Search

An execution trace is a model of a single system behavior. Execution traces occur everywhere in the system's lifecycle as they can typically be produced by executable models, by prototypes of (sub)systems, and by the system itself during its operation. An execution trace can be visualized and analyzed with various techniques, providing insight into the dynamic behavior, performance, bottlenecks, etc., of the system. In this paper, we present the Trace tool of the Eclipse Trace4cps project for the visualization and analysis of execution traces. A prominent application is the trace-based performance engineering of embedded or cyber-physical systems. Performance is an important system quality, as it can give a competitive advantage. Reasoning about system-level performance in such systems, however, is hard due to its cross-cutting nature. We show how the Trace tool can support this by various examples. Performance engineering is not the only application of the Trace tool, however: it supports system analysis in a wide range of situations. [ABSTRACT FROM AUTHOR]

The expanding amounts of information created by Internet of Things (IoT) devices places a strain on cloud computing, which is often used for data analysis and storage. This paper investigates a different approach based on edge cloud applications, which involves data filtering and processing before being delivered to a backup cloud environment. This Paper suggest designing and implementing a low cost, low power cluster of Single Board Computers (SBC) for this purpose, reducing the amount of data that must be transmitted elsewhere, using Big Data ideas and technology. An Apache Hadoop and Spark Cluster that was used to run a test application was containerized and deployed using a Raspberry Pi cluster and Docker. To obtain system data and analyze the setup's performance a Prometheusbased stack monitoring and alerting solution in the cloud based market is employed. This Paper assesses the system's complexity and demonstrates how containerization can improve fault tolerance and maintenance ease, allowing the suggested solution to be used in industry. An evaluation of the overall performance is presented to highlight the capabilities and limitations of the suggested architecture, taking into consideration the suggested solution's resource use in respect to device restrictions. [ABSTRACT FROM AUTHOR]

Performance and scalability issues in Enterprise applications are leading to multimillion dollar losses to Businesses across the globe. Such issues lead to bad customer experiences, lower employee productivity, reduced operational efficiency and ultimately impact the business. Regular performance audit of the IT systems across the entire technology stack is a proven method to identify and fix the pending and chronic performance issues, thereby improving performance. This complements other methods like performance monitoring and testing. A large enterprise can have thousands of applications and identifying the critical applications to undergo regular performance audit can be challenging, particularly in enterprises that have many teams. Selecting the right application set can be a subjective, unstructured and error prone process. Wrong selection can lead to sub-optimal coverage and no reduction in the risk profile of the applications that matter the most. How should enterprises address this challenge and objectively select an optimal set of applications for the audit activity? Analytical Hierarchy Process (AHP) is discussed in this paper as a technique for introducing objectivity and dealing with the complex decision-making process while selecting the application set. The paper applies the technique to a typical banking enterprise and explains how the critical applications can be selected. [ABSTRACT FROM AUTHOR]

This special issue contains four papers selected from the 26th UK Performance Engineering Workshop (UKPEW), which was held during 8–9 July 2010 in the Department of Computer Science at the University of Warwick. UKPEW is the leading UK forum for the presentation of work relating to all aspects of performance modelling and the analysis of computer and telecommunication systems. This year's selected papers demonstrate the variety of topics on offer at UKPEW: performance issues in cloud computing, energy-aware routing in packet networks, source location privacy and multi-hop ad hoc networks for wireless networks. [ABSTRACT FROM AUTHOR]

This special issue includes research papers related to a workshop held at the Royal Society on 21st of September 2010, honouring Erol Gelenbe on his 65th birthday. The workshop brought together some 40 participants including some of Erol's former and current PhD students, colleagues and friends with whom he has interacted over the years, both in Europe and the USA. The papers focus on probability models in computer science, at the core of the field of computer performance engineering, which Erol pioneered. [ABSTRACT FROM AUTHOR]

In this paper, we address the question of how to automatically map computational kernels to highly efficient code for a wide range of computing platforms and establish the correctness of the synthesized code. More specifically, we focus on two fundamental problems that software developers are faced with: performance portability across the ever-changing landscape of parallel platforms and correctness guarantees for sophisticated floating-point code. The problem is approached as follows: We develop a formal framework to capture computational algorithms, computing platforms, and program transformations of interest, using a unifying mathematical formalism we call operator language (OL). Then we cast the problem of synthesizing highly optimized computational kernels for a given machine as a strongly constrained optimization problem that is solved by search and a multistage rewriting system. Since all rewrite steps are semantics preserving, our approach establishes equivalence between the kernel specification and the synthesized program. This approach is implemented in the SPIRAL system, and we demonstrate it with a selection of computational kernels from the signal and image processing domain, software-defined radio, and robotic vehicle control. Our target platforms range from mobile devices, desktops, and server multicore processors to large-scale high-performance and supercomputing systems, and we demonstrate performance comparable to expertly hand-tuned code across kernels and platforms. [ABSTRACT FROM AUTHOR]

In this paper we discuss a new approach to Perfor mance Engineering called the Object-Oriented Per formance Engineering Approach (OOPEA). OOPEA is an innovative approach to performance modeling, simulation and design where an object-oriented per formance analysis model, a discrete-event simula tion model and an operational model are integrated into a modeling framework. We emphasize simula tion in this paper and present a case study on the UICCELL II (University of Illinois at Chicago CELLular II switching system), a software simula tor of a cellular phone system. The system can be divided into three main components: the Mobile Switching Center, Base Site, and Mobile Telephone Unit, which have client-server relationships and communicate with each other using IPC facilities, via message passing. The UICCELL II is implemented in a distributed environment on SUN SPARC 1 and 10 workstations using an Ethernet LAN. Sim ulation is used to examine the dynamic behavior of the system, and helps to capture the performance constraints, mainly the contention for resources. [ABSTRACT FROM PUBLISHER]

Block variants of the Jacobi-Davidson method for computing a few eigenpairs of a large sparse matrix are known to improve the robustness of the standard algorithm when it comes to computing multiple or clustered eigenvalues. In practice, however, they are typically avoided because the total number of matrix-vector operations increases. In this paper we present the implementation of a block Jacobi-Davidson solver. By detailed performance engineering and numerical experiments we demonstrate that the increase in operations is typically more than compensated by performance gains through better cache usage on modern CPUs, resulting in a method that is both more efficient and robust than its single vector counterpart. The steps to be taken to achieve a block speedup involve both kernel optimizations for sparse matrix and block vector operations, and algorithmic choices to allow using blocked operations in most parts of the computation. We discuss the aspect of avoiding synchronization in the algorithm and show by numerical experiments with our hybrid parallel implementation that a significant speedup through blocking can be achieved for a variety of matrices on up to 5 120 CPU cores as long as at least about 20 eigenpairs are sought. [ABSTRACT FROM AUTHOR]

This study aims to identify the most effective input parameters for performance modelling of container-based web systems. We introduce a method using queueing Petri nets to model web system performance for containerized structures, leveraging prior measurement data for resource demand estimation. This approach eliminates intrusive interventions in the production system. Our research evaluates the accuracy of various formal estimation methods, pinpointing the most suitable for container environments. With the use of a stock exchange web system benchmark for data collection and simulation verification, our findings reveal that the proposed method ensures precise response time parameter accuracy for such architectural configurations. [ABSTRACT FROM AUTHOR]

Developments in numerical simulation of flows and high-performance computing influence one another. More detailed simulation methods create a permanent need for more computational power, while new hardware developments often require changes to the software to exploit new hardware features. This dependency is very pronounced in the case of vector-units which are featured by all modern processors to increase their numerical throughput but require vectorization of the software to be used efficiently. We study the vectorization of a simulation method that exhibits an inherent level of vector-parallelism. This is of particular interest as SIMD operations will hopefully be available with std::simd in a future C++ standard. The simulation method considered here results in the simultaneous solution of multiple sparse linear systems of equations which only differ by their main diagonal and right-hand sides. Such structure arises in the simulation of unsteady flow in turbomachinery by means of a frequency domain approach called harmonic balance. [ABSTRACT FROM AUTHOR]

The desire for high performance on scalable parallel systems is increasing the complexity and tunability of MPI implementations. The MPI Tools Information Interface (MPI_T) introduced as part of the MPI 3.0 standard provides an opportunity for performance tools and external software to introspect and understand MPI runtime behavior at a deeper level to detect scalability issues. The interface also provides a mechanism to fine-tune the performance of the MPI library dynamically at runtime. In this paper, we propose an infrastructure that extends existing components — TAU, MVAPICH2, and BEACON to take advantage of the MPI_T interface and offer runtime introspection, online monitoring, recommendation generation, and autotuning capabilities. We validate our design by developing optimizations for a combination of production and synthetic applications. Using our infrastructure, we implement an autotuning policy for AmberMD (a molecular dynamics package) that monitors and reduces the internal memory footprint of the MVAPICH2 MPI library without affecting performance. For applications such as MiniAMR whose collective communication is latency sensitive, our infrastructure is able to generate recommendations to enable hardware offloading of collectives supported by MVAPICH2. By implementing this recommendation, the MPI time for MiniAMR at 224 processes reduces by 15%. [ABSTRACT FROM AUTHOR]

This paper describes a model-driven performance engineering framework for applications embedded in individual nodes of wireless sensor networks (WSNs). The framework, called Moppet, is designed for developers, even non-programmers, to rapidly implement WSN applications, estimate their performance and feedback the estimated performance results for customizing their design/implementation. By leveraging the notion of feature modeling, Moppet allows developers to graphically and intuitively specify features (e.g. functionalities and configuration policies) required in their applications. It also validates a set of constraints among features and generates a lightweight application code. Moreover, with event calculus and network calculus, Moppet estimates a WSN application's performance without generating its code nor running it on simulators and real networks. It can approximate data yield, data transmission latency and network lifetime as well as each node's power and bandwidth consumption. Evaluation results show that, in a large-scale WSN of 400 nodes, Moppet's performance estimation is 46% more efficient than empirical performance measurement and its estimation error is lower than 10%. Moppet scales well to network size with respect to estimation efficiency and accuracy. [ABSTRACT FROM AUTHOR]

Performance engineering can be described as a collection of techniques and methodologies whose aim is to provide reliable prediction, measurement and validation of the performance of applications on a variety of computing platforms. This paper reviews techniques for performance estimation and performance engineering developed at the University of Southampton and presents application case studies in task scheduling for engineering meta-applications, and capacity engineering for a financial transaction processing system. These show that it is important to describe performance in terms of a resource model, and that the choice of models may have to trade accuracy for utility in addressing the operational issues. We then present work from the on-going EU funded Grid project GRIA, and show how lessons learned from the earlier work have been applied to support a viable business model for Grid service delivery to a specified quality of service level. The key in this case is to accept the limitations of performance estimation methods, and design business models that take these limitations into account rather than attempting to provide hard guarantees over performance. We conclude by identifying some of the key lessons learned in the course of our work over many years and suggest possible directions for future investigations. Copyright © 2005 John Wiley & Sons, Ltd. [ABSTRACT FROM AUTHOR]

This paper includes parametric study and optimization of non-linear ceiling fan blades by combining the techniques of Design of Experiments (DOE), Response Surface Methods (RSM) and Computational Fluid Dynamics (CFD). Specifically, the nonlinear (elliptical) planform shape of ceiling fan blade is investigated in conjunction with blade tip width, root and tip angle of attack. Sixteen cases are designed for three blade ceiling fan using two level full factorial model. The flow field is modeled using Reynolds-Averaged-Navier-Stokes approach. The performance variables used to formulate a multi-objective optimization problem are volumetric flow rate, torque and energy efficiency. Response Surface Method is used to generate the optimized design for non-linear ceiling fan blade profile. The results reveal that the interactions between the design variables play a significant role in determining the performance. It is concluded that the nonlinear forward sweep has a moderate effect on response parameters. [ABSTRACT FROM AUTHOR]

A data analytics application submitted to a Spark cluster often has to finish executing by a specified time target. To use cluster resources effectively, the key challenge is having the ability to gain quick insights on how the execution time of any given application is likely to be impacted by the resources allocated to the application, e.g., the number of Spark executor cores assigned, and the size of the data to be processed. Such insights can be used to quickly estimate the required resources and configure a Spark application for a desired execution time using the least amount of resources. Our paper proposes an automated execution time estimation approach called PERIDOT that involves executing a given application under a fixed resource setting with two different-sized, small subsets of its input data to offer fast, lightweight execution time predictions. It analyzes logs from these two executions to estimate the dependencies between internal stages of the application. Information on these dependencies combined with knowledge of Spark's data partitioning mechanisms is used to derive an analytic model that can estimate execution times for other resource settings and input data sizes. Our results from a wide range of applications and multiple Spark clusters show that PERIDOT can accurately estimate the execution time of an application from limited historical data, and suggest the minimum amount of resources required to closely meet an execution time target.

Reuse-based software development provides an opportunity for better quality and increased productivity in the software products. One of the most critical aspects of the quality of a software system is its performance. The systematic application of software performance engineering techniques throughout the development process can help to identify design alternatives that preserve desirable qualities such as extensibility and reusability while meeting performance objectives. In the present scenario, most of the performance failures are due to a lack of consideration of performance issues early in the development process, especially in the design phase. These performance failures results in damaged customer relations, lost productivity for users, cost overruns due to tuning or redesign, and missed market windows. In this paper, we propose UML based Performance Models for design assessment in a reuse based software development scenario. [ABSTRACT FROM AUTHOR]

Abstract: Several techniques and applications have been proposed to aid the decision taking process in the system performance domain. Most of these techniques have depicted the performance model of systems through annotations of performance measurements coming from specific software descriptive syntactical languages. However, the semantic representation of performance information provides the possibility of its ulterior machine-processable logical interpretation and therefore the applicability of inference rules about a particular domain. Moreover, ontologies ease the interchange and reuse of knowledge of particular domains, e.g. system performance. In this work, we propose a performance ontology together with the system performance analysis technique as an example of framework building for intelligent applications based on semantic web. The paper also shows the construction of performance rules through OWL to automatically infer new performance constraints and QoS knowledge about the system on execution. [Copyright &y& Elsevier]

The development of complex software systems satisfying performance requirements is achievable only spending careful attention to performance goals throughout the lifecycle, and especially from its very beginning. Unified modeling language (UML) is quickly becoming a standard notation for specification and design of software systems. UML offers several diagrams for separating concerns of different system views, and this feature is helpful to derive early performance models that take into account combined data from these diagrams. In this paper, we introduce a methodology Performance Incremental Validation in UML (PRIMA-UML) aimed at generating a queueing network based performance model from UML diagrams that are usually available early in the software lifecycle. PRIMA-UML is incremental in that it combines information extracted from (and annotated into) different UML diagrams to piecewise build the performance model. Besides, this is not a “black box” approach, as the methodology is open to embed information coming from other UML diagrams (possibly in late lifecycle phases) for detailing, refining or domain tailoring the performance model. This work is a contribute to encompass the performance validation task as an integrated activity within the development process of complex systems. We apply the methodology to a quite simple example to show how effective it can be to get early performance insights. [Copyright &y& Elsevier]

A high-performance implementation of a multiphase lattice Boltzmann method based on the conservative Allen-Cahn model supporting high-density ratios and high Reynolds numbers is presented. Meta-programming techniques are used to generate optimized code for CPUs and GPUs automatically. The coupled model is specified in a high-level symbolic description and optimized through automatic transformations. The memory footprint of the resulting algorithm is reduced through the fusion of compute kernels. A roofline analysis demonstrates the excellent efficiency of the generated code on a single GPU. The resulting single GPU code has been integrated into the multiphysics framework waLBerla to run massively parallel simulations on large domains. Communication hiding and GPUDirect-enabled MPI yield near-perfect scaling behavior. Scaling experiments are conducted on the Piz Daint supercomputer with up to 2048 GPUs, simulating several hundred fully resolved bubbles. Further, validation of the implementation is shown in a physically relevant scenario--a three-dimensional rising air bubble in water. [ABSTRACT FROM AUTHOR]

General matrix-matrix multiplications with double-precision real and complex entries (DGEMM and ZGEMM) in vendor-supplied BLAS libraries are best optimized for square matrices but often show bad performance for tall & skinny matrices, which are much taller than wide. NVIDIA's current CUBLAS implementation delivers only a fraction of the potential performance as indicated by the roofline model in this case. We describe the challenges and key characteristics of an implementation that can achieve close to optimal performance. We further evaluate different strategies of parallelization and thread distribution and devise a flexible, configurable mapping scheme. To ensure flexibility and allow for highly tailored implementations we use code generation combined with autotuning. For a large range of matrix sizes in the domain of interest we achieve at least 2/3 of the roofline performance and often substantially outperform state-of-the art CUBLAS results on an NVIDIA Volta GPGPU. [ABSTRACT FROM AUTHOR]