Job Deadlines Research Articles

Job-Deadline-Guarantee-Based Joint Flow Scheduling and Routing Scheme in Data Center Networks.

Many emerging Internet of Things (IoT) applications deployed on cloud platforms have strict latency requirements or deadline constraints, and thus meeting the deadlines is crucial to ensure the quality of service for users and the revenue for service providers in these delay-stringent IoT applications. Efficient flow scheduling in data center networks (DCNs) plays a major role in reducing the execution time of jobs and has garnered significant attention in recent years. However, only few studies have attempted to combine job-level flow scheduling and routing to guarantee meeting the deadlines of multi-stage jobs. In this paper, an efficient heuristic joint flow scheduling and routing (JFSR) scheme is proposed. First, targeting maximizing the number of jobs for which the deadlines have been met, we formulate the joint flow scheduling and routing optimization problem for multiple multi-stage jobs. Second, due to its mathematical intractability, this problem is decomposed into two sub-problems: inter-coflow scheduling and intra-coflow scheduling. In the first sub-problem, coflows from different jobs are scheduled according to their relative remaining times; in the second sub-problem, an iterative coflow scheduling and routing (ICSR) algorithm is designed to alternately optimize the routing path and bandwidth allocation for each scheduled coflow. Finally, simulation results demonstrate that the proposed JFSR scheme can significantly increase the number of jobs for which the deadlines have been met in DCNs.

PED: Probabilistic Energy-efficient Deadline-aware scheduler for heterogeneous SoCs

Heterogeneous systems-on-chip (SoCs) integrate diverse cores with different performance and energy tradeoffs. Scheduling applications with soft deadline constraints is highly complex in such heterogeneous platforms, and the complexity is further exacerbated by the streaming jobs generated by applications from domains such as communication and radar systems. Existing deadline-aware schedulers typically first translate the job deadlines to task-level slacks before scheduling, which is the time available for a processing element (PE) to execute a specific task. Task-level slacks are critically dependent on the task-to-PE allocation of all other tasks from the same job (intra-job) or concurrent jobs (inter-job). However, this allocation is usually unknown before the start of the scheduling process. To address the problem, we propose PED, a probabilistic energy-efficient deadline-aware scheduler for heterogeneous SoCs. PED minimizes the average tardiness of streaming jobs with the least energy consumption by accurately predicting the task-to-PE allocation using Neural Network and considering intra- and inter-job contentions when scheduling tasks. Our extensive experimental results in a domain-specific SoC (DSSoC) designed for radar and communication domains show that PED can reduce tardiness by 6.9× with comparable energy consumption; and reduce energy consumption by 14% without any loss in tardiness, when compared with state-of-the-art schedulers.

Load Balancer using Whale-Earthworm Optimization for Efficient Resource Scheduling in the IoT-Fog-Cloud Framework

Cloud-Fog environment is useful in offering optimized services to customers in their daily routine tasks. With the exponential usage of IoT devices, a huge scale of data is generated. Different service providers use optimization scheduling approaches to optimally allocate the scarce resources in the Fog computing environment to meet job deadlines. This study introduces the Whale-EarthWorm Optimization method (WEOA), a powerful hybrid optimization method for improving resource management in the Cloud-Fog environment. Striking a balance between exploration and exploitation of these approaches is difficult, if only Earthworm or Whale optimization methods are used. Earthworm technique can result in inefficiency due to its investigations and additional overhead, whereas Whale algorithm, may leave scope for improvement in finding the optimal solutions using its exploitation. This research introduces an efficient task allocation method as a novel load balancer. It leverages an enhanced exploration phase inspired by the Earthworm algorithm and an improved exploitation phase inspired by the Whale algorithm to manage the optimization process. It shows a notable performance enhancement, with a 6% reduction in response time, a 2% decrease in cost, and a 2% improvement in makespan over EEOA. Furthermore, when compared to other approaches like h-DEWOA, CSDEO, CSPSO, and BLEMO, the proposed method achieves remarkable results, with response time reductions of up to 82%, cost reductions of up to 75%, and makespan improvements of up to 80%.

An effective fault tolerance aware scheduling using hybrid horse herd optimisation‐reptile search optimisation approach for a cloud computing environment

AbstractIT services can be requested via cloud computing, a model based on services as well as the Internet. It includes all computer systems resources from hardware components to software platforms and software applications in a distributed environment. Scheduling is a key step in processing tasks using remote resources. Serious issues have been brought forward, including the ineffective use of resources and task execution failure. The concurrent provision of fault tolerance and resource optimisation is achallenging task. In the context of cloud computing, this research offers a brand‐new job scheduling and fault‐tolerant system. Tasks submitted by users are taken as an input for the proposed method. Several virtual machines (VM) are initially arranged for scheduling work and execution process. Initially, Horse Herd Optimisation is employed here to allocate the job based on key factors such as deadline and user budget. Once the jobs are assigned to each VM, then each job's deadline is confirmed and transferred to VM which has sufficient capacity. Here, the Reptile Search Optimisation technique is applied to identify the VM error. Any VM that does not have enough capacity is the one that has a problem. When a fault is found, a fault‐tolerant process is instantly started. A replication‐based fault‐tolerant mechanism is used in this manuscript. The proposed approach is tested with several metrics which attains better performance like 80 s response time, turnaround time of 32 s, 17% resource utilisation and a success rate of 92%. Thus the designed model is the best choice for fault‐tolerant aware task scheduling.

An Instability-Resilient Renewable Energy Allocation System for a Cloud Datacenter

Renewable energy supply is a promising solution for datacenters' increasing electricity monetary cost, energy consumption and harmful gas emissions. However, due to the instability of renewable energy, insufficient renewable energy supply may lead to the use of stored energy or brown energy. To handle this problem, in this paper, we propose an instability-resilient renewable energy allocation system. We define a job's service-level-objective (SLO) as the successful running probability by only using supplied renewable energy. The system allocates jobs with the same SLO level to the same physical machine (PM) group, and powers each PM group with renewable energy generators that have probability no less than its SLO to produce the amount no less than its energy demand. We use a deep learning technique to predict the probability of producing the amount no less than each value of each renewable energy source, and predict the energy demands of each PM area. We formulate an optimization problem to match renewable energy resources with different instabilities to different PM groups for supply, and use reinforcement learning method and linear programming method to solve it. We further propose an energy-driven computing resource assignment method, which adjusts the amount of computing resource of each job based on job deadline and failure probability in each PM group, and a failure prediction based energy saving method. Real trace driven experiments show that our methods achieve much lower SLO violations, total energy monetary cost and total carbon emission compared to other methods and the effectiveness of individual methods.

Machine Learning Feature Based Job Scheduling for Distributed Machine Learning Clusters

With the rapid proliferation of Machine Learning (ML) and Deep learning (DL) applications running on modern platforms, it is crucial to satisfy application performance requirements such as meeting deadline and ensuring accuracy. To this end, researchers have proposed several job schedulers for ML clusters. However, none of the previously proposed schedulers consider ML model parallelism, though it has been proposed as an approach to increase the efficiency of running large-scale ML and DL jobs. Thus, in this paper, we propose an ML job Feature based job Scheduling system (MLFS) for ML clusters running both data parallelism and model parallelism ML jobs. MLFS first uses a heuristic scheduling method that considers an ML job’s spatial and temporal features to determine task priority for job queue ordering in order to improve job completion time (JCT) and accuracy performance. It uses the data from the heuristic scheduling method for training a deep reinforcement learning (RL) model. After the RL model is well trained, it then switches to the RL method to automatically make decisions on job scheduling. In addition, MLFS has a system load control method that selects tasks from overloaded servers to move to underloaded servers based on task priority, and also intelligently removes the tasks that generate little or no improvement on the desired accuracy performance when the system is overloaded to improve JCT and accuracy by job deadline. Furthermore, we propose Optimal ML iteration stopping method that determines the proper time to stop training ML model when this model reaches the minimum loss value. Our real experiments and large-scale simulation based on real trace show that MLFS reduces JCT by up to 53% and makespan by up to 52%, and improves accuracy by up to 64% when compared with existing ML job schedulers. We also open sourced our code.

Tight necessary feasibility analysis for recurring real-time tasks on a multiprocessor

One of the important design issues for time-critical embedded systems is to derive necessary conditions that meet all job deadlines invoked by a set of recurring real-time tasks under a computing resource (called feasibility). To this end, existing studies focused on how to derive a tight lower-bound of execution requirement (i.e., demand) of a target set of real-time tasks. In this paper, we address the following question regarding the supply provided by a multiprocessor resource: is it possible for a real-time task set to always utilize all the provided supply? We develop a systematic approach that i) calculates the amount of supply proven unusable, ii) finds a partial schedule that yields a necessary condition to minimize the amount of unusable supply, and iii) uses the partial schedule to further reclaim unusable supply. While the systematic approach can be applied to most (if not all) recurring real-time task models, we show two examples how the approach can yield tight necessary feasibility conditions for the sequential task model and the gang scheduling model. We demonstrate the proposed approach finds a number of additional infeasible task sets which have not been proven infeasible by any existing studies for the task models.

SLA-Based Scheduling of Spark Jobs in Hybrid Cloud Computing Environments

Big data frameworks such as Apache Spark is becoming prominent to perform large-scale data analytics jobs in various domains. However, due to limited resource availability, the local or on-premise computing resources are often not sufficient to run these jobs. Therefore, public cloud resources can be hired on a pay-per-use basis from the cloud service providers to deploy a Spark cluster entirely on the cloud. Nevertheless, using only cloud resources can be costly. Hence, both local and cloud resources nowadays are used together to deploy a hybrid cloud computing cluster. However, scheduling jobs in a cluster deployed on hybrid clouds is challenging in the presence of various Service-Level Agreement (SLA) demands such as cost minimization and job deadline guarantee. Most of the existing works either consider a public or a locally deployed cluster and mainly focus on improving job performance in the cluster. In this article, we propose efficient scheduling algorithms that leverage from different VM instance pricing in a hybrid cloud deployed cluster to optimize the Virtual Machine (VM) usage cost for both local and cloud resources and maximize the job deadline met percentage. We have conducted extensive simulation-based experiments to compare our proposed algorithms with the baseline approaches. In addition, we have developed a prototype system on top of Apache Mesos cluster manager and performed real experiments to evaluate the applicability of our proposed approaches in a real platform with benchmark applications. The results show that our proposed algorithms are highly scalable and reduce the cost of VM usage of a hybrid cluster for up to 20 percent.

Workplace Stressors of Deaf and Hearing Workers

The occurrence of stress in the workplace is frequent and present in all areas of work. Population deaf workers are at some risk when it comes to the impact of stress in the workplace. The research aims is to determine factors that directly cause stress in the working environment deaf and hearing workers. The sample consisted of 30 deaf workers and 33 hearing workers. For the needs of the research, the measuring instrument of the International labor organizations was used (2003). Research results show that isolated stressors factors are more common in a subsample of deaf subjects. Fear of losing a job, fatigue and short deadlines, lack of support from colleagues and managers, constant change of obligations, feeling insecurities, fear of position in the firm, and communication difficulties are isolated causes of stress in deaf workers. Workflow errors are associated with a feeling of fear and insecurity, they appear due to their insufficient information due to communication difficulties.

Statistical analysis of family based dispatching rules and preemption

Family based dispatching rules reduce time spent on setups to improve system performance. However, they can lead to bulky arrivals at subsequent stages, causing additional delays. Few studies have considered this effect so far. Preemption, on the other hand, can enforce the deadlines of important jobs in real situations. This study deals with a new production control problem of family based dispatching rules and preemption. No research has addressed both family based dispatching rules and preemption. A new framework of decision making is presented to solve this problem. The analysis of family based dispatching rules and preemption is carried out at a two-stage flow shop by simulation modeling and statistical analysis to discover several key factors. In addition, this research investigates the performance differences among different preemption types with respect to family based dispatching rules. The results of this paper can offer insight into the use of preemption, especially with family based dispatching rules.

Energy-aware adaptive offloading of soft real-time jobs in mobile edge clouds

We present a model for measuring the impact of offloading soft real-time jobs over multi-tier cloud infrastructures. The jobs originate in mobile devices and offloading strategies may choose to execute them locally, in neighbouring devices, in cloudlets or in infrastructure cloud servers. Within this specification, we put forward several such offloading strategies characterised by their differential use of the cloud tiers with the goal of optimizing execution time and/or energy consumption. We implement an instance of the model using Jay, a software framework for adaptive computation offloading in hybrid edge clouds. The framework is modular and allows the model and the offloading strategies to be seamlessly implemented while providing the tools to make informed runtime offloading decisions based on system feedback, namely through a built-in system profiler that gathers runtime information such as workload, energy consumption and available bandwidth for every participating device or server. The results show that offloading strategies sensitive to runtime conditions can effectively and dynamically adjust their offloading decisions to produce significant gains in terms of their target optimization functions, namely, execution time, energy consumption and fulfilment of job deadlines.

Preemptive Mobile Computation Offloading With Hard Deadlines and Concurrent Local Execution

This paper considers Preemptive Mobile Computation Offloading when concurrent local execution (CLE) is used to guarantee task execution time constraints. By allowing simultaneous local and remote execution, CLE ensures that job deadlines are always satisfied in the face of unforeseen wireless channel conditions. In the preemptive offloading case, at the start of each time slot, a decision is made to either continue or temporarily interrupt the offload. This mechanism allows the system to adapt when channel conditions change. The paper considers the case for homogeneous Markovian wireless channels. Using Markovian decision process stopping theory, an online energy-optimal computation offloading algorithm is formulated for preemptive offloading, referred to as Optimal Preemptive Offloading (OPO). Since the computational complexity of OPO can be prohibitive, the paper introduces three computationally efficient techniques motivated by OPO: 1) water-filling; 2) water-filling with scheduling; and 3) generalized water-filling. For each method, two variations are considered. The first (Equ) uses the equilibrium channel state probabilities in offloading decision calculations, and the second (Exp) uses Markovian transition matrix exponentiation. This results in six algorithms with a wide variety of energy performance and computational complexity. Performance of the algorithms is compared that shows the tradeoffs between complexity and mobile energy saving performance.

Energy-aware offloading based on priority in mobile cloud computing

Smartphones and portable devices have been widely used in our daily life. However, these portable devices cannot be used in a lot of environments due to limitations in battery capacity and computing speed. Therefore, how to save energy consumption and improve processing ability has become a hot issue. The previous works were mainly to transfer some local tasks to remote compute nodes by intelligently selecting the execution route of tasks based on Directed Acyclic Graph (DAG). Different from previous work, this work mainly attempts to consider task urgency and system load based on Dynamic Voltage and Frequency Scaling (DVFS) model. Firstly, tasks are given different priorities according to the energy consumption of different locations and task urgency. Secondly, a reference working state is given according to system load to ensure that each task (regardless of job deadline) is completed with minimum energy consumption under the average system load. Finally, a heuristic algorithm is used to schedule tasks on mobile devices and remote cloud resources based on the priority of tasks and reference status of resources. Our method is called AODVFS-Adaptive Offloading based on DVFS model. Comparison with other methods shows that our method saves average energy consumption and increases the number of completed tasks.

Fault Tolerant List Scheduler for Time-Triggered Communication in Time-Sensitive Networks

The performance and dependability of modern mission-critical systems significantly depends on the communication infrastructure. In this context, the Time-Sensitive Networking (TSN) task group addressed different requirements of cyber-physical systems such as timing and reliability constraints. TSN provides real-time capabilities through sharing a global time reference and employing transmission schedule tables called Gate Control Lists (GCL). On the other hand, TSN masks faulty behaviors within a system through a technique called Frame Replication and Elimination for Reliability (FRER). FRER fulfills the safety requirements of mission-critical systems by message replication and transmission of message replicas over redundant paths. The scheduling problem for the GCL synthesis is NP-complete. For simplification of the scheduling process, several state-of-the-art solutions provide schedulers for fault-free networks. However, this assumption is very optimistic and in practice networks experience different faulty-behaviors over time. This paper extends our heuristic TSN scheduler which was developed for fault-free TSN systems to support the FRER mechanism. Our fault-tolerant TSN scheduler focuses on enhancing the reliability of a mission-critical system while meeting the deadlines of time-critical jobs. To achieve this goal, we introduce a novel reliability analysis approach for a mission-critical system with a TSN communication infrastructure. This approach models and evaluates the reliability of a system based on the reliability of message transmissions between safety-critical jobs. The reliability of message transmissions is computed based on the reliability of the network components that form the forwarding paths. Thereby, our reliability model enables the system designers to plan networks more optimally.

Batch Auction Design for Cloud Container Services

Cloud containers represent a new, light-weight alternative to virtual machines in cloud computing. A user job may be described by a container graph that specifies the resource profile of each container and container dependence relations. This work is the first in the cloud computing literature that designs efficient market mechanisms for container based cloud jobs. Our design targets simultaneously incentive compatibility, computational efficiency, and economic efficiency. It further adapts the idea of batch online optimization into the paradigm of mechanism design, leveraging agile creation of cloud containers and exploiting delay tolerance of elastic cloud jobs. The new and classic techniques we employ include: (i) compact exponential optimization for expressing and handling non-traditional constraints that arise from container dependence and job deadlines; (ii) the primal-dual schema for designing efficient approximation algorithms for social welfare maximization; and (iii) posted price mechanisms for batch decision making and truthful payment design. Theoretical analysis and trace-driven empirical evaluation verify the efficacy of our container auction algorithms.

PowerCoord: Power capping coordination for multi-CPU/GPU servers using reinforcement learning

Modern supercomputers and cloud providers rely on server nodes that are equipped with multiple CPU sockets and general purpose GPUs (GPGPUs) to handle the high demand for intensive computations. These nodes consume much higher power than commodity servers, and integrating them with power capping systems used in modern clusters presents new challenges. In this paper, we propose a new power capping controller, PowerCoord, that is specifically designed for servers with multiple CPU and GPU sockets that are running multiple jobs at a time. PowerCoord coordinates among the various power domains (e.g., CPU sockets and GPUs) inside a node server to meet target power caps, while seeking to maximize throughput. Our approach also takes into consideration job deadlines and priorities. Because performance modeling for co-located jobs is error-prone, PowerCoord uses a learning method. PowerCoord has a number of heuristic policies to allocate power among the various CPUs and GPUs, and it uses reinforcement learning for policy selection during runtime. Based on the observed state of the system, PowerCoord shifts the distribution of selected policies. We implement our power cap controller on a real multi-CPU/GPU server with low overhead, and we demonstrate that it is able to meet target power caps while maximizing the throughput, and balancing other demands such as priorities and deadlines. Our results show PowerCoord improves the server throughput on average by 18% compared with the case when power is not coordinated among CPU/GPU domains. Also, PowerCoord improves the server throughput on average by 11% compared with prior work that uses a heuristic approach to coordinate the power among domains.

CPU–GPU Utilization Aware Energy-Efficient Scheduling Algorithm on Heterogeneous Computing Systems

Nowadays, heterogeneous computing systems have proven to be a good solution for processing computation intensive high-performance applications. The main challenges for such large-scale systems are energy consumption, computing node CPU-GPU utilization dynamic variability, and so on. In response to these challenges, this study first provides heterogeneous computing systems architecture and parallel application job model. Then, we build system computing node CPU-GPU utilization model and analyze job execution energy consumption. We also deduce the optimal CPU-GPU utilization and job deadline scheduling constraint. Third, we propose a systems CPU-GPU utilization aware and energy-efficient heuristic greedy strategy (UEJS) to solve this job scheduling problem. To improve the algorithm global optimization ability, we design a hybrid particle swarm optimization algorithm (H-PSO), which incorporates the heuristic greedy strategy into the bio-inspired search optimization technique. The rigorous experimental results clearly demonstrate that our proposed H-PSO outperforms heuristic greedy strategy, Max-EAMin, and Genetic Algorithm in terms of the average energy consumption of jobs and system job rejection rate. In particular, H-PSO is better than UEJS by 36.5%, Max-EAMin by 36.3%, and GA by 46.7% in term of the job average energy consumption for heterogeneous system with high workload.

High Performance Fault Tolerant Resource Scheduling in Computational Grid Environment

Virtual resources team up to create a computational grid, which is used in computation-intensive problem solving. A majority of these problems require high performance resources to compute and generate results, making grid computation another type of high performance computing. The optimization in computational grids relates to resource utilization which in turn is achieved by the proper distribution of loads among participating resources. This research takes up an adaptive resource ranking approach, and improves the effectiveness of NDFS algorithm by scheduling jobs in those ranked resources, thereby increasing the number of job deadlines met and service quality agreements met. Moreover, resource failure is taken care of by introducing a partial backup approach. The benchmark codes of Fast Fourier Transform and Matrix Multiplication are executed in a real test bed of a computational grid, set up by Globus Toolkit 5.2 for the justification of propositions made in this article.

Energy efficient VM scheduling strategies for HPC workloads in cloud data centers

In virtualized environments, virtual machine (VM) scheduling strategies incorporated with energy efficient techniques are needed to reduce the operational cost of the system while delivering high Quality of Service (QoS). It is widely accepted that the cost of the energy consumption in the environment is a dominant part of the owner’s budget. However, when considering energy efficiency, VM scheduling decisions become more constrained, leading in the violation of job deadlines and hence compromising QoS. This paper studies energy efficient VM scheduling strategies in virtualized environments to minimize the queue time and makespan under the fulfillment of SLA requirements (i.e., deadline). Specifically, six energy efficient VM scheduling strategies are investigated incorporated with the dynamic voltage and frequency scaling (DVFS) power management technique. They consist of user-oriented and system-oriented policies. The strategies are extensively simulated and compared with three power management governing methods provided at hypervisor level (i.e., userspec, ondemand, and performance). To conduct simulation experiments, we employ real-world high performance computing (HPC) workloads collected from a production data center. For comparison and evaluation, we analyze the: (a) energy consumption, (b) runtime, (c) queue time, (d) makespan, and (e) slowdown ratio. Lastly, we highlight the strengths and weaknesses of the VM scheduling strategies that can help to choose the most appropriate VM scheduling strategy for a given scenario.

Deadline-Aware MapReduce Job Scheduling with Dynamic Resource Availability

As MapReduce is becoming ubiquitous in large-scale data analysis, many recent studies have shown that the performance of MapReduce could be improved by different job scheduling approaches, e.g., Fair Scheduler and Capacity Scheduler. However, most exiting MapReduce job schedulers focus on the scenario that MapReduce cluster is stable and pay little attention to the MapReduce cluster with dynamic resource availability. In fact, MapReduce cluster resources may fluctuate as there is a growing number of Hadoop clusters deployed on hybrid systems, e.g., infrastructure powered by mix of traditional and renewable energy, and cloud platforms hosting heterogeneous workloads. Thus, there is a growing need for providing predictable services to users who have strict requirements on job completion times in such dynamic environments. In this paper, we propose, RDS , a Resource and Deadline-aware Hadoop job Scheduler that takes future resource availability into consideration when minimizing job deadline misses. We formulate the job scheduling problem as an online optimization problem and solve it using an efficient receding horizon control algorithm. To aid the control, we design a self-learning model to estimate job completion times. We further extend the design of RDS scheduler to support flexible performance goals in various dynamic clusters. In particular, we use flexible deadline time bounds instead of the single fixed job completion deadline. We have implemented RDS in the open-source Hadoop implementation and performed evaluations with various benchmark workloads. Experimental results show that RDS substantially reduces the penalty of deadline misses by at least 36 and 10 percent compared with Fair Scheduler and Earliest Deadline First (EDF) scheduler, respectively. In a Hadoop cluster running partially on renewable energy, the experimental result shows the green power based resource prediction approach can further reduce the penalty of deadline misses by 16 percent compared to Auto-Regressive Integrated Moving Average (ARIMA) prediction approach.