Runtime Prediction Research Articles

تحسين خوارزمية الجولة الدائرية باستخدام نماذج التعلم الآلي (KNN) في جدولة وحدة المعالجة المركزية

Round Robin is considered as one of the most practically recognized process scheduling algorithms in CPU scheduling because it is simple and fair. However, the efficiency of Round Robin depends a lot upon the selection of an optimal time quantum. If the quantum is too small, then frequent context switches are needed by the CPU; therefore, the overhead increases, and thus, the average waiting time for the processes also becomes long. As a result, system performance is reduced as more and more CPU time is used up in context switching, instead of task execution. it may behave similarly to the First Come First Serve (FCFS) algorithm if the time quantum is excessively long, which leads to extended average waiting time. This paper proposes an improved Round Robin algorithm by incorporating machine learning, which optimally determines the time quantum dynamically. More precisely, the K-Nearest Neighbors algorithm will be used, with NumPy in charge of data processing, for the runtime prediction of an optimal time quantum considering characteristics of processes. The experimental results showed a considerable improvement in the parameters of average waiting, turnaround time, and the number of context switches with respect to the traditional Round Robin algorithm. Results indicated that machine learning efficiently modifies the predictable scheduling algorithms to make the scheduling process adaptive and efficient in operating systems. Key words: dynamic round robin, classical round robin, burst time, machine learning.

A Disturbance-Driven Textual Model for Train Running Time Prediction on High-Speed Railways

A Disturbance-Driven Textual Model for Train Running Time Prediction on High-Speed Railways

Machine learning approaches to predict the execution time of the meteorological simulation software COSMO

AbstractPredicting the execution time of weather forecast models is a complex task, since these models are usually performed on High Performance Computing systems that require large computing capabilities. Indeed, a reliable prediction can imply several benefits, by allowing for an improved planning of the model execution, a better allocation of available resources, and the identification of possible anomalies. However, to make such predictions is usually hard, since there is a scarcity of datasets that benchmark the existing meteorological simulation models. In this work, we focus on the runtime predictions of the execution of the COSMO (COnsortium for SMall-scale MOdeling) weather forecasting model used at the Hydro-Meteo-Climate Structure of the Regional Agency for the Environment and Energy Prevention Emilia-Romagna. We show how a plethora of Machine Learning approaches can obtain accurate runtime predictions of this complex model, by designing a new well-defined benchmark for this application task. Indeed, our contribution is twofold: 1) the creation of a large public dataset reporting the runtime of COSMO run under a variety of different configurations; 2) a comparative study of ML models, which greatly outperform the current state-of-practice used by the domain experts. This data collection represents an essential initial benchmark for this application field, and a useful resource for analyzing the model performance: better accuracy in runtime predictions could help facility owners to improve job scheduling and resource allocation of the entire system; while for a final user, a posteriori analysis could help to identify anomalous runs.

Combining Machine Learning techniques and Genetic Algorithm for predicting run times of High Performance Computing jobs

This study proposes a novel approach combining Machine Learning (ML) techniques and Genetic Algorithms (GA) for predicting High-Performance Computing (HPC) job run times. The objective is to create a prediction method universally applicable to any HPC system, irrespective of workload characteristics, application specific parameters, user behavior, or hardware architecture. Since user-supplied run time estimates are often inaccurate, we aim to categorize job runtimes into several classes, allowing users to select appropriate classes for their jobs. A Genetic Algorithm is developed to optimally define these runtime classes, determining both the number of classes and the time intervals represented by them. Four Machine Learning algorithms (K Nearest Neighbours, Support Vector Regression, Extreme Gradient Boosting and Deep Neural Networks) are implemented for run time prediction. A unique set of features extracted from historical job data serves as input to the Machine Learning models. The generalized nature of our method is demonstrated by validating its performance on data from six clusters with distinct configurations, applications and runtime distributions. Our results illustrate the superior performance of Machine Learning models incorporating GA-defined runtime classes for all datasets. Across all six datasets, our method achieves R2 scores exceeding 0.8, and accuracy greater than 0.7.

Workload Placement on Heterogeneous CPU-GPU Systems

The popularity of heterogeneous CPU-GPU processing has increased considerably in recent years. To efficiently utilize heterogeneous resources, data processing systems depend on an appropriate workload placement strategy to assign the right amount of compute to the right processor. However, finding an optimal placement strategy is not trivial due to various complex and conflicting tradeoffs related to the characteristics of processors, the nature of the workload, and data locality. In addition, placement decisions impact workload runtime and performance cost, and also depend on the availability of potentially different implementations for CPUs and GPUs, which adds extra complexity in such heterogeneous environments. In this tutorial, we review and compare state-of-the-art strategies for workload placement on heterogeneous CPU-GPU architectures, along with runtime prediction techniques and methods to support multi-device code. We also discuss open issues and identify potentially promising future research directions.

LS-ICE: A Load State Intercase Encoding framework for improved predictive monitoring of business processes

Research on developing techniques for predictive process monitoring has generally relied on feature encoding schemes that extract intra-case features from events to make predictions. In doing so, the processing of cases is assumed to be solely influenced by the attributes of the cases themselves. However, cases are not processed in isolation and can be influenced by the processing of other cases or, more generally, the state of the process under investigation. In this work, we propose the LS-ICE (load state intercase encoding) framework for encoding intercase features that enriches events with a depiction of the state of relevant load points in a business process. To assess the benefits of the intercase features generated using the LS-ICE framework, we compare the performance of predictive process monitoring models constructed using the encoded features against baseline models without these features. The models are evaluated for remaining trace and runtime prediction using five real-life event logs. Across the board, a consistent improvement in performance is noted for models that integrate intercase features encoded through the proposed framework, as opposed to baseline models that lack these encoded features.

Performance Prediction of Power Beacon-Aided Wireless Sensor-Powered Non-Orthogonal Multiple-Access Internet-of-Things Networks under Imperfect Channel State Information

In this paper, we investigate a novel power beacon (PB)-aided wireless sensor-powered non-orthogonal multiple-access (NOMA) Internet-of-Things (IoT) network under imperfect channel state information (CSI). Furthermore, the exact expression outage probability (OP) of two IoT users is derived to analyze the performance of the considered network. To give further insight, the expression asymptotic OP and diversity order are also expressed when the transmit power at the PB goes to infinity. Furthermore, a deep neural network (DNN) framework is proposed to concurrently forecast IoT users’ OP in relation to real-time setups for IoT users. Additionally, when compared to the traditional analysis, our created DNN shows the shortest run-time prediction, and the outcomes predicted by the DNN model almost match those of the simulation. In addition, numerical results validate our analysis, simulation, and prediction through a Monte Carlo Simulation. Furthermore, the results show the impact of the main parameter on our proposed system. Finally, these findings show that NOMA performs better than the conventional orthogonal multiple-access (OMA) techniques.

Efficient and accurate compound scaling for convolutional neural networks

Designing efficient and accurate network architectures to support various workloads, from servers to edge devices, is a fundamental problem as the use of Convolutional Neural Networks (ConvNets) becomes increasingly widespread. One simple yet effective method is to scale ConvNets by systematically adjusting the dimensions of the baseline network, including width, depth, and resolution, enabling it to adapt to diverse workloads by varying its computational complexity and representation ability. However, current state-of-the-art (SOTA) scaling methods for neural network architectures overlook the inter-dimensional relationships within the network and the impact of scaling on inference speed, resulting in suboptimal trade-offs between accuracy and inference speed. To overcome those limitations, we propose a scaling method for ConvNets that utilizes dimension relationship and runtime proxy constraints to improve accuracy and inference speed. Specifically, our research notes that higher input resolutions in convolutional layers lead to redundant filters (convolutional width) due to increased similarity between information in different positions, suggesting a potential benefit in reducing filters while increasing input resolution. Based on this observation, the relationship between the width and resolution is empirically quantified in our work, enabling models with higher parametric efficiency to be prioritized through our scaling strategy. Furthermore, we introduce a novel runtime prediction model that focuses on fine-grained layer tasks with different computational properties for more accurate identification of efficient network configurations. Comprehensive experiments show that our method outperforms prior works in creating a set of models with a trade-off between accuracy and inference speed on the ImageNet datasets for various ConvNets.

A Direct Data Aware LSTM Neural Network Architecture for Complete Remaining Trace and Runtime Prediction

Developing LSTM neural networks that can accurately predict the future trajectory of ongoing cases and their remaining runtime is an active area of research in predictive process monitoring. In this work a novel complete remaining trace prediction (CRTP) LSTM is proposed. This model is trained to directly predict the complete remaining trace and runtime of cases in contrast to single event prediction as is considered in previously published research on this topic. This makes the CRTP-LSTM robust in terms of utilizing all available attributes of previously observed events for prediction, consequently it can be considered natively data aware. In an extensive experimental assessment the authors show that CRTP-LSTMs consistently outperform other considered approaches for both remaining trace and runtime prediction. Furthermore, the authors show that including all available information contained in previously observed events has a positive impact on the performance of the CRTP-LSTM model. This indicates that valuable information can be extracted from attributes of events in order to make more accurate trace and runtime predictions. This opens up interesting avenues for future research including the incorporation of inter-case features into a modeling setup when predicting the remaining trace and runtime of cases.

Job runtime prediction of HPC cluster based on PC-Transformer

Job runtime prediction of HPC cluster based on PC-Transformer

A time-sensitive learning-to-rank approach for cloud simulation resource prediction

Predicting the computing resources required by simulation applications can provide a more reasonable resource-allocation scheme for efficient execution. Existing prediction methods based on machine learning, such as classification/regression, typically must accurately predict the runtime of simulation applications and select the optimal computing resource allocation scheme by sorting the length of the simulation runtime. However, the ranking results are easily affected by the simulation runtime prediction accuracy. This study proposes a time-sensitive learning-to-rank (LTR) approach for cloud simulations resource prediction. First, we use the Shapley additive explanation (SHAP) value from the field of explainable artificial intelligence (XAI) to analyze the impact of relevant factors on the simulation runtime and to extract the feature dimensions that significantly affect the simulation runtime. Second, by modifying the target loss function of the rankboost algorithm and training a time-sensitive LTR model based on simulation features, we can accurately predict the computing resource allocation scheme that maximizes the execution efficiency of simulation applications. Compared with the traditional machine learning prediction algorithm, the proposed method can improve the average sorting performance by 3%–48% and can accurately predict the computing resources required for the simulation applications to execute in the shortest amount of time.

Battery-Powered RSU Running Time Monitoring and Prediction Using ML Model Based on Received Signal Strength and Data Transmission Frequency in V2I Applications

The application of the Internet of Things (IoT), vehicles to infrastructure (V2I) communication and intelligent roadside units (RSU) are promising paradigms to improve road traffic safety. However, for the RSUs to communicate with the vehicles and transmit the data to the remote location, RSUs require enough power and good network quality. Recent advances in technology have improved lithium-ion battery capabilities. However, other complementary methodologies including battery management systems (BMS) have to be developed to provide an early warning sign of the battery's state of health. In this paper, we have evaluated the impact of the received signal strength indication (RSSI) and the current consumption at different transmission frequencies on a static battery-based RSU that depends on the global system for mobile communications (GSM)/general packet radio services (GPRS). Machine learning (ML) models, for instance, Random Forest (RF) and Support Vector Machine (SVM), were employed and tested on the collected data and later compared using the coefficient of determination (R2). The models were used to predict the battery current consumption based on the RSSI of the location where the RSUs were imposed and the frequency at which the RSU transmits the data to the remote database. The RF was preferable to SVM for predicting current consumption with an R2 of 98% and 94%, respectively. It is essential to accurately forecast the battery health of RSUs to assess their dependability and running time. The primary duty of the BMS is to estimate the status of the battery and its dynamic operating limits. However, achieving an accurate and robust battery state of charge remains a significant challenge. Referring to that can help road managers make alternative decisions, such as replacing the battery before the RSU power source gets drained. The proposed method can be deployed in other remote WSN and IoT-based applications.

Exploring job running path to predict runtime on multiple production supercomputers

There are massive jobs submitted in the supercomputer, and the job management system is typically deployed to schedule these jobs and allocate compute resources. FCFS (First Come First Serve) is a popular scheduling policy and the job's priority is determined based on the arrival time. However, under the FCFS strategy, if existing idle resources cannot meet the requirement of the head job in the waiting queue, they cannot be allocated to other jobs, which suffers from resource waste. To optimize the resource utilization, the backfilling method is proposed, which allocates the reserved idle compute nodes to a small-size and short-running non-head job, on the premise of not delaying the original head job. Obtaining the job's runtime in advance is necessary for backfilling and the traditional method relies on the user's estimation. Unfortunately, the estimated runtime provided by users is generally overestimated. Many studies extract features from historical job logs and adopt machine learning to predict the runtime. However, traditional features are insufficient to characterize the job. In this paper, we collect job logs from two supercomputers and present a novel runtime prediction framework called PREP. PREP explores the job's running path as a new feature, which implies plentiful information about the job's properties, such as the user, the project, the scale of data sets, and the parameters used. As there is a strong correlation between the job's runtime and its running path, we group jobs with similar paths into a cluster and train a runtime prediction model for each cluster respectively. Extensive evaluations demonstrate that introducing the new feature can achieve higher prediction accuracy (88.5% and 82.3% in two production supercomputers respectively), and our framework has a more desirable prediction performance than other popular strategies like Last-2 and IRPA. In addition, the predicted runtime is inserted into the real job trace of a slurm simulator to verify the advantages of PREP.

Reevaluating Graph-Neural-Network-Based Runtime Prediction of SAT-Based Circuit Deobfuscation

Logic locking is a technique that can help hinder reverse-engineering-based attacks in the IC supply chain from untrusted foundries or end-users. In 2015, the Boolean Satisfiability (SAT) attack was introduced. Although the SAT attack is effective in deobfuscating a wide range of logic locking schemes, its execution time varies widely from a few seconds to months. Previous research has shown that Graph Convolutional Networks (GCN) may be used to estimate this deobfuscation time for locked circuits with varied key sizes. In this paper, we explore whether GCN models truly understand/capture the structural/functional sources of deobfuscation hardness. In order to tackle this, we generate different curated training datasets: traditional ISCAS benchmark circuits locked with varying key sizes, as well as an important novel class of synthetic benchmarks: Substitution-Permutation Networks (SPN), which are circuit structures used to produce the most secure and efficient keyed-functions used today: block-ciphers. We then test whether a GCN trained on a traditional benchmark can predict the simple fact that a deeper SPN is superior to a wide SPN of the same size. We find that surprisingly the GCN model fails at this. We propose to overcome this limitation by proposing a set of circuit features motivated by block-cipher design principles. These features can be used as stand-alone or combined with GCN models to provide deeper topological cues than what GCNs can access.

Characterizing and Optimizing EDA Flows for the Cloud

Design space exploration in logic synthesis and parameter tuning in physical design require a massive amount of compute resources in order to meet tapeout schedules. To address this need, cloud computing provides semiconductor and electronics companies with instant access to scalable compute resources. However, deploying electronic design automation (EDA) jobs on the cloud requires EDA teams to deeply understand the characteristics of their jobs in cloud environments. Unfortunately, there has been little to no public information on these characteristics. Thus, in this article, we first formulate the problem of moving EDA jobs to the cloud. To address the problem, we characterize the performance of four EDA main applications, namely: 1) synthesis; 2) placement; 3) routing; and 4) static timing analysis. We show that different EDA jobs require different compute configurations in order to achieve the best performance. Using observations from our characterization, we propose a novel model based on graph convolutional networks to predict the total runtime of a given stage on different configurations. Our model achieves a prediction accuracy of 87%. Furthermore, we present a new formulation for optimizing cloud deployments in order to reduce costs while meeting deadline constraints. We present a pseudopolynomial optimal solution using a multichoice knapsack mapping that reduces deployment costs by 35.29%, with minimal overhead to the total runtime. In addition, we describe a cloud-ready solution, called EDA analytics central, for the continuous optimization of a design across an EDA flow. We used this system in building our runtime prediction model.

Performance Analysis and Deep Learning Design of Wireless Powered Cognitive NOMA IoT Short-Packet Communications With Imperfect CSI and SIC

In this article, we study wireless-powered cognitive nonorthogonal multiple access (NOMA) Internet of Things (IoT) networks with short-packet communications to improve spectrum utilization and sustainability, as well as reduce the latency under imperfect channel state information (CSI) and successive interference cancelation (SIC). For performance evaluation, closed-form expressions for the block error rate (BLER) of the NOMA users, goodput, energy efficiency, latency, and reliability are derived. To gain some further insights into the system design, two scenarios can be taken into account for the positions of the primary receivers: 1) they are located near the secondary network and 2) they are located far away from the secondary network. Moreover, we propose an effective algorithm to minimize the BLERs of the NOMA users by optimizing power allocation coefficients. In addition, a novel multi-output deep-learning (DL) framework is designed to simultaneously predict the BLERs and goodputs of users towards real-time configurations for IoT systems. Numerical results show the outstanding performance of the proposed system over the orthogonal multiple access (OMA) one in terms of the BLER and goodput. Moreover, the proposed system achieves a lower latency and higher reliability compared to the long packet communications under the same channel settings. Furthermore, our designed multioutput DL also exhibits the lowest error performance and a short run-time prediction compared to the other multioutput regression models, while the predicted results using the DL model are almost matched with the simulation ones.

Scalable performance analysis method for SPMD applications

The analysis of parallel scientific applications allows us to understand their computational and communication behavior. One way of obtaining performance information is through performance tools. One such tool is parallel application signatures for performance prediction (PAS2P), based on parallel application repeatability, focusing on performance analysis and prediction. The same resources that execute the parallel application are used to perform its analysis, creating a machine independent model of the application and identifying its common patterns. However, the analysis is costly in terms of execution time due to the high number of synchronization communications performed by PAS2P, degrading performance as the number of processes increases. To solve this problem, we propose a model that reduces data dependency between processes, reducing the number of communications performed by PAS2P in the analysis stage and taking advantage of the characteristics of single program, multiple sata applications. Our analysis proposal allows us to decrease the analysis time by 29 times when the application scales to 256 processes, while keeping error levels below 11% in the runtime prediction. It is important to mention that the analysis time is not considerably affected by increasing the number of application processes.

KubeSC‐RTP: Smart scheduler for Kubernetes platform on CPU‐GPU heterogeneous systems

SummaryHeterogeneous systems composed of multiple CPUs and GPUs are progressively attractive as platforms for high performance computing because of their higher performance. Especially with the use of containers which are rapidly replacing virtual machines as the compute instance of choice in cloud‐based deployments such in Kubernetes clusters. The task scheduling in a heterogeneous environment became one of the most important issues considered by the platform providers. The ability to choose the appropriate device, CPU or GPU, has a direct impact on the performance of a particular system. It reduces total processing time and increases customer satisfaction. In heterogeneous systems, optimizing resource consumption is a critical aspect for cloud service providers. Adequate scheduling of an application implies optimization of its execution time, which results in resource consumption for the service provider. The development of algorithms for scheduling applications in heterogeneous computing systems has received a significant amount of attention in recent years. A variety of efforts are dedicated to the design of such scheduling algorithms. This article is one of those efforts. We present in this work, KubeSC‐RTP, a scheduler for Kubernetes environment using machine learning based on runtime prediction of the applications in order to better select the appropriate device, CPU or GPU.

Runtime prediction of big data jobs: performance comparison of machine learning algorithms and analytical models

Due to the rapid growth of available data, various platforms offer parallel infrastructure that efficiently processes big data. One of the critical issues is how to use these platforms to optimise resources, and for this reason, performance prediction has been an important topic in the last few years. There are two main approaches to the problem of predicting performance. One is to fit data into an equation based on a analytical models. The other is to use machine learning (ML) in the form of regression algorithms. In this paper, we have investigated the difference in accuracy for these two approaches. While our experiments used an open-source platform called Apache Spark, the results obtained by this research are applicable to any parallel platform and are not constrained to this technology. We found that gradient boost, an ML regressor, is more accurate than any of the existing analytical models as long as the range of the prediction follows that of the training. We have investigated analytical and ML models based on interpolation and extrapolation methods with k-fold cross-validation techniques. Using the interpolation method, two analytical models, namely 2D-plate and fully-connected models, outperform older analytical models and kernel ridge regression algorithm but not the gradient boost regression algorithm. We found the average accuracy of 2D-plate and fully-connected models using interpolation are 0.962 and 0.961. However, when using the extrapolation method, the analytical models are much more accurate than the ML regressors, particularly two of the most recently proposed models (2D-plate and fully-connected). Both models are based on the communication patterns between the nodes. We found that using extrapolation, kernel ridge, gradient boost and two proposed analytical models average accuracy is 0.466, 0.677, 0.975, and 0.981, respectively. This study shows that practitioners can benefit from analytical models by being able to accurately predict the runtime outside of the range of the training data using only a few experimental operations.

Wireless Powered Cognitive NOMA-Based IoT Relay Networks: Performance Analysis and Deep Learning Evaluation

In this article, we study novel wireless powered cognitive nonorthogonal multiple access (NOMA)-based Internet-of-Things (IoT) relay networks to improve the performance of a cell-edge user under perfect and imperfect successive interference cancelation (SIC). In the secondary networks, a source node communicates with a cell-center user via direct link and with a cell-edge user through the assistance of a master IoT node under cognitive radio constraint. Exact closed-form analytical expressions for the outage probability (OP) of NOMA users and the overall system throughput are derived. To provide further insights, a performance floor analysis is also carried out considering two power-setting scenarios: 1) the transmit powers at the power beacon goes to infinity and 2) the maximum allowable power constraint goes to infinity. Moreover, we develop two iterative algorithms for minimizing OP users and maximizing system throughput subject to time-switching and power-allocation factors in two-hop transmission. Direct derivation of the closed-form expression for the ergodic capacity (EC) becomes unfeasible due to the high complexity of the proposed system model. To overcome this issue, we design a deep neural network (DNN) framework for the EC prediction toward real-time configurations. Our results show that the predicted results based on this DNN framework perfectly align with the simulations, validating our design framework. In addition, the DNN approach exhibits the lowest root-mean-square error and low run-time predictions among other regression models.