Job Queue Research Articles

Q-Learning-Based Priority Dispatching Rule Preference Model for Non-Permutation Flow Shop

Non-permutation flow shop scheduling (NPFS) is an extension of the traditional permutation flow shop scheduling, with a broader solution space. The effectiveness of reinforcement learning in solving flow shop scheduling problems has been demonstrated through its powerful combinatorial optimization capabilities. However, the design and training of the end-to-end policy network is complex, leading to long online training time and limited adaptability of offline training. To overcome these problems, we introduced a NPFS dynamic decision-making process and then proposed a novel NPFS method that combines the Q-learning algorithm with the priority dispatching rule (PDR) set. The NPFS dynamic decision-making process involves decomposing the entire process into multiple sub-job queues for scheduling. The PDR demonstrates better scheduling performance when applied to smaller job queues. By utilizing the Q-learning algorithm, PDRs with superior performance are assigned to sub-scheduling queues based on the generation process of sub-job sequences, resulting in an optimized NPFS strategy. The limited number of PDRs in the PDR set and the small number of sub-job queues in the NPFS process contribute to the efficiency of Q-learning in solving NPFS problem. Finally, we demonstrate the superiority of the proposed NPFS method by a series of numerical experiments using the machine-based assignment PDR method and NSGA-II algorithms as performance benchmarks.

Towards a Low Carbon Proof-of-work Blockchain

Proof of Work (PoW) blockchains burn a lot of energy. Proof-of-work algorithms are expensive by design and often only serve to compute blockchains. In some sense, carbon-based and non-carbon based regional electric power is fungible. So the total carbon and non-carbon electric power mix plays a role. Thus, generally PoW algorithms have large CO2 footprints solely for computing blockchains. A proof of technology is described towards replacing hashcash or other PoW methods with a lottery and proof-of-VM (PoVM) emulation. PoVM emulation is a form of PoW where an autonomous blockchain miner gets a lottery ticket in exchange for providing a VM (virtual Machine) for a specified period. These VMs get their jobs from a job queue. Managing and ensuring, by concensus, that autonomous PoVMs are properly configured and running as expected gives several gaps for a complete practical system. These gaps are discussed. Our system is similar to a number of other blockchain systems. We briefly survey these systems. This paper along with our proof of technology was done as a senior design project.

Iterative Reconstruction of Micro Computed Tomography Scans Using Multiple Heterogeneous GPUs.

Graphics processing units (GPUs) facilitate massive parallelism and high-capacity storage, and thus are suitable for the iterative reconstruction of ultrahigh-resolution micro computed tomography (CT) scans by on-the-fly system matrix (OTFSM) calculation using ordered subsets expectation maximization (OSEM). We propose a finite state automaton (FSA) method that facilitates iterative reconstruction using a heterogeneous multi-GPU platform through parallelizing the matrix calculations derived from a ray tracing system of ordered subsets. The FSAs perform flow control for parallel threading of the heterogeneous GPUs, which minimizes the latency of launching ordered-subsets tasks, reduces the data transfer between the main system memory and local GPU memory, and solves the memory-bound of a single GPU. In the experiments, we compared the operation efficiency of OS-MLTR for three reconstruction environments. The heterogeneous multiple GPUs with job queues for high throughput calculation speed is up to five times faster than the single GPU environment, and that speed up is nine times faster than the heterogeneous multiple GPUs with the FIFO queues of the device scheduling control. Eventually, we proposed an event-triggered FSA method for iterative reconstruction using multiple heterogeneous GPUs that solves the memory-bound issue of a single GPU at ultrahigh resolutions, and the routines of the proposed method were successfully executed on each GPU simultaneously.

The Saturated Multiserver Job Queuing Model with Two Classes of Jobs: Exact and Approximate Results

We study a Multiserver Job Queuing Model (MJQM), i.e., a FIFO (First-In-First-Out) queue in which jobs request varying numbers of servers. Our investigation focuses on the saturation analysis of MJQM with two job classes. Jobs in the first class need only one server for an exponentially distributed service time, while jobs in the second class require either all servers or a fraction of them. We provide exact results for the scenario in which jobs in the second class require all servers and have generally distributed service times. Additionally, we develop an accurate approximate model of the MJQM behavior when jobs in the second class require a fraction of the available servers, and have exponentially distributed service times.

Exploring Real-Time Sentiment Analysis Prototype for Retail Industry

The rise of social media platforms has revolutionized the way consumers interact with retailers and express their opinions on products and services. Online retailers particularly need to keep a close eye on customer sentiment in real-time to make informed decisions about their offerings and improve customer satisfaction. However, efficiently analysing large volumes of unstructured text data from social media in real-time poses a significant challenge. This research aimed to develop a scalable, real-time sentiment analysis system tailored for online retailers using Reddit as the data source. The system comprises three main components: a data extraction and streaming pipeline, a sentiment analysis model, and a web application with real-time analytics. To address the data extraction challenge, a job queue-based system was implemented using Node.js, ‘BullMQ’, and Redis to create and manage campaigns for data streaming from Reddit. The data was streamed using Kafka, a distributed streaming platform, to enable efficient real-time processing. The sentiment analysis model was developed using a Naive Bayes classifier after experimenting with other machine learning and deep learning techniques. In the conducted study, the sentiment analysis model's performance was evaluated using standard metrics tailored to the context of online retail sentiment analysis. An accuracy of 0.6737 was achieved, reflecting the model's ability to correctly classify approximately 67.37 per cent of the sentiments in the test data. Concurrently, an F1 score of 0.7894 was recorded and the Area Under the Curve (AUC) value on the test data was measured at 0.5468, a metric that, while acceptable, suggests room for further refinement in the model's discriminatory ability between classes. The integration of the Data Version Control (DVC) system provided a mechanism for fine-tuning the model according to specific data requirements of various tenants. These results, taken together, not only validate the feasibility of employing a Naive Bayes classifier for real-time sentiment analysis in the retail context, but also provide a baseline for future research aimed at enhancing both the accuracy and efficiency of sentiment classification. The project’s evaluation focused on the performance of the sentiment analysis model, the efficiency of the Kafka streaming and real-time Spark pre-processing pipeline, and the backend infrastructure, including the job queuing system and WebSocket implementation. Various evaluation techniques, such as graphs and literature comparisons, were used to assess the system’s performance. In conclusion, this project successfully demonstrated the feasibility of a scalable, real-time sentiment analysis system for online retailers using Reddit data. The system has the potential to help retailers better understand customer opinions and make data-driven decisions for their businesses. Future work could include exploring alternative data sources, experimenting with more advanced sentiment analysis techniques, and enhancing the web application’s user interface and analytics capabilities.

The saturated Multiserver Job Queuing Model with two classes of jobs: Exact and approximate results

We consider a multiserver queue where jobs request for a varying number of servers for a random service time. The requested number of servers is assigned to each job following a First-In First-Out (FIFO) order. When the number of free servers is not sufficient to accommodate the next job in line, that job and any subsequent jobs in the queue are forced to wait. As a result, not all available servers are allocated to jobs if the next job requires more servers than are currently free. This queuing system is often called a Multiserver Job Queuing Model (MJQM).In this paper, we study the behavior of a MJQM under saturation, i.e., when the waiting line always contains jobs to be served. We categorize jobs into two classes: the first class consists of jobs that only require one server, while the second class includes jobs that require a larger number of servers. We obtain the system utilization and the throughput of the two job classes for the case in which the number of servers requested by jobs in the second class is equal to the number of available servers, using a simple approach that allows for a general distribution of the service time of jobs in the second class. Hence, we derive the stability condition of the non-saturated MJQM under these assumptions. Additionally, we develop an approximate analysis for the case in which the jobs of the second class require a fraction of the available servers.Based on analytical and numerical results, we highlight interesting system properties and insights.

Employability and Education Mismatch in the Moroccan Urban Workplace

This study explores the skill demand and supply of the current Moroccan urban labor workforce. It seeks to understand the experience of skill mismatch, and how it affects employability. Using semi-structured interviews and a phenomenological approach, the study explores the nature of job queues as informed by recruiters and educators. A total of 27 interviews were conducted with recruits, recruiters, and university professors, all of whom were selected using a criterion-based approach coupled with snowball sampling. Over-education was affirmed among employees, and the findings yielded a crossover of skill sets. The job queue for a balanced labor market is oriented towards personal attitudes and capabilities related to the field of study. Finally, responses from both recruiters’ and professors’ interviews yielded a framework of education-job matches for the demand-supply sides, and oriented workplace-based employability.

Job Scheduling By Using Swarm Intelligent Algorithms: survey

The job scheduling problem is used in industry, manufacturing, job planning, and the network environment to manage users’ jobs on the right machines with different limitations. There are three types of job scheduling problems. Firstly, job shop scheduling means each job is executed on a determined sequence of machines specified previously. While in the flow shop scheduling problem, each job can be carried on one idle sequencing using a job queue. The third type, open shop, the sequence of jobs can be carried out on any free machine. This paper describes the various types of job scheduling problems and the various swarm intelligence algorithms, particularly cuckoo search algorithm, that can used to solve them.

Efficient Decentralized Multi-agent Learning in Asymmetric Bipartite Queueing Systems

New Algorithm Enables Efficient Decentralized Learning in Bipartite Queueing Systems Bipartite queueing systems, where agents with individual job queues request service from a pool of heterogeneous servers, are standard models for service applications like data networks and call centers. Traditionally, a central controller schedules agent requests with full knowledge of system parameters. However, emerging applications require decentralized operation without this central coordination or complete system information. This presents challenges as agents lack the global knowledge needed to efficiently route jobs. Recent research into efficient decentralized learning algorithms for such systems faces limitations in nonoptimal throughput, demanding computations, or degrading efficiency with exponential queue growth. In contrast, this paper introduces an algorithm that enables queues to efficiently learn decentralized scheduling policies while ensuring throughput optimality. The approach is computationally lightweight, achieving queue length bounds that scale polynomially rather than exponentially in system size. Experiments demonstrate faster convergence and robustness of our algorithm compared with prior decentralized algorithms.

Automatic performance tuning for Albany Land Ice

Accurate simulation of the evolution of polar ice-sheets requires a massive amount of computational power. In order to take advantage of the newest generation of supercomputing clusters, the Albany Land Ice code has been modernized for performance portability across a variety of parallel architectures, with a focus on enabling end-to-end GPU capability. Albany uses a multigrid preconditioning approach for solving linear systems via performance portable smoothers from the Trilinos package Ifpack2. Since the Albany Land Ice code is constantly evolving and both Albany and Trilinos are in constant development, it is likely that the optimal choice of solver parameters will change over time. It is therefore critical to have an automatic performance tuning framework to ensure that the best possible performance is maintained. Toward this effect, we have developed an automatic performance tuning framework to determine the best fine- and coarse-grid smoothing algorithms and parameters. We treat the underlying performance model of the linear solve as a black box and use the python-based GPTune Bayesian optimization library to determine the optimal smoother choice and parameters. Using this approach, we have found smoothers and their corresponding parameters that result in, on average, 1.2 times faster, and up to 1.5 times faster solve-times than our manually-tuned parameters. We also show that the proposed auto-tuning approach produces reliably better parameters than naive black box optimization techniques like random search for a given function evaluation budget. By implementing our tuning framework in the Python-based workflow management tool parsl, we also ensure that we efficiently use available computing resources during the tuning process and avoid unnecessary long wait times in computing cluster job queues.

A Machine Learning Approach for an HPC Use Case: the Jobs Queuing Time Prediction

High-Performance Computing (HPC) domain provided the necessary tools to support the scientific and industrial advancements we all have seen during the last decades. HPC is a broad domain targeting to provide both software and hardware solutions as well as envisioning methodologies that allow achieving goals of interest, such as system performance and energy efficiency. In this context, supercomputers have been the vehicle for developing and testing the most advanced technologies since their first appearance. Unlike cloud computing resources that are provided to the end-users in an on-demand fashion in the form of virtualized resources (i.e., virtual machines and containers), supercomputers’ resources are generally served through State-of-the-Art batch schedulers (e.g., SLURM, PBS, LSF, HTCondor). As such, the users submit their computational jobs to the system, which manages their execution with the support of queues. In this regard, predicting the behaviour of the jobs in the batch scheduler queues becomes worth it. Indeed, there are many cases where a deeper knowledge of the time experienced by a job in a queue (e.g., the submission of check-pointed jobs or the submission of jobs with execution dependencies) allows exploring more effective workflow orchestration policies. In this work, we focused on applying machine learning (ML) techniques to learn from the historical data collected from the queuing system of real supercomputers, aiming at predicting the time spent on a queue by a given job. Specifically, we applied both unsupervised learning (UL) and supervised learning (SL) techniques to define the most effective features for the prediction task and the actual prediction of the queue waiting time. For this purpose, two approaches have been explored: on one side, the prediction of ranges on jobs’ queuing times (classification approach) and, on the other side, the prediction of the waiting time at the minutes level (regression approach). Experimental results highlight the strong relationship between the SL models’ performances and the way the dataset is split. At the end of the prediction step, we present the uncertainty quantification approach, i.e., a tool to associate the predictions with reliability metrics, based on variance estimation.

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.

The perceived labour market value of Massive Open Online Courses (MOOCs) in Europe and the USA

ABSTRACT The advantages of higher education have received significant attention over time. However, recent research seems to challenge this assumption. It highlights that returns to education may be subject to inflation, may vary in relation to skills, and may not be equally distributed, thus posing new questions about the role of formal education. Against this background, the growing popularity of new forms of online education and training such as MOOCs (Massive Open Online Courses) have emerged. Investigating the way different learners use MOOCs may contribute to a deeper understanding of the evolution of the labour market outcomes of both traditional and technologically-mediated educational qualifications. Based on 43 semi-structured interviews conducted with MOOC users in the USA and in Europe, this article explores the potential of MOOCs on the labour market. The positional competition approach can help frame the results, inasmuch as MOOCs emerge as ‘soft credentials’. These accessible and flexible educational tools seem to provide applicants in the job queue with additional resources, although their labour market value remains modest and ancillary to formal educational qualifications. From a long-term perspective, however, increasing reliance on this type of training may contribute to further shifting of responsibilities from collective actors to individual workers.

Optimized scheduling of multi‐user Map‐Reduce jobs in heterogeneous environment

SummaryMap‐Reduce is a programming paradigm widely used for data intensive applications in distributed computing. Divisible load theory (DLT) is another model to partition divisible load for parallel processing. In this work, a new job scheduler named virtual job scheduler (VJS) is developed to schedule the MapReduce jobs based on DLT. VJS constructs a virtual job set from the queue of jobs awaiting execution by considering the CPU and IO resource utilization levels of each job. The core of VJS is in its partitioning algorithm. Two novel partitioning algorithms, namely, two level successive partitioning (TLSP) and predictive partitioning (PRED) have been proposed. TLSP applies DLT to both map and reduce phase in succession. This second load partitioning performed during the comparatively longer reduce phase, exploits the advantage of DLT better. PRED is a modification of TLSP aimed at optimizing the overall schedule length, by taking the idle time of the reducers into consideration. Evaluation of these two models across various execution environments has indicated a profound decrease in the wait time of reducers and as a result has shown a significant reduction in the makespan of the whole job as such. VJS altered to incorporate PRED partitioning algorithm has proved to be suitable for heterogeneous environment, with high resource utilization compared with the default Hadoop MapReduce scheduler.

On the problem of the switching of the fast server

A Markovian model of a system with two heterogeneous servers is investigated, in which the slow server is always on, and the fast server can turn on when the job queue length reaches a certain value. Models with unlimited and limited size of the buffer of waiting jobs are considered. A numerical method for calculating the probability distribution of states and characteristics of the systems under study is developed. The results of numerical experiments are given.

Optimal Policy for Controlling Two-Server Queueing Systems with Jockeying

This paper studies the optimal policy for joint control of admission, routing, service, and jockeying in a queueing system consisting of two exponential servers in parallel. Jobs arrive according to a Poisson process. Upon each arrival, an admission/routing decision is made, and the accepted job is routed to one of the two servers with each being associated with a queue. After each service completion, the servers have an option of serving a job from its own queue, serving a jockeying job from another queue, or staying idle. The system performance is inclusive of the revenues from accepted jobs, the costs of holding jobs in queues, the service costs and the job jockeying costs. To maximize the total expected discounted return, we formulate a Markov decision process (MDP) model for this system. The value iteration method is employed to characterize the optimal policy as a hedging point policy. Numerical studies verify the structure of the hedging point policy which is convenient for implementing control actions in practice.

A queuing theory based delay-efficient packet scheduler for machine type communication

Lately, there is a drastic change in the telecommunication industry due to emergence of Internet of Things (IoT). Machine to Machine (M2M) plays a very important role in enabling IoT. As the LTE is mainly suitable for Human-to-Human (H2H) communication, MTC faces some challenges in LTE like radio resource management and uplink-scheduling algorithms are not suitable for MTC. In this paper, we use a queuing theory and propose a packet scheduler, which would model the regular and event-driven traffic and schedule the jobs in queue for request processing. Then using the Markov Chain process, we calculate the queue length and determine the blocking probability. If there is high blocking probability and still there are some requests in queue, then those requests will be processed in virtual server for maximum network usage.

Concurrent WIP and an application to clearing functions for complex heterogenous systems

Complex manufacturing systems are challenging to study because of the high level of information required and the inaccessibility of most of it. Their tractability is however essential for the efficiency of state-of-the-art industries. This is particularly the case in the semiconductor industry that faces high mix and low volume conditions, and for which traditional methods fail to capture the high complexity and require continuous actions and corrections to adjust to heterogeneous toolsets and product-mix.We present the Concurrent WIP (CWIP), a new way of studying such systems at the level of a process-cluster by identifying each job's queue from its own perspective. CWIP is designed to be practical, with a low level of resource investments, yet informative. We explain how CWIP can be computed based on historical data and then used to derive capacity estimates and clearing functions without any assumptions on the system or on the form of the functions. In the process, we derive not only an average workload-dependent capacity, but also a confidence interval on this capacity. The relevance and efficiency of the proposed estimates are experimentally tested on a simulated system mimicking a small but complex process-cluster of the semiconductor industry. The estimates are used to predict WIP absorption times and we show how they characterize well not only the average behavior but also the full range of possible behaviors of the system. Finally, we discuss further applications of CWIP, that could be used to compute refined clearing functions or to monitor complex systems.

Responding to rapidly changing product demand through a coordinated additive manufacturing production system: a COVID-19 case study

COVID-19’s lockdown policies saw Modern Manufacturing Practice (MMP) – batch/mass/just-in-time supply chains – severed and societal demands rapidly change from products such as vehicles and clothing to Personal Protection Equipment (PPE), ventilators and equipment for remote working. The critical and, in many cases, life preserving, need for responsive manufacturing resulted in government and frontline services turning to society’s Additive Manufacturing (AM) capability in homes, schools, universities, and industry to provide essential products and product replacements. While AM managed to respond and support some government and frontline services, the highly distributed and diverse nature of the nation’s AM resources resulted in potentially avoidable production inefficiencies and delays. This paper develops and evaluates a series of strategies for coordinating AM for rapidly changing product demand to further enhance the responsiveness and productivity of AM. The strategies presented in the paper employ a host-client agent-based architecture that enable local governance of production thereby enabling distributed AM resource to come together to tackle society’s production needs without the need for centralised coordination. To enable and support local governance, it is necessary to understand how the combination of production logics impact the overall performance of the production system. Correspondingly, the contribution of this paper lies in the characterisation and quantification of the impacts of production logic through the metrics of Time in System, Lateness and Jobs in Queue and the consequences for responsive AM production systems.

Coordinated global and private job-flow scheduling in grid virtual organizations

In this work, an approach for a preference-based job-flow scheduling in Grid virtual organizations is proposed and studied. Virtual organizations provide uniform rules of the resources sharing and consumption and should respect individual and common preferences of its stakeholders: users, resource providers and administrators. In most cases, a virtual organization’s stakeholders pursue contradictory market-based goals, which may be expressed in terms of time and cost. Still, mutually beneficial job-flow scheduling scenarios are possible in such multi-agent computing environments when the different optimization criteria are coordinated. For this purpose we consider different resources allocation heuristics and scenarios based on a simple linear combination of global (group) and private (user) job scheduling criteria. We study backfilling-based job queue scheduling procedure with a secondary preference-based optimization performed only on the resources allocation step. This approach allows us to respect backfilling-based jobs execution priorities together with more specific individual jobs criteria, including total execution cost or time minimization.