CPU Demand Research Articles

Variational mode decomposition and sample entropy optimization based transformer framework for cloud resource load prediction

The efficient prediction of cloud resource demand plays a crucial role in resource allocation and scheduling in cloud data centers, helping to optimize resource utilization and improve service quality. However, accurately predicting cloud resource demand poses challenges due to the failure of prediction models in real-world scenarios, such as extreme load peaks, and the limitation of computation burden on the global characterization capability. To effectively handle single-variable cloud resource load time series with multidimensional hidden factors, we propose a sample entropy-optimized variational model decomposition transformer (VMDSE-Tformer) for cloud resource scheduling. Hereby, we decompose the time series through variational model decomposition, and then reconstruct the subsequence collection using sample entropy calculation. Then, we use a class Transformer framework with a multi-head self-attention mechanism to learn deep features and obtain encoding representations of each component sequence. We conduct sufficient experiments on three benchmark datasets by comparing them with five state-of-the-art models. Notably, the MAPE of VMDSE-Tformer is improved by about 60% compared to LSTNet. The results demonstrate the superior performance of our VMDSE-Tformer in terms of predicting task sequence intensity, CPU, and RAM resource demand. Therefore, VMDSE-Tformer can serve as a powerful and efficient tool to predict resource demand in cloud data centers, with implications for more effective resource management and service delivery.

Energy reference guidance for drag-modulated aerocapture

Aerocapture is a method of achieving orbit insertion via a single pass through the atmosphere of the central body. Single-event jettison drag modulation is a simple way of achieving control during atmospheric flight by effecting a discrete change in the aerodynamics of the vehicle. A novel guidance algorithm, energy reference guidance, is developed and implemented for a reference scenario of a small satellite aerocapture demonstration at Earth, and is compared to the baseline numerical predictor–corrector solution. The new approach is shown to achieve equivalent apoapsis targeting performance as the baseline algorithm with significantly lower CPU demand during atmospheric flight, although more onboard memory is required in exchange. The relationship between targeting performance and required memory is quantitatively explored for the new algorithm; the selected configuration generates approximately 3.3 MB of data, which is expected to be well within the capability of relevant avionics systems.

A novel approach for IoT tasks offloading in edge-cloud environments

Recently, the number of Internet of Things (IoT) devices connected to the Internet has increased dramatically as well as the data produced by these devices. This would require offloading IoT tasks to release heavy computation and storage to the resource-rich nodes such as Edge Computing and Cloud Computing. Although Edge Computing is a promising enabler for latency-sensitive related issues, its deployment produces new challenges. Besides, different service architectures and offloading strategies have a different impact on the service time performance of IoT applications. Therefore, this paper presents a novel approach for task offloading in an Edge-Cloud system in order to minimize the overall service time for latency-sensitive applications. This approach adopts fuzzy logic algorithms, considering application characteristics (e.g., CPU demand, network demand and delay sensitivity) as well as resource utilization and resource heterogeneity. A number of simulation experiments are conducted to evaluate the proposed approach with other related approaches, where it was found to improve the overall service time for latency-sensitive applications and utilize the edge-cloud resources effectively. Also, the results show that different offloading decisions within the Edge-Cloud system can lead to various service time due to the computational resources and communications types.

Decomposition Based Cloud Resource Demand Prediction Using Extreme Learning Machines

Cloud computing has drastically transformed the means of computing in past few years. Apart from numerous advantages, it suffers with a number of issues including resource under-utilization, load balancing and power consumption. The workload prediction is being widely explored to solve these issues using time series analysis regression and neural networks based models. The time series analysis based models are unable to capture the dynamics in the workload behavior whereas neural network based models offer better accuracy on the cost of high training time. This paper presents a workload prediction model based on extreme learning machines (ELM) whose learning time is very low and forecasts the workload more accurately. The performance of the model is evaluated over two real world cloud server workloads i.e. CPU and Memory demand traces of Google cluster and compared with predictive models based on state-of-art techniques including Auto Regressive Integrated Moving Average (ARIMA), Support Vector Regression (SVR), Linear Regression (LR), Differential Evolution (DE), Blackhole Algorithm (BhA), and Propagation (BP). It is observed that the proposed model outperforms the state-of-art techniques by reducing the mean prediction error up to 100% and 99% on CPU and memory request traces respectively.

Ensemble learning based predictive framework for virtual machine resource request prediction

The cloud service providers require a large number of computing resources to provide services on-demand that consume the electricity at large and leave high carbon footprints which must be minimized. A cloud system must optimally use its resources to achieve a low operational cost without degrading the quality of services. In this context, an ensemble learning based workload forecasting method is presented that uses extreme learning machines and their corresponding forecasts are weighted by a voting engine. A metaheuristic algorithm inspired by blackhole theory is used to select the optimal weights. The accuracy of the approach is tested on CPU and memory demand requests of Google cluster trace. The method is also compared with recent existing work in the literature on CPU utilization of Google cluster and PlanetLab traces. The results validate the superiority of the approach over existing methods with an improvement up to 99.20% in root mean squared error.

An Energy and SLA-Aware Resource Management Strategy in Cloud Data Centers

Reducing energy consumption of data centers is an important way for cloud providers to improve their investment yield, but they must also ensure that the services delivered meet the various requirements of consumers. In this paper, we propose a resource management strategy to reduce both energy consumption and Service Level Agreement (SLA) violations in cloud data centers. It contains three improved methods for subproblems in dynamic virtual machine (VM) consolidation. For making hosts detection more effective and improving the VM selection results, first, the overloaded hosts detecting method sets a dynamic independent saturation threshold for each host, respectively, which takes the CPU utilization trend into consideration; second, the underutilized hosts detecting method uses multiple factors besides CPU utilization and the Naive Bayesian classifier to calculate the combined weights of hosts in prioritization step; and third, the VM selection method considers both current CPU usage and future growth space of CPU demand of VMs. To evaluate the performance of the proposed strategy, it is simulated in CloudSim and compared with five existing energy–saving strategies using real-world workload traces. The experimental results show that our strategy outperforms others with minimum energy consumption and SLA violation.

Evaluation of a Straight-Ray Forward Model for Bayesian Inversion of Crosshole Ground Penetrating Radar Data

Bayesian inversion of crosshole ground penetrating radar (GPR) data is capable of characterizing the subsurface dielectric properties and qualifying the associated uncertainties. Markov chain Monte Carlo (MCMC) simulations within the Bayesian inversion usually require thousands to millions of forward model evaluations for the parameters to hit their posterior distributions. Therefore, the CPU cost of the forward model is a key issue that influences the efficiency of the Bayesian inversion method. In this paper we implement a widely used straight-ray forward model within our Bayesian inversion framework. Based on a synthetic unit square relative permittivity model, we simulate the crosshole GPR first-arrival traveltime data using the finite-difference time-domain (FDTD) and straight-ray solver, respectively, and find that the straight-ray simulator runs 450 times faster than its FDTD counterpart, yet suffers from a modeling error that is more than 7 times larger. We also perform a series of numerical experiments to evaluate the performance of the straight-ray model within the Bayesian inversion framework. With modeling error disregarded, the inverted posterior models fit the measurement data nicely, yet converge to the wrong set of parameters at the expense of unreasonably large number of iterations. When the modeling error is accounted for, with a quarter of the computational burden, the main features of the true model can be identified from the posterior realizations although there still exist some unwanted artifacts. Finally, a smooth constraint on the model structure improves the inversion results considerably, to the extent that it enhances the inversion accuracy approximating to those of the FDTD model, and further reduces the CPU demand. Our results demonstrate that the use of the straight-ray forward model in the Bayesian inversion saves computational cost tremendously, and the modeling error correction together with the model structure constraint are the necessary amendments that ensure that the model parameters converge correctly.

Effect of ordered set on feasibility analysis of static-priority system

Exact feasibility conditions for real-time system under preemptive fixed-priority systems are NP-hard and attempts have been made to lower the computation cost of such tests by restricting the set of scheduling points or applying lowest priority first approach. Under feasibility tests based on scheduling points, feasibility of a system is tested at all scheduling points starting with the lowest task period. However, it has been observed that it is very unlikely that the cumulative demand of a lower priority is addressed at such smaller points. Consequently, feasibility is tested at a huge number of scheduling points before the schedulability of a task is concluded. In this work, we show that it is more appropriate to test schedulability of a task at larger potential scheduling points instead of starting with smallest point. Since there is a logical OR involved in feasibility analysis at task level, schedulability of a task is affirmative when the CPU demand is fulfilled at any scheduling point or else the task is declared unschedulable. The complexity of our proposed solution is pseudo-polynomial, but our results are promising when the system utilization is low or when task periods vary largely.

APEM — Approximate Performance Evaluation for Multi-Core Computers

Mean Value Analysis (MVA) has long been a standard approach for performance analysis of computer systems. While the exact load-dependent MVA algorithm is an efficient technique for computer system performance modeling, it fails to address multi-core computer systems with Dynamic Frequency Scaling (DFS). In addition, the load-dependent MVA algorithm suffers from numerical difficulties under heavy load conditions. The goal of our paper is to find an efficient and robust method which is easy to use in practice and is also accurate for performance prediction for multi-core platforms. The proposed method, called Approximate Performance Evaluation for Multi-core computers (APEM), uses a flow-equivalent performance model designed specifically to address multi-core computer systems and identify the influence on the CPU demand of the effect of DFS. We adopt an approximation technique to estimate resource demands to parametrize MVA algorithms. To validate the application of our method, we investigate three case studies with extended TPC-W benchmark kits, showing that our method achieves better accuracy compared with other commonly used MVA algorithms. We compare the three different performance models, and we also extend our approach to multi-class models.

Effect of mobility and convection-dominated flow on evaluation of reservoir dynamic performance by fast marching method

The determination of optimum well locations and number of wells needed during green field development always comes with unprecedented challenges because of the geological uncertainty, and the non-linear relationship between the input and output variables associated with real reservoirs. These variables are key sources affecting the viability and validity of the results. Reservoir simulation is one of the least uncertain and most reliable prediction tools for dynamic performance of any reservoir. As field development progresses, more information becomes available, enabling us to continually update and, if needed, correct the reservoir description. The simulator can then be used to perform a variety of exercises or scenarios, with the goal of optimizing field development and operation strategies. Optimizing well numbers or locations under such geological uncertainty is achieved by using a reservoir simulator under several geological realizations, and these require multiple reservoir simulations to estimate the field performance for a given well configuration at a given location. Using reservoir simulation becomes impractical when dealing with real field cases incorporating multi-million cells because of the associated CPU demand constraints (Bouzarkouna et al. 2011). For instance, to determine the optimum well locations in a giant field that will result in the most efficient production rate scenario, one requires a large number of simulation runs for different realizations and well configurations. A large amount of runs is technically difficult to achieve even if we have access to super computers. The fast marching method (FMM), which is based on a solution of Eikonal equation, can be used to find the optimum well locations in a green oil field by tracking the pressure distribution in the reservoir. The FMM will enable us to calculate the radius of investigation or pressure front as a function of time without running any simulation and with a high degree of accuracy under primary depletion conditions. The main purpose of this paper is to study the effect of mobility on FMM and extend the investigation of its validity to include two-phase flow and convection-dominated flow and evaluate the ability of the methodology to predict the dynamic performance of the reservoir during pseudo-steady-state flow regime.

Fast and accurate determination of the relative binding affinities of small compounds to HIV-1 protease using non-equilibrium work.

The fast pulling ligand (FPL) out of binding cavity using non-equilibrium molecular dynamics (MD) simulations was demonstrated to be a rapid, accurate and low CPU demand method for the determination of the relative binding affinities of a large number of HIV-1 protease (PR) inhibitors. In this approach, the ligand is pulled out of the binding cavity of the protein using external harmonic forces, and the work of pulling force corresponds to the relative binding affinity of HIV-1 PR inhibitor. The correlation coefficient between the pulling work and the experimental binding free energy of R=-0.95 shows that FPL results are in good agreement with experiment. It is thus easier to rank the binding affinities of HIV-1 PR inhibitors, that have similar binding affinities because the mean error bar of pulling work amounts to δW=7%. The nature of binding is discovered using the FPL approach. © 2016 Wiley Periodicals, Inc.

Switched reluctance machines control with a minimized sampling frequency

This paper is focused on reducing the Switched Reluctance Machines (SRMs) control sampling frequency in order to save processor real time resources, while keeping the stability and also the performance, in terms of average torque and torque ripple. Reducing the CPU cost either by implementing the control algorithm in a less performing CPU or more importantly reducing the percentage of the CPU demand is an attractive goal, especially for the electrical vehicle industry from where the SRM used in this research has been designed for. Once low sampling periods are applied in the current loop, a strong degradation in the averaged torque and torque ripple arises. Such problem degenerates with the speed, becoming unbearable at high speeds and eventually making the control unstable. In this paper two solutions are proposed. The first one, which is just software feasible, consists on anticipating the voltage supply in order to tackle the noncoincident calculated turn on and off angles and the actual sampling instants. The second solution, which must be implemented at a very low hardware level, uses a basic function to allow the process to emulate continuous control and therefore independent of the sampling instants. Finally, experimental results on a 8/6 SRM illustrate the validity of the novel strategies in terms of average torque performance and torque ripple minimization.

Hierarchical Architecture for HEVC Parallel Encoder

The Emerging High Efficiency Video Coding (HEVC) standard offers a significantly better compression rate and higher video quality. HEVC encoders create very high CPU demand, and it is hard for a single core computer to deal with such complex coding computations. To process the high workload of HEVC coding for large-scale video data, the current HEVC draft contains several parallelizing approaches: tile-based parallelization and WPP-level parallelization. In this paper, we adopt additional data parallelism, GOP partitioning, to implement a parallel encoder. We propose the scalable cluster architecture of the HEVC encoder to achieve scalability and a high encoding speed by combining two levels of parallelism, GOP-level parallelism and the HEVC parallel method. The proposed scheme can reduce the large encoding time and significantly improve the coding efficiency. The proposed scalable cluster system is very suitable for high-resolution video such as 4K or 8K containing large amounts of video data.

Modeling the effect of clay drapes on pumping test response in a cross-bedded aquifer using multiple-point geostatistics

Summary This study investigates whether fine-scale clay drapes can cause an anisotropic pumping test response at a much larger scale. A pumping test was performed in a sandbar deposit consisting of cross-bedded units composed of materials with different grain sizes and hydraulic conductivities. The measured drawdown values in the different observation wells reveal an anisotropic or elliptically-shaped pumping cone. The major axis of the pumping ellipse is parallel with the strike of cm to m-scale clay drapes that are observed in several outcrops. To determine (1) whether this large-scale anisotropy can be the result of fine-scale clay drapes and (2) whether application of multiple-point geostatistics can improve interpretation of pumping tests, this pumping test is analyzed with a local 3D groundwater model in which fine-scale sedimentary heterogeneity is modelled using multiple-point geostatistics. To reduce CPU and RAM demand of the multiple-point geostatistical simulation step, edge properties indicating the presence of irregularly-shaped surfaces are directly simulated. Results show that the anisotropic pumping cone can be attributed to the presence of the clay drapes. Incorporating fine-scale clay drapes results in a better fit between observed and calculated drawdowns. These results thus show that fine-scale clay drapes can cause an anisotropic pumping test response at a much larger scale and that the combined approach of multiple-point geostatistics and cell edge properties is an efficient method for integrating fine-scale features in larger scale models.

Direct Multiple-Point Geostatistical Simulation of Edge Properties for Modeling Thin Irregularly Shaped Surfaces

Thin, irregularly shaped surfaces such as clay drapes often have a major control on flow and transport in heterogeneous porous media. Clay drapes are often complex, curvilinear three-dimensional surfaces and display a very complex spatial distribution. Variogram-based stochastic approaches are also often not able to describe the spatial distribution of clay drapes since complex, curvilinear, continuous, and interconnected structures cannot be characterized using only two-point statistics. Multiple-point geostatistics aims to overcome the limitations of the variogram. The premise of multiple-point geostatistics is to move beyond two-point correlations between variables and to obtain (cross) correlation moments at three or more locations at a time using training images to characterize the patterns of geological heterogeneity. Multiple-point geostatistics can reproduce thin irregularly shaped surfaces such as clay drapes, but this is often computationally very intensive. This paper describes and applies a methodology to simulate thin, irregularly shaped surfaces with a smaller CPU and RAM demand than the conventional multiple-point statistical methods. The proposed method uses edge properties for indicating the presence of thin irregularly shaped surfaces. Instead of pixel values, edge properties indicating the presence of irregularly shaped surfaces are simulated using snesim. This method allows direct simulation of edge properties instead of pixel properties to make it possible to perform multiple-point geostatistical simulations with a larger cell size and thus a smaller computation time and memory demand. This method is particularly valuable for three-dimensional applications of multiple-point geostatistics.

CPU demand for web serving

Managing the resources in a large Web serving system requires knowledge of the resource needs for service requests of various types. In order to investigate the properties of Web traffic and its de...

A regression-based analytic model for capacity planning of multi-tier applications

The multi-tier implementation has become the industry standard for developing scalable client-server enterprise applications. Since these applications are performance sensitive, effective models for dynamic resource provisioning and for delivering quality of service to these applications become critical. Workloads in such environments are characterized by client sessions of interdependent requests with changing transaction mix and load over time, making model adaptivity to the observed workload changes a critical requirement for model effectiveness. In this work, we apply a regression-based approximation of the CPU demand of client transactions on a given hardware. Then, we use this approximation in an analytic model of a simple network of queues, each queue representing a tier, and show the approximation's effectiveness for modeling diverse workloads with a changing transaction mix over time. Using two case studies, we investigate factors that impact the efficiency and accuracy of the proposed performance prediction models. Experimental results show that this regression-based approach provides a simple and powerful solution for efficient capacity planning and resource provisioning of multi-tier applications under changing workload conditions.

CPU demand for web serving: Measurement analysis and dynamic estimation

Managing the resources in a large Web serving system requires knowledge of the resource needs for service requests of various types. In order to investigate the properties of Web traffic and its demand, we collected measurements of throughput and CPU utilization and performed some data analyses. First, we present our findings in relation to the time-varying nature of the traffic, the skewness of traffic intensity among the various types of requests, the correlation among traffic streams, and other system-related phenomena. Then, given such nature of web traffic, we devise and implement an on-line method for the dynamic estimation of CPU demand. Assessing resource needs is commonly performed using techniques such as off-line profiling, application instrumentation, and kernel-based instrumentation. Little attention has been given to the dynamic estimation of dynamic resource needs, relying only on external and high-level measurements such as overall resource utilization and request rates. We consider the problem of dynamically estimating dynamic CPU demands of multiple kinds of requests using CPU utilization and throughput measurements. We formulate the problem as a multivariate linear regression problem and obtain its basic solution. However, as our measurement data analysis indicates, one is faced with issues such as insignificant flows, collinear flows, space and temporal variations, and background noise. In order to deal with such issues, we present several mechanisms such as data aging, flow rejection, flow combining, noise reduction, and smoothing. We implemented these techniques in a Work Profiler component that we delivered as part of a broader system management product. We present experimental results from using this component in scenarios inspired by real-world usage of that product.

Efficient gradual deformation using a streamline-based proxy method

The application of state-of-the-art history matching methods to large heterogeneous reservoirs is hampered by two main problems: (1) reservoir models should be constrained jointly to dynamic data and a large variety of geological continuity information, and (2) CPU demand should not be prohibitive for large models. This paper proposes a method contributing to alleviating these two main concerns. By combining three existing ideas, namely gradual deformation, multiple-point geostatistics and a fast streamline-based history matching method, an algorithm is proposed that could honor a large variety of geological scenarios and is obtained with a limited amount of flow simulations. The method expands on the traditional gradual deformation methodology in three ways: (1) multiple-point geostatistics is used to generate models that are geologically more realistic than traditional variogram-based models, (2) a streamline simulator is used to define “zones-of-influence” of producers in order to locally deform an initial model toward jointly matching a large amount of wells and most importantly (3) the number of flow calculations is limited by defining a proxy to the streamline simulator in terms of streamline-based harmonic averages. Using synthetic examples of increasing complexity, the method's efficiency and generality is assessed.

Chemical transferability of single- and multiple-scattering EXAFS Debye-Waller factors.

Single- and multiple-scattering EXAFS Debye-Waller factors are amplitude reduction parameters that appear in the EXAFS chi(k) equation accounting for the structural and thermal disorder of a given sample. These parameters must be known accurately in order to obtain quantitative agreement between theory and experiment. Since experimental data can only support a limited number of fitted parameters these factors must be known from another source. Although various approaches have been considered in the past with a variety of results, the self-consistent ab initio Density functional theory stands for the most accurate and reliable method regardless of molecular symmetry or other specific sample requirements. Since DFT scales as N3 where N is the number of atomic basis set, an ab initio calculation on a large structure is not feasible due to enormous CPU demand and in many cases due to hard energy/geometry convergence. In this paper we present two ways of overcoming this problem. Both they use the idea that by reducing the structure, the DWFs are still chemically transferable. In order to test this we use the Zn tetraimidazole. This molecule represents typical metalorganic ring samples that can be seen in active sites of metaloproteins. Results are compared to experimental EXAFS spectra.