines and the objective is to find the best assignment of jobs on machines maximising the system availability. We assume that the production system can be subject to potentially costly failures then PM actions are performed at the end of production jobs. We have proposed a branch and bound algorithm, dominance rules and an efficient upper bound to solve the proposed model optimally. Computational experiments are carried out on randomly generated test problems and results show the efficiency of the proposed upper bound and dominance rules. [Submitted 23 December 2016; Accepted 27 October 2018]

Using social network analysis for industrial plant layout analysis in the context of industry 4.0

Social network analysis (SNA) is a widely studied research topic, which has been increasingly applied for solving different kinds of problems, including industrial manufacturing ones. This paper focuses on the application of SNA to an industrial plant layout problem. The study aims at analysing the importance of using SNA techniques to study the important relations between entities in a manufacturing environment, such as jobs and resources in the context of industrial plant layout analysis. Here, performance measures such as maximum completion time of jobs (makespan), resource utilisation, and throughput time have been considered to evaluate the system performance. Later, with the simulation analysis, the relationships between entities and their impact on the system performance are evaluated. The experimental results revealed that the proposed SNA approach supports to find the key machines of the systems that ultimately lead to the effective performance of the whole system. Finally, the identification of relations among these entities supported the establishment of an appropriate plant layout for producing the jobs in the context of industry 4.0.

A two-phase simulated annealing algorithm to minimise the completion time variance of jobs in flowshops

In this paper, simulated annealing (SA) algorithm is employed to solve the flowshop scheduling problem with the objective of minimising the completion time variance (CTV) of jobs. Four variants of the two-phase SA algorithm (SA-I to SA-IV) are proposed to solve flowshop scheduling problem with the objective of minimising the CTV of jobs without considering the right shifting of completion time of jobs on the last machine. The proposed SA algorithms have been tested on 90 benchmark flowshop scheduling problems. The solutions yielded by the proposed SA variants are compared with the best CTV of jobs reported in the literature, and the proposed SA-III is found to perform well in minimising the chosen performance measure (CTV) particularly in medium and large size problems.

An Effective Heuristic Method to Minimize Makespan and Flow Time in a Flow Shop Problem

In this paper, it is presented a heuristic method for solving the multi-objective flow shop problem. The work carried out considers the simultaneous optimization of the makespan and the flow time; both objectives are essential in measuring the production system's performance since they aim to reduce the completion time of jobs, increase the efficiency of resources, and reduce waiting time in queue. The proposed method is an adaptation of multi-objective Newton's method, which is applied to problems with functions of continuous variables. In this adaptation, the method seeks to improve a sequence of jobs through local searches recursively. The computational experiments show the potential of the proposed method to solve medium-sized and large instances compared with other existing literature methods.

A novel approach for improving data locality of MapReduce applications in cloud environment through intelligent data placement

In this world of big data, hosting storage and analytics as cloud service is extremely relevant. In multi-user environments, there are chances for load imbalance during data placement. MapReduce like frameworks move computation towards data. However, because of load imbalance, some nodes cannot start computation on the node on which data is stored and may be compelled to start computation on some other nodes. This results in deteriorating data locality. In this case, data have to be copied to the computing node. This data transfer increases the job completion time. This paper proposes a data placement policy for clouds in which the data and virtual machines are collocated in the same set of physical servers. The physical servers in the cloud are grouped into partitions created using the minimum spanning tree. Experimental results show that this proposal improves node utilisation and reduces execution time over default placement in the cloud environment.

A Game-Theoretic Analysis of Shared/Buy-in Computing Systems

High performance computing clusters are increasingly operating under a shared/buy-in paradigm. Under this paradigm, users choose between two tiers of services: shared services and buy-in services. Shared services provide users with access to shared resources for free, while buy-in services allow users to purchase additional buy-in resources in order to shorten job completion time. An important feature of shared/buy-in computing systems consists of making unused buy-in resources available to all other users of the system. Such a feature has been shown to enhance the utilization of resources. Alongside, it creates strategic interactions among users, hence giving rise to a non-cooperative game at the system level. Specifically, each user is faced with the questions of whether to purchase buy-in resources, and if so, how much to pay for them. Under quite general conditions, we establish that a shared/buy-in computing game yields a unique Nash equilibrium, which can be computed in polynomial time. We provide an algorithm for this purpose, which can be implemented in a distributed manner. Moreover, by establishing a connection to the theory of aggregative games, we prove that the game converges to the Nash equilibrium through best response dynamics from any initial state. We justify the underlying game-theoretic assumptions of our model using real data from a computing cluster, and conduct numerical simulations to further explore convergence properties and the influence of system parameters on the Nash equilibrium. In particular, we point out potential unfairness and abuse issues and discuss solution venues.

A novel approach for improving data locality of MapReduce applications in cloud environment through intelligent data placement

In this world of big data, hosting storage and analytics as cloud service is extremely relevant. In multi-user environments, there are chances for load imbalance during data placement. MapReduce like frameworks move computation towards data. However, because of load imbalance, some nodes cannot start computation on the node on which data is stored and may be compelled to start computation on some other nodes. This results in deteriorating data locality. In this case, data have to be copied to the computing node. This data transfer increases the job completion time. This paper proposes a data placement policy for clouds in which the data and virtual machines are collocated in the same set of physical servers. The physical servers in the cloud are grouped into partitions created using the minimum spanning tree. Experimental results show that this proposal improves node utilisation and reduces execution time over default placement in the cloud environment.