Daniele Tessera

Workload Characterization Issues and Methodologies

The performance of any type of system cannot be determined without knowing the workload, that is, the requests being processed. Workload characterization consists of a description of the workload by means of quantitative parameters and functions; the objective is to derive a model able to show, capture, and reproduce the behavior of the workload and its most important features.

A hierarchical approach to workload characterization for parallel systems

Performance evaluation studies are to be an integral part of the design and tuning of parallel applications. We propose a hierarchical approach to the systematic characterization of the workload of a parallel system, to be kept as modular and flexible as possible. The methodology is based on three different, but related, layers: the application, the algorithm, and the routine layer. For each of these layers different characteristics representing functional, sequential, parallel, and quantitative descriptions have been identified. These characteristics are specified in a system independent way to clearly separate between the workload description and the architecture description. Taking also architectural and mapping features into consideration, the hierarchical workload characterization can be applied to any type of performance studies.

Medea: a tool for workload characterization of parallel systems

The Medea (MEasurements Description, Evaluation and Analysis) software tool provides a user-friendly environment for systematically applying workload characterization techniques to raw data produced by monitoring parallel programs. Medeas models are especially useful for program tuning and performance debugging, for testing alternative system configurations and for supporting benchmarking studies. >

Workload Characterization: A Survey Revisited

Workload characterization is a well-established discipline that plays a key role in many performance engineering studies. The large-scale social behavior inherent in the applications and services being deployed nowadays leads to rapid changes in workload intensity and characteristics and opens new challenging management and performance issues. A deep understanding of user behavior and workload properties and patterns is therefore compelling. This article presents a comprehensive survey of the state of the art of workload characterization by addressing its exploitation in some popular application domains. In particular, we focus on conventional web workloads as well as on the workloads associated with online social networks, video services, mobile apps, and cloud computing infrastructures. We discuss the peculiarities of these workloads and present the methodological approaches and modeling techniques applied for their characterization. The role of workload models in various scenarios (e.g., performance evaluation, capacity planning, content distribution, resource provisioning) is also analyzed.

Redistribution strategies for portable parallel FFT: a case study

The best approach to parallelize multidimensional FFT algorithms has long been under debate. Distributed transposes are widely used, but they also vary in communication policies and hence performance. In this work we analyze the impact of different redistribution strategies on the performance of parallel FFT, on various machine architectures. We found that some redistribution strategies were consistently superior, while some others were unexpectedly inferior. An in‐depth investigation into the reasons for this behavior is included in this work. Copyright

A methodology towards automatic performance analysis of parallel applications

Tuning and debugging the performance of parallel applications is an iterative process consisting of several steps dealing with identification and localization of inefficiencies, repair, and verification of the achieved performance. In this paper, we address the analysis of the performance of parallel applications from a methodological viewpoint with the aim of identifying and localizing inefficiencies. Our methodology is based on performance metrics and criteria that highlight the properties of the applications and the load imbalance and dissimilarities in the behavior of the processors. A few case studies illustrate the application of the methodology.

Workloads in the Clouds

Despite the fast evolution of cloud computing, up to now the characterization of cloud workloads has received little attention. Nevertheless, a deep understanding of their properties and behavior is essential for an effective deployment of cloud technologies and for achieving the desired service levels. While the general principles applied to parallel and distributed systems are still valid, several peculiarities require the attention of both researchers and practitioners. The aim of this chapter is to highlight the most relevant characteristics of cloud workloads as well as identify and discuss the main issues related to their deployment and the gaps that need to be filled.

Modeling and predicting temporal patterns of web content changes

The technologies aimed at Web content discovery, retrieval and management face the compelling need of coping with its highly dynamic nature coupled with complex user interactions. This paper analyzes the temporal patterns of the content changes of three major news websites with the objective of modeling and predicting their dynamics. It has been observed that changes are characterized by a time dependent behavior with large fluctuations and significant differences across hours and days. To explain this behavior, we represent the change patterns as time series. The trend and seasonal components of the observed time series capture the weekly and daily periodicity, whereas the irregular components take into account the remaining fluctuations. Models based on trigonometric polynomials and ARMA components accurately reproduce the dynamics of the empirical change patterns and provide extrapolations into the future to be used for forecasting.

Integration of a Compilation System and a Performance Tool: The HPF+ Approach

The performance of HPF codes is influenced by the characteristics of the parallel system and by the efficiency of the compilation system. Performance analysis has to take into account all these aspects. We present the integration of a compilation system with a performance analysis tool aimed at the evaluation of HPF+ codes. The analysis is carried out at the source level. The “costs” of the parallelization strategies applied by the compiler are also captured such that a comprehensive view of the performance is provided.

Parallel Performance Evaluation: The MEDEA Tool

The performance of parallel programs is influenced by the multiplicity of hardware and software components involved in their executions. Experimental approaches, where trace files collected at run-time by monitors are the basis of the analyses, allow a detailed evaluation of the performance. Quantitative as well as qualitative information related to the behavior of the programs are required. Medea is a parallel performance evaluation tool which provides various types of statistical and numerical techniques integrated with visualization facilities such that both quantitative and qualitative descriptions of the programs are obtained. A large variety of studies dealing with tuning, performance debugging, and code optimization profitably benefits of Medea.