Execution Time Estimation Research Articles

Task graph scheduling in the presence of performance fluctuations ofcomputational resources

Most of the existing work in the area of task graph scheduling considers resources with fixed processing capacity. The algorithms in these works rely on an estimation of the execution times of tasks on different resources. However, in practice, due to fluctuations in performance of cloud resources, these algorithms have challenges in these environments. In this paper, we focus on the problem of fault-tolerant scheduling of task graphs in the presence of performance fluctuations of computational resources. With the aim of reducing the adverse impacts of both soft errors and resource performance degradations, we propose an opportunistic task replication scheme that uses idle durations of resources for replicating tasks. Unlike the previous works, the proposed algorithm does not rely on estimation of task execution times for finding idle resources. We introduce the notion of concurrency graphs and propose a graph theory-based algorithm for finding the number of idle resources during the execution of a set of tasks. The appropriate redundancy for each task is chosen with respect to the number of idle resources and the characteristics of the set of tasks that are being processed concurrently. Simulation experiments show that, in most situations, the proposed algorithm outperforms the previous algorithms in terms of average execution time and cost.

An enhanced priority-based scheduling heuristic for DAG applications with temporal unpredictability in task execution and data transmission

Abstract In the context of large-scale distributed computing environments, among various DAG scheduling heuristics aiming at coping with severe uncertainties during task execution, those ones that try to maximize the parallelism of ready tasks seems to be promising. However, most of such heuristics consider only the DAG structure and/or task execution cost when making scheduling decision, while the data transmission cost, which receives particular attention in big data era, is often ignored. When the information about task execution time and data transmission time (even though might be inaccurate) is given, it might be interesting to investigate whether this information should be taken into account during scheduling, and if yes, how to. This paper presents an enhanced priority-based heuristic, which properly considers the DAG structure attributes, the task execution time and data transmission time estimates in the task prioritization. The evaluation result shows the proposed approach obtains better scheduling performance than existing solutions.

Employing rule mining and multi-objective search for dynamic test case prioritization

Test case prioritization (TP) is widely used in regression testing for optimal reordering of test cases to achieve specific criteria (e.g., higher fault detection capability) as early as possible. In our earlier work, we proposed an approach for black-box dynamic TP using rule mining and multi-objective search (named as REMAP) by defining two objectives (fault detection capability and test case reliance score) and considering test case execution results at runtime. In this paper, we conduct an extensive empirical evaluation of REMAP by employing three different rule mining algorithms and three different multi-objective search algorithms, and we also evaluate REMAP with one additional objective (estimated execution time) for a total of 18 different configurations (i.e., 3 rule mining algorithms × 3 search algorithms × 2 different set of objectives) of REMAP. Specifically, we empirically evaluated the 18 variants of REMAP with 1) two variants of random search while using two objectives and three objectives, 2) three variants of greedy algorithm based on one objective, two objectives, and three objectives, 3) 18 variants of static search-based prioritization approaches, and 4) six variants of rule-based prioritization approaches using two industrial and three open source case studies. Results showed that the two best variants of REMAP with two objectives and three objectives significantly outperformed the best variants of competing approaches by 84.4% and 88.9%, and managed to achieve on average 14.2% and 18.8% higher Average Percentage of Faults Detected per Cost (APFDc) scores.

Determination of Worst-Case Data Using an Adaptive Surrogate Model for Real-Time System

The estimation of worst-case execution time (WCET) is a critical activity in the analysis of real-time systems. Evolutionary algorithms are frequently employed for the determination of worst-case data, used in the estimation of WCET. However, in order to employ an evolutionary algorithm, several executions of the application program are required, either on the target hardware or using its simulator. Multiple executions of the application program consume a huge amount of time. In order to reduce the huge execution time, this paper proposes the use of an adaptive surrogate model. The initial training of surrogate model is performed with a cycle-accurate simulator. The initially trained model is then used to assist the evolutionary algorithm by predicting the execution time of an application program. However, contrary to the direct training approach, the surrogate model in this paper is updated (adapted) during the evolution process. The adaptive training of a surrogate model increases its prediction accuracy and reduces the overall time. The validity of proposed methodology is illustrated with multiple sorting algorithms, extensively used in real-time systems.

Assessing mobile applications performance and energy consumption through experiments and Stochastic models

Energy consumption, execution time, and availability are common terms in discussions on application development for mobile devices. Mobile applications executing in a mobile cloud computing (MCC) environment must consider several issues, such as Internet connections problems and CPU performance. Misconceptions during the design phase can have a significant impact on costs and time-to-market, or even make the application development unfeasible. Anticipating the best configuration for each type of application is a challenge that many developers are not prepared to tackle. In this work, we propose models to rapidly estimate execution time, availability, and energy consumption of mobile applications executing in an MCC environment. We defined a methodology to create and validate Deterministic and Stochastic Petri net (DSPN) models to evaluate these three critical metrics. The DSPNs results were compared with results obtained through experiments performed on a testbed environment. We analyzed an image processing application, regarding connections type (WLAN, WiFi, and 3G), servers type (MCC or cloudlet), and functionalities performance. Our numerical analyses indicate, for instance, that the use of a cloudlet significantly improves performance and energy efficiency. Besides, the baseline scenario took us one month to implement, while modeling and evaluation the three scenarios required less than one day. In this way, our DSPN models represent a powerful tool for mobile developers to plan efficient and cost-effective mobile applications. They allow rapidly assess execution time, availability, and energy consumption metrics to improve the quality of mobile applications.

EXTES: An Execution-Time Estimation Scheme for Efficient Computational Science and Engineering Simulation via Machine Learning

In recent years, computational science and engineering (CSE) simulations using high-performance computing resources are actively exploited to solve complex domain-specific problems. Thanks to the remarkable advance of IT technology, the CSE community is challenging more complex and difficult problems than ever before, by running these simulations online. In this regard, we often witness that 1) online simulation users suffer from knowing little about the estimated termination time of their launched simulations and 2) the limited computing resources are squandered by wrong input that leads the simulations to run forever. To address such issues, we propose a novel execution time estimation scheme, termed EXTES , using machine learning techniques for more efficient online CSE simulations. With a large amount of existing provenance data, the EXTES scheme trains a suite of models rooted from classification, regression, and a hybrid of the two and utilize these models to estimate the execution time for specified input parameters for simulations. In the experiments on real simulation data, our proposed models achieved about 73% accuracy on average in execution time estimation across 16 simulation programs taken from a variety of CSE fields. In the meantime, the overhead incurred by the training and estimation is almost negligible.

Resource Allocation and Task Offloading for Heterogeneous Real-Time Tasks With Uncertain Duration Time in a Fog Queueing System

Fog computing has become the primary infrastructure on the Internet for improving the quality of service. We consider a fog queueing system with limited infrastructure resources to accommodate real-time tasks with heterogeneities in task types and execution deadlines. Owing to the uncertain execution duration, such a fog system should jointly consider fog resource allocation and a task offloading to satisfy the deadline requirements. To consider the task heterogeneity, a parallel virtual queue model is applied to buffer each type of task in a separate queue. Subsequently, we use a framework, including three parallel algorithms, namely, offloading, buffering, and resource allocation, to improve resource allocation balance, throughput, and task completion ratio. The task offloading is decided according to the task urgencies in terms of the laxity times, which accounts for the deadline, estimated execution time, and transmission delay to the cloud. The buffering process rearranges the arriving tasks based on their laxity time and the estimated task execution time at the fog tier. The resource allocation uses an adaptive queue weight based on the Lyapunov drift to avoid task starvation that may lead to a long queueing delay for tasks with long execution time. The simulation results indicate that our proposed policies can avoid task starvation and yields a tradeoff between high throughput and a high task completion ratio.

A static analytical performance model for GPU kernel

Graphics processing units (GPUs) have shown increased popularity and play an important role as kind of coprocessor in heterogeneous co-processing environment. Heavily data parallel problems can be solved efficiently due to tens of thousands threads collaborative work in parallel GPU architecture. The achieved performance, therefore, depends on the capability of multiple threads in parallel collaboration. This paper, a static analytical kernel performance model (SAKP) was proposed to estimate the execution time of GPU kernel. Especially, a set of kernel and device features for target GPU is generated in the proposed model. We determine the performance limiting factors and generate an estimation of kernel execution time with this model. Matrix multiplication (MM) and histogram generation (HG) in NVIDIA GTX680 GPU card were performed to verify our proposed model and showed an absolute error in prediction less than 6.8%.

A static analytical performance model for GPU kernel

Graphics processing units (GPUs) have shown increased popularity and play an important role as kind of coprocessor in heterogeneous co-processing environment. Heavily data parallel problems can be solved efficiently due to tens of thousands threads collaborative work in parallel GPU architecture. The achieved performance, therefore, depends on the capability of multiple threads in parallel collaboration. This paper, a static analytical kernel performance model (SAKP) was proposed to estimate the execution time of GPU kernel. Especially, a set of kernel and device features for target GPU is generated in the proposed model. We determine the performance limiting factors and generate an estimation of kernel execution time with this model. Matrix multiplication (MM) and histogram generation (HG) in NVIDIA GTX680 GPU card were performed to verify our proposed model and showed an absolute error in prediction less than 6.8%.

Time Estimation and Resource Minimization Scheme for Apache Spark and Hadoop Big Data Systems With Failures

Apache Spark and Hadoop are open source frameworks for big data processing, which have been adopted by many companies. In order to implement a reliable big data system that can satisfy processing target completion times, accurate resource provisioning and job execution time estimations are needed. In this paper, time estimation and resource minimization schemes for Spark and Hadoop systems are presented. The proposed models use the probability of failure in the estimations to more accurately formulate the characteristics of real big data operations. The experimental results show that the proposed Spark adaptive failure-compensation and Hadoop adaptive failure-compensation schemes improve the accuracy of resource provisions by considering failure events, which improves the scheduling success rate of big data processing tasks.

Comparative performance evaluation of MMSE-based speech enhancement techniques through simulation and real-time implementation

In this article, a generic model of minimum mean square error (MMSE) based speech enhancement technique has been presented and implemented on the hardware. Models for different MMSE-based methods were obtained by changing the gain function in the generic model. In all the methods, modified cascaded median-based noise estimation method has been used for noise estimation. Performances of all these MMSE-based methods were compared, among themselves and also with the spectral subtraction method for speech enhancement. The results have been evaluated using objective measures, subjective measure and composite objective measures for different noisy speech files. Results, in terms of objective evaluation parameters, indicated that the adaptive β-order MMSE method yielded better performance than the other methods. In subjective quality test (according to MOS listening test), β-order MMSE and adaptive β-order MMSE method yielded high scores. Real-time implementation has been carried out using TMS320C6416T DSP starter kit and code composer studio software. Estimation of memory consumption and execution time has been done for all the methods.

EOmesh: Combined Flow Balancing and Deterministic Routing for Reduced WCET Estimates in Embedded Real-Time Systems

The increasing performance needs in critical real-time embedded systems (CRTESs) can only be satisfied with the use of high-performance manycore processors. While NoC-based manycore systems are popular in the high-performance domain due to their high average performance, they challenge deriving tight worst-case execution time (WCET) estimates, as needed in CRTES. Weighted meshes have been proposed to alleviate NoCs pathological behavior—caused by large bandwidth imbalance—by making locally unbalanced arbitration decisions to reach globally balanced bandwidth. In this paper, we show that existing weighted mesh solutions do not completely remove unwanted imbalance, in particular for nodes subject to high congestion. We propose even/odd mesh (EOmesh), an approach that combines heterogeneous predictable routing and weight allocations that delivers near-optimal bandwidth allocation across cores without increasing NoC complexity. EOmesh, which can be implemented either by hardware means or by software means on top of regular weighted meshes, improves the average performance and WCET results of the reference weighted mesh design.

ESPRET: A tool for execution time estimation of manual test cases

Manual testing is still a predominant and an important approach for validation of computer systems, particularly in certain domains such as safety-critical systems. Knowing the execution time of test cases is important to perform test scheduling, prioritization and progress monitoring. In this work, we present, apply and evaluate ESPRET (EStimation and PRediction of Execution Time) as our tool for estimating and predicting the execution time of manual test cases based on their test specifications. Our approach works by extracting timing information for various steps in manual test specification. This information is then used to estimate the maximum time for test steps that have not previously been executed, but for which textual specifications exist. As part of our approach, natural language parsing of the specifications is performed to identify word combinations to check whether existing timing information on various test steps is already available or not. Since executing test cases on the several machines may take different time, we predict the actual execution time for test cases by a set of regression models. Finally, an empirical evaluation of the approach and tool has been performed on a railway use case at Bombardier Transportation (BT) in Sweden.

Estimation of travelling wave arrival time in longitudinal differential protections for multi‐terminal HVDC systems

Multi-terminal high-voltage DC (HVDC) systems equipped with HVDC breakers require that detection and clearing of faults occur within a few milliseconds. Although detection without communication is generally preferred with regard to detection delay and simplicity, detection schemes relying on communication can enhance the overall performance due to their reliability. One such principle is the travelling wave differential protection that is based on the comparison of waves at both ends rather than solely on current. However, as with conventional differential protections, some kind of synchronisation must be applied to properly compensate for the time difference between samples obtained at different geographical locations. In this study, a synchronisation algorithm is presented that applies real-time signal processing to automatically obtain a proper compensation between the signals from both ends. The algorithm is examined by injection of data samples, generated from the simulation of an external fault in a PSCAD/EMTDC model. Additionally, the suitability for a real-time implementation is evaluated by execution time estimation using a simulated DSP. The algorithm is found to perform well, not only in identifying the compensation delay but also in terms of computational efficiency, thus allowing for real-time application.

Scheduling in volunteer computing networks, based on neural network prediction of the job execution time

ABSTRACTImprovement of scheduling which broadly speaking means the distribution of jobs to volunteers is very important for improving the effectiveness of volunteer computing networks operating on the basis of computing resources connected to the Internet. The scheduling strategy based on the prediction of the job execution time is chosen as the main strategy to solve this problem. Suggested approach includes a neural network mechanism for computing the predictive estimate of the job execution time and a genetic algorithm for distributing jobs to volunteers with adjustment of parameters that makes it possible to respond to changes in the computing environment. The features of the approach are illustrated by computing experiments. In addition, we consider an example of the distribution of jobs to two volunteers for a project consisting of three applications. Even approximate (interval) estimates of job execution time allowed reducing the total execution time of the project and thereby optimising the computing process.

A multiswarm for composite SaaS placement optimization based on PSO

SummaryRecently, the demand for software as a service (SaaS) has witnessed an increasing interest, which has raised many challenges for SaaS management. One of these challenges is to deliver a high performance composite SaaS for users while optimizing the resources used. In this paper, we focus on the problem of SaaS placement. This problem occurs in the deployment of SaaS components in Cloud. It deals with the way a composite SaaS should be placed in a Cloud by the Cloud's provider such that its performance is optimal based on its estimated execution time. Previous work, including metaheuristic methods and particle swarm optimization, focuses on resolving the problem in a static environment. Moreover, in a Cloud data center, the workloads of applications and resources capacities keep changing over time, and the environment is dynamic, so the solution found for the initial placement may need to be reconfigured to maintain the SaaS performance and to optimize the resource used. As multiswarm technique has attracted increasing attention during the last decade, in this paper, we propose a new placement solution based on such technique enhanced with a cooperative learning strategy to cope with the dynamic aspect of the Cloud.

Scalable and precise estimation and debugging of the worst-case execution time for analysis-friendly processors: a comeback of model checking

Estimating the Worst-Case Execution Time (WCET) of an application is an essential task in the context of developing real-time or safety-critical software, but it is also a complex and error-prone process. Conventional approaches require at least some manual inputs from the user, such as loop bounds and infeasible path information, which are hard to obtain and can lead to unsafe results if they are incorrect. This is aggravated by the lack of a comprehensive explanation of the WCET estimate, i.e., a specific trace showing how WCET was reached. It is therefore hard to spot incorrect inputs and hard to improve the worst-case timing of the application. Meanwhile, modern processors have reached a complexity that refutes analysis and puts more and more burden on the practitioner. In this article we show how all of these issues can be significantly mitigated or even solved, if we use processors that are amenable to WCET analysis. We define and identify such processors, and then we propose an automated tool set which estimates a precise WCET without unsafe manual inputs, and also reconstructs a maximum-detail view of the WCET path that can be examined in a debugger environment. Our approach is based on Model Checking, which however is known to scale badly with growing application size. We address this issue by shifting the analysis to source code level, where source code transformations can be applied that retain the timing behavior, but reduce the complexity. Our experiments show that fast and precise estimates can be achieved with Model Checking, that its scalability can even exceed current approaches, and that new opportunities arise in the context of "timing debugging".

МЕТОДИКА ДИНАМИЧЕСКОГО АНАЛИЗА ВРЕМЕНИ ВЫПОЛНЕНИЯ ПРОГРАММ В ГЕТЕРОГЕННЫХ РАСПРЕДЕЛЕННЫХ ВЫЧИСЛИТЕЛЬНЫХ СРЕДАХ

The PURPOSE of the study is to develop the methodology for estimating the program execution time in a heterogeneous distributed computing environment. Today, the problem of obtaining such estimation is important and nontrivial in many practical applications related to the computation planning and resource allocation. METHODS. The study uses the method of frequency response characteristics that is based on the use of special tools for the dynamic analysis of programs. It has well proved itself in practice. RESULTS. A new methodology is proposed that provides program execution time estimation and takes into account the characteristics of reference and target computational nodes as well as software parameters that reflect the computational load on the components of these nodes. The estimates are calculated based on the amount of input data. This methodology has been successfully applied when analyzing the program execution for solving the problem of matrix multiplication. The paper provides the test cases of obtaining the time estimation of such problems solution in which the error is less than 10%. The obtained results demonstrate that the estimation error decreases as the dimension of matrices (both for integer and real values) grows. CONCLUSIONS. The proposed technique was used for the real job flow in the heterogeneous distributed computing environment based on the resources of the public access computer center “Irkutsk Supercomputer Center SB RAS”. Its application showed the significant improvement in the obtained estimations of the program execution time as compared with the estimates of the required time for solving the problems in users’ queries, as well as their values corrected on the basis of the computational history of executed jobs.

DTFM

Many-core processors are expected to be hardware targets to support the execution of real-time applications. In a many-core processor, cores communicate through a Network-On-Chip (NoC), which offers high bandwidth and scalability, but also introduces contentions leading to additional variability to task execution times. Such contentions also strongly increase the pessimistic trend of worst case execution time estimation. Consequently, modeling and analysis of network contentions interferences on many-core processors is a challenge to support real-time applications. In this article, we formalize a dual task and flow model called DTFM. From the specification of a real-time application composed of a set of tasks and their communication dependencies, DTFM allows us to compute flow requirements and to assess predictability of the tasks. DTFM is extensible enough to be adapted to various NoCs and task models, allowing designers to compare candidate software and NoC architectures. Furthermore, we introduce an original validation approach based on the cross-use of a task level real-time scheduling analysis tool and a cycle-accurate SystemC NoC simulator.

On the analysis of random replacement caches using static probabilistic timing methods for multi-path programs

Probabilistic hard real-time systems, based on hardware architectures that use a random replacement cache, provide a potential means of reducing the hardware over-provision required to accommodate pathological scenarios and the associated extremely rare, but excessively long, worst-case execution times that can occur in deterministic systems. Timing analysis for probabilistic hard real-time systems requires the provision of probabilistic worst-case execution time (pWCET) estimates. The pWCET distribution can be described as an exceedance function which gives an upper bound on the probability that the execution time of a task will exceed any given execution time budget on any particular run. This paper introduces a more effective static probabilistic timing analysis (SPTA) for multi-path programs. The analysis estimates the temporal contribution of an evict-on-miss, random replacement cache to the pWCET distribution of multi-path programs. The analysis uses a conservative join function that provides a proper over-approximation of the possible cache contents and the pWCET distribution on path convergence, irrespective of the actual path followed during execution. Simple program transformations are introduced that reduce the impact of path indeterminism while ensuring sound pWCET estimates. Evaluation shows that the proposed method is efficient at capturing locality in the cache, and substantially outperforms the only prior approach to SPTA for multi-path programs based on path merging. The evaluation results show incomparability with analysis for an equivalent deterministic system using an LRU cache. For some benchmarks the performance of LRU is better, while for others, the new analysis techniques show that random replacement has provably better performance.