Pace—A Toolset for the Performance Prediction of Parallel and Distributed Systems

Abstract

This paper describes a methodology that provides detailed predictive performance information throughout the software design and implementation cycles. It is structured around a hierarchy of performance models that describe the computing system in terms of its software, parallelization, and hardware components. The methodology is illustrated with an implementation, the performance analysis and characterization environment (PACE) system, which provides information concerning execution time, scalability, and resource use. A principal aim of the work is to provide a capability for rapid calculation of relevant performance numbers without sacrificing accuracy. The predictive nature of the approach provides both pre and post implementation analyses and allows implementation alternatives to be explored prior to the commitment of an application to a system. Because of the relatively fast analysis times, these techniques can be used at runtime to assist in application steering and scheduling with reference to dynamically changing systems and metacomputing.

Towards Solving the Problem of Virtual Machine Placement in Cloud Computing: A Job Classification Approach

“…• Code analysis (Nudd et al, 2000;Reistad and Gifford, 1994) • Analytic benchmarking/code profiling (Yang et al, 1993) • Historical/Statistical prediction (Sanjay and Vadhiyar, 2008;Iverson et al, 1996) • Empirical analysis (Berman et al, 2005) Using these solutions, or by executing part of the code, other information related to user jobs can be obtained. In addition, there exists many studies which proposed models to estimate specific parameters (e.g., Cycle Per Instruction (CPI) (Chen and John, 2011;Intel Corporation, 2008), Memory Access Per Instruction (MPI) (Zhang and Chang, 2014) and estimated bandwidth (Zhu et al, 2012).…”

A Taxonomy of Workflow Management Systems for Grid Computing

“…In the analytical modeling approach [31][36] [93], a scheduler predicts the performance of tasks in workflow on a given set of resources based on an analytic metric. For example, within GrADS [31], two types of performance models are developed, namely memory hierarchy performance model and computational model.…”

“…ARMS has integrated Titan [111], which utilizes performance data obtained from PACE [93], a toolset for resource performance and usage analysis, with iterative heuristic algorithms to minimize the makespan and idle time of a grid resource. PACE can exact control flow, and use an analytical model approach based on queuing theory, to predict application performance on a given set of resources such as time, scalability and system resource usage.…”

Performance prediction and procurement in practice: assessing the suitability of commodity cluster components for wavefront codes

“…Wisconsin Wind Tunnel [19], PRO-TEUS [3] and the PACE toolkit [7,12] ) were originally envisaged as a mechanism to lower the burden of performance modelling by eliminating the need to manually inspect application source code. The automated replay of applications either in source or binary form allowed developers and performance modellers alike to experiment with performance by making direct changes to the application and simulating execution without requiring direct access to the specific machine in question.…”