Execution Time Estimation Research Articles

Predictive modelling of MapReduce job performance in cloud environments using machine learning techniques

Within the Hadoop ecosystem, MapReduce stands as a cornerstone for managing, processing, and mining large-scale datasets. Yet, the absence of efficient solutions for precise estimation of job execution times poses a persistent challenge, impacting task allocation and distribution within Hadoop clusters. In this study, we present a comprehensive machine learning approach for predicting the execution time of MapReduce jobs, encompassing data collection, preprocessing, feature engineering, and model evaluation. Leveraging a rich dataset derived from comprehensive Hadoop MapReduce job traces, we explore the intricate relationship between cluster parameters and job performance. Through a comparative analysis of machine learning models, including linear regression, decision tree, random forest, and gradient-boosted regression trees, we identify the random forest model as the most effective, demonstrating superior predictive accuracy and robustness. Our findings underscore the critical role of features such as data size and resource allocation in determining job performance. With this work, we aim to enhance resource management efficiency and enable more effective utilisation of cloud-based Hadoop clusters for large-scale data processing tasks.

Estimate of the time required to perform a nonadiabatic holonomic quantum computation

Nonadiabatic holonomic quantum computation has been proposed as a method to implement quantum logic gates with robustness comparable to that of adiabatic holonomic gates but with shorter execution times. In this paper, we establish an isoholonomic inequality for quantum gates, which provides a lower bound on the lengths of cyclic transformations of the computational space that generate a specific gate. Then, as a corollary, we derive a nonadiabatic execution time estimate for holonomic gates. In addition, we demonstrate that under certain dimensional conditions, the isoholonomic inequality is tight in the sense that every gate on the computational space can be implemented holonomically and unitarily in a time-optimal way. We illustrate the results by showing that the procedures for implementing a universal set of holonomic gates proposed in a pioneering paper on nonadiabatic holonomic quantum computation saturate the isoholonomic inequality and are thus time optimal. Published by the American Physical Society 2024

PerfTop: Towards performance prediction of distributed learning over general topology

Distributed learning with multiple GPUs has been widely adopted to accelerate the training process of large-scale deep neural networks. However, misconfiguration of the GPU clusters with various communication primitives and topologies could potentially diminish the gains in parallel computation and lead to significant degradation in training efficiency. Predicting the performance of distributed learning enables service providers to identify potential bottlenecks beforehand. In this work, we propose a Performance prediction framework over General Topologies, called PerfTop, for accurate estimation of per-iteration execution time. The main strategy is to integrate computation time prediction with an analytical model to map the nonlinearity in communication and fine-grained computation-communication patterns. This enables accurate prediction of a variety of neural network models over general topologies, such as tree, hierarchical, and exponential. Our extensive experiments show that PerfTop outperforms existing methods in estimating both computation and communication time, particularly for communication, surpassing the existing methods by over 45%. Meanwhile, it achieves an accuracy of above 85% in predicting the execution time over general topologies compared to simple topologies such as star and ring from the previous works.

Query execution time estimation in graph databases based on graph neural networks

Query execution time estimation is an essential task for databases, accurate estimation results can help administrators to manage and monitor systems. This study proposes an interaction-aware and dependency-aware query execution time estimation approach that utilizes graph neural networks to capture dependence and interaction relationships. We divide graph query execution time estimation tasks into three stages: workload generation and running, graph-based feature modeling and representation, training and estimation. Specifically, we generate query workloads and run them to collect the database and plan information when queries are executed. Then, the collected plan and database components are modeled into vertexes, the interaction and dependency between them are modeled into edges of graph-based feature representation. We develop an estimation model based on graph neural networks, in which the vertex embedding network is proposed to deal with the vertex heterogeneity, and the message passing network is proposed to aggregate the local representation into the global representation to obtain an embedding that can represent the higher-order feature information of the whole graph, and the estimation network is proposed to estimate execution times. The experiment results on datasets show that our estimation approach can improve estimation quality and outperform other estimation approaches in terms of estimation accuracy.

LOGISTICS MODEL VS GOMPERTZ MODEL IN APPROXIMATION OF “S” CURVES IN WATER SUPPLY SYSTEMS PROJECTS

Much of the success of an engineering project depends on good planning. Most engineers use an S curve to estimate the relationship between the execution time variables and the financial resources available for construction. Mathematically speaking, the functions that can be used for said estimation are those that pass through the origin of the Cartesian plane, given that at the initial time of the project the cost used is zero, they are also increasing and upon reaching the estimated execution time it is expected to have used the total amount assigned for the work. This investigation began in 2019 where the forms of several executed water supply systems were taken and with this data, the planning curve was modeled with a logistic equation adjusted with least squares and validated with new projects to analyze its effectiveness. Now in 2023, data from new projects is added and a new model was made based on the Gompertz curve that has a graph similar to the logistic curve, but with the difference that the curvature at the beginning and end of the time interval it is not always the same, as is the case with the logistic equation model, and it is expected that this characteristic will improve the approximation to the real execution curve of the project. It is concluded that the Gompertz model is indeed better than the logistic equation model for this type of approach and a tool is presented that is used for decision making.

Performance-driven scheduling for malleable workloads

The development of adaptive scheduling algorithms that take advantage of malleability has become a crucial area of research in many large-scale projects. Malleable workloads can improve the system’s performance but, at the same time, provide an extra dimension to the scheduling problem. This paper proposes an adaptive, performance-based job scheduling method that emphasizes the backfilling concept with malleability. The proposed method performs the malleability operations only when the estimated execution time of the involved applications is better than or equal to the execution time on the allocated resources without reconfiguration. The reconfiguration feasibility is determined by performance models considering the application scalability and reconfiguration overheads. Different policies for implementing malleability are presented, each targeting a specific workload in terms of job size and scalability. The comprehensive evaluation shows an improvement in the slowdown up to 49% compared to the non-adaptive baseline scheduling algorithm.

Alk: A Formal-Methods-based Educational Platform for Enhancing Algorithmic Thinking

Algorithm design courses are fundamental to computer science curricula, but fostering algorithmic thinking in students is challenging due to the diverse skills and creativity required. Dedicated teaching support tools can help both course instructors and students in this effort. We have developed the Alk platform to promote algorithmic thinking, leveraging the theoretical foundations of Matching Logic. Alk features an intuitive algorithm language that provides a flexible computational model suitable for analysis, symbolic execution, and checking properties of algorithms. In this paper, we present an overview of the Alk platform tool and demonstrate, through use cases, how it fosters various algorithmic thinking skills. We conclude that the Alk platform is a valuable tool for learning and teaching algorithms, effectively enhancing students’ understanding and skills. Future work will extend its capabilities of supporting symbolic execution, probabilistic algorithms, as well as estimation of execution time, further broadening its impact on computer science education.

Worst Case Execution Time and Power Estimation of Multicore and GPU Software: A Pedestrian Detection Use Case

Worst Case Execution Time estimation of software running on parallel platforms is a challenging task, due to resource interference of other tasks and the complexity of the underlying CPU and GPU hardware architectures. Similarly, the increased complexity of the hardware, challenges the estimation of worst case power consumption. In this paper, we employ Measurement Based Probabilistic Timing Analysis (MBPTA), which is capable of managing complex architectures such as multicores. We enable its use by software randomisation, which we show for the first time that is also possible on GPUs. We demonstrate our method on a pedestrian detection use case on an embedded multicore and GPU platform for the automotive domain, the NVIDIA Xavier. Moreover, we extend our measurement based probabilistic method in order to predict the worst case power consumption of the software on the same platform.

Neural Network Mapping of Industrial Robots’ Task Times for Real-Time Process Optimization

The ability to predict the maximal performance of an industrial robot executing non-deterministic tasks can improve process productivity through time-based planning and scheduling strategies. These strategies require the configuration and the comparison of a large number of tasks in real time for making a decision; therefore, an efficient task execution time estimation method is required. In this work, we propose the use of neural network models to approximate the task time function of a generic multi-DOF robot; the models are trained using data obtained from sophisticated motion planning algorithms that optimize the shape of the trajectory and the executed motion law, taking into account the kinematic and dynamic model of the robot. For scheduling purposes, we propose to evaluate only the neural network models, thus confining the online use of the motion planning software to the full definition of the actually scheduled task. The proposed neural network model presents a uniform interface and an implementation procedure that is easily adaptable to generic robots and tasks. The paper’s results show that the models are accurate and more efficient than the full planning pipeline, having evaluation times compatible with real-time process optimization.

Predictable GPU Wavefront Splitting for Safety-Critical Systems

We present a predictable wavefront splitting (PWS) technique for graphics processing units (GPUs). PWS improves the performance of GPU applications by reducing the impact of branch divergence while ensuring that worst-case execution time (WCET) estimates can be computed. This makes PWS an appropriate technique to use in safety-critical applications, such as autonomous driving systems, avionics, and space, that require strict temporal guarantees. In developing PWS on an AMD-based GPU, we propose microarchitectural enhancements to the GPU, and a compiler pass that eliminates branch serializations to reduce the WCET of a wavefront. Our analysis of PWS exhibits a performance improvement of 11% over existing architectures with a lower WCET than prior works in wavefront splitting.

Mitigating Mode-switch through Run-time Computation of Response Time

Mixed-critical systems consist of applications with different criticality. In these systems, different confidence levels of Worst-Case Execution Time (WCET) estimations are used. Dual criticality systems use a less pessimistic, but with lower level of assurance, WCET estimation, and a safe, but pessimistic, WCET estimation. Initially, both high and low criticality tasks are executed. When a high criticality task exceeds its less pessimistic WCET, the system switches mode and low criticality tasks are usually dropped, reducing the overall system Quality of Service (QoS). To postpone mode switch, and thus, improve QoS, existing approaches explore the slack, created dynamically, when the actual execution of a task is faster than its WCET. However, existing approaches observe this slack only after the task has finished execution. To enhance dynamic slack exploitation, we propose a fine-grained approach that is able to expose the slack during the progress of a task, and safely uses it to postpone mode switch. The evaluation results show that the proposed approach has lower cost and achieves significant improvements in avoiding mode-switch, compared to existing approaches.

Design of a optimization algorithm for binary classification

In the present work, the design of a system to classify data is carried out, using the Scrum methodology. The validation was carried out by expert judgment, having favorable results in terms of different criteria such as; integrity, ease of use, innovation, and scalability. Regarding the development of the functional elements of the system, it was obtained; he developed the architecture of the system, the database, and the prototypes, among other points considered. From the implementation of the system, the equation of a classifying plane in three-dimensional space will be obtained, as well as the number of internal iterations that the algorithm develops, the estimated execution time, and the graph of the plane. This system is based on a recently introduced symmetric cone proximal multiplier algorithm to solve separable optimization problems, this algorithm made an application for classification-related support vector machines.

Performance models of data parallel DAG workflows for large scale data analytics

Directed Acyclic Graph (DAG) workflows are widely used for large-scale data analytics in cluster-based distributed computing systems. The performance model for a DAG on data-parallel frameworks (e.g., MapReduce) is a research challenge because the allocation of preemptable system resources among parallel jobs may dynamically vary during execution. This resource allocation variation during execution makes it difficult to accurately estimate the execution time. In this paper, we tackle this challenge by proposing a new cost model, called Bottleneck Oriented Estimation (BOE), to estimate the allocation of preemptable resources by identifying the bottleneck to accurately predict task execution time. For a DAG workflow, we propose a state-based approach to iteratively use the resource allocation property among stages to estimate the overall execution plan. Furthermore, to handle the skewness of various jobs, we refine the model with the order statistics theory to improve estimation accuracy. Extensive experiments were performed to validate these cost models with HiBench and TPC-H workloads. The BOE model outperforms the state-of-the-art models by a factor of five for task execution time estimation. For the refined skew-aware model, the average prediction error is under 3% when estimating the execution time of 51 hybrid analytics (HiBench) and query (TPC-H) DAG workflows.

Analysis of benchmark program results of worst case execution time for multithreaded programs

<span lang="EN-US">Worst case execution time (WCET) estimation by static analyzers is being investigated with keen interest in view of their importance in designing applications for embedded systems that have real- time requirements. Recent work reported on improving precision of estimates of WCET of multithreaded programs, by improving precision of shared instruction cache analysis, shows significant improvement in WCET estimates. An abstraction of a multithreaded program as Hoare’s communicating sequential processes (CSP) specification program is realized to enable higher precision in micro-architectural modelling unit of WCET analyzer of multithreaded programs. A thread is viewed as a composition of CSP. The WCET of a thread may be viewed as dependent on WCET of processes in a thread and in turn WCET of each process is the WCET of the sub-graph of basic block nodes in the process. Corresponding CSP in interacting threads, based on calls to synchronization primitives wait and notify, generate shared cache interferences to the process in a thread whose WCET is being estimated by the analyzer. A detailed study of how partitioning of a thread into processes yields higher reduction in WCET is performed on benchmark programs. Furthermore, which processes in a thread yield higher reduction in WCET is performed.</span>

Estimation of execution time for computing tasks

This work aims to estimate the execution time of data processing tasks (specific executions of a program or an algorithm) before their execution. The paper focuses on the estimation of the average-case execution time (ACET). This metric can be used to predict the approximate cost of computations, e.g. when resource consumption in a High-Performance Computing system has to be known in advance. The presented approach proposes to create machine learning models using historical data. The models use program metadata (properties of input data) and parameters of the run-time environment as their explanatory variables. Moreover, the set of these variables can be easily expanded with additional parameters of the specific programs. The program code itself is treated as a black box. The response variable of the model is the execution time. The models have been validated within a Large-Scale Computing system that allows for a unified treatment of programs as computation modules. We present the process of training and validation for several different computation modules and discuss the suitability of the proposed models for ACET estimation in various computing environments.

Cost-based Optimization of Multistore Query Plans

Multistores are data management systems that enable query processing across different and heterogeneous databases; besides the distribution of data, complexity factors like schema heterogeneity and data replication must be resolved through integration and data fusion activities. Our multistore solution relies on a dataspace to provide the user with an integrated view of the available data and enables the formulation and execution of GPSJ queries. In this paper, we propose a technique to optimize the execution of GPSJ queries by formulating and evaluating different execution plans on the multistore. In particular, we outline different strategies to carry out joins and data fusion by relying on different schema representations; then, a self-learning black-box cost model is used to estimate execution times and select the most efficient plan. The experiments assess the effectiveness of the cost model in choosing the best execution plan for the given queries and exploit multiple multistore benchmarks to investigate the factors that influence the performance of different plans.

Execution Time Estimation of Multithreaded Programs With Critical Sections

The ideal benefit of parallelizing/multithreading a program is diminished in practice by several factors such as hardware scaling, memory bandwidth, power constraints, and synchronization due to critical sections. Several models have been proposed in the past to estimate the resulting performance and extend the traditional Amdahl’s law. In this work, we focus on the effect of synchronization, and develop a model for the execution time estimation of multithreaded programs under the presence of critical sections. The proposed model is applicable to multiple different critical sections and generalizes and improves previously proposed models. Experimental results on simulated, synthetic and benchmark examples show that the proposed model provides accurate approximations.

Sampling Rate Optimization and Execution Time Analysis for Two-Degrees-of-Freedom Control Systems

The current journal paper proposes an end-to-end analysis for the numerical implementation of a two-degrees-of-freedom (2DOF) control structure, starting from the sampling rate selection mechanism via a quasi-optimal manner, along with the estimation of the worst-case execution time (WCET) for the specified controller. For the sampling rate selection, the classical Shannon–Nyquist sampling theorem is replaced by an optimization problem that encompasses the trade-off between the fidelity of the controllers’ representation, along with the fidelity of the resulting closed-loop systems, and the implementation difficulty of the controllers. Additionally, the WCET analysis can be seen as a verification step before automatic code generation, a computational model being provided. The proposed computational model encompasses infinite-impulse response (IIR) and finite-impulse response (FIR) filter models for the controller implementation, along with additional relevant phenomena being discussed, such as saturation, signal scaling and anti-windup techniques. All proposed results will be illustrated on a DC motor benchmark control problem.

Efficient Obstacle Detection and Tracking Using RGB-D Sensor Data in Dynamic Environments for Robotic Applications.

Obstacle detection is an essential task for the autonomous navigation by robots. The task becomes more complex in a dynamic and cluttered environment. In this context, the RGB-D camera sensor is one of the most common devices that provides a quick and reasonable estimation of the environment in the form of RGB and depth images. This work proposes an efficient obstacle detection and tracking method using depth images to facilitate quick dynamic obstacle detection. To achieve early detection of dynamic obstacles and stable estimation of their states, as in previous methods, we applied a u-depth map for obstacle detection. Unlike existing methods, the present method provides dynamic thresholding facilities on the u-depth map to detect obstacles more accurately. Here, we propose a restricted v-depth map technique, using post-processing after the u-depth map processing to obtain a better prediction of the obstacle dimension. We also propose a new algorithm to track obstacles until they are within the field of view (FOV). We evaluate the performance of the proposed system on different kinds of data sets. The proposed method outperformed the vision-based state-of-the-art (SoA) methods in terms of state estimation of dynamic obstacles and execution time.

Spline-based trajectory generation to estimate execution time in a robotic assembly cell

Spline-based trajectory generation to estimate execution time in a robotic assembly cell