The elusive goal of workload characterization

Characterizing the I/O requirements of parallel applications that manipulate huge amounts of data, such as scienti c codes, is critical to the achievement of good application performance and to the e ective use of parallel systems. In this paper we formulate a compositional stochastic model of the behavior of I/O intensive scienti c applications, which can be applied at various granularity levels of characterization. Based on the observation of the interaction of CPU and I/O activity in a set of scienti c codes, we exercise the stochastic model. The model is in excellent agreement with experimental data in the set of the examined codes. For the model to be used for performance prediction, we also propose a set of functional forms for forecasting the scalability of computation and I/O components of an application. The complexity of the functional form adopted a ects the accuracy of performance prediction.

ECMS 2014 Proceedings edited by: Flaminio Squazzoni, Fabio Baronio, Claudia Archetti, Marco Castellani, 2014

Multicore architectures are now available for a wide range of high performance applications, ranging from embedded systems to large scale servers deployed in cloud environments. Multicore architectures are usually subject to two conflicting goals: obtaining a full utilization of the cores while achieving given performance objectives, such as throughput, response time or reduced energy consumption. Moreover, there is a strong interdependence between the software characteristics of the applications, and the underlying CPU architecture. In this scenario, simulation and analytical techniques can provide solid tools to properly design the considered class of systems: however, properly characterize the workload on multithreaded application in multicore environment is not an easy task, and thus is an hot research topic. In this paper we present several models, of increasing complexity, that can characterize multithreaded applications running on multicore architectures.

… Computing, 1997. Proceedings. The Sixth IEEE …, 1996

We develop a workload model based on observations of parallel computers at the San Diego Supercomputer Center and the Cornell Theory Center. This model gives us insight into the performance of strategies for scheduling moldable jobs on space-sharing parallel computers. We find that Adaptive Static Partitioning (ASP), which has been reported to work well for other workloads, does not perform as well as strategies that adapt better to system load. The best of the strategies we consider is one that explicitly reduces allocations when load is high (a variation of Sevcik's A+ strategy (1989))

International Journal of Embedded and Real-Time Communication Systems, 2013

Modern mobile nomadic devices for example internet tablets and high end mobile phones support diverse distributed and stand-alone applications that were supported by single devices a decade back. Furthermore the complex heterogeneous platforms supporting these applications contain multi-core processors, hardware accelerators and IP cores and all these components can possibly be integrated into a single integrated circuit (chip). The high complexity of both the platform and the applications makes the design space very complex due to the availability of several alternatives. Therefore the system designer must be able to quickly evaluate the performance of different application architectures and implementations on potential platforms. The most popular technique employed nowadays is termed as system-level-performance evaluation which uses abstract workload and platform capacity models. The platform capacity models and application workload models reside at a higher abstraction-level. The...

Parallel Computing - Fundamentals and Applications - Proceedings of the International Conference ParCo99, 2000

This paper describes the collection and analysis of usage data for a large (hundreds of nodes) distributed memory machine over a period of 31 months during which 178,000 batch jobs were submitted. A number of data items were collected for each job, including the queue wait times, elapsed (wall clock) execution times, the number of nodes used, as well as the actual job CPU, system and wait times in node hours. This data set represents perhaps the most comprehensive such information on the use of a 100 Gflop parallel machine by a large (over 1,200 users in any given month) and diverse set of users. The results of this analysis provide some insights on how much machines are used and on workload profiles in terms of the number of nodes used, average queue wait times, elapsed and CPU times, as well as their distributions. A longitudinal analysis shows how these have changed over time and how scheduling policies affect user behavior. Some of these results confirm earlier studies, while others reveal new information. That knowledge has been used to develop a new scheduler for the system which has increased system node utilization from the 60% to the 90-95% range if there are sufficient jobs waiting in the queue.

In computing, the workload is the amount of processing that the computer has been given to do at a given time. Workloads are two types namely synthetic and real workload. Real workloads are not publicly available and some workloads are available in the internet like Google trace, world cup 98 trace and Clark Net trace. Synthetic workloads are generated based on our experiments. The real trace is downloaded from Google cluster data which consists of two workloads. In first trace it refers to 7 hours period with set of tasks. In second trace it refers to 30 days period of work. Each dataset is packed as set of one or more files, each provided in compressed Common Separated Values(CSV) format. In this paper we are analyzing the Google cluster data version 2 trace in IBM SPSS statistics and generating another workload called synthetic workload with the same characteristics and behavior of real workload based on formulae which is generated using linear regression in IBM SPSS statistics.

The Mermaid simulation environment facilitates the performance evaluation of a wide range of design options in MIMD multicomputer architectures. Because the Mermaid architectural simulators are driven by events that are more abstract than real machine instructions, workloads must explicitly be modelled. In this paper, we provide an overview of the workload modelling techniques used within Mermaid. This includes the tracing of real programs and the stochastic modelling of application behaviour. Moreover, we present several validation and performance results indicating that our simulation methodology obtains good accuracy while being fairly efficient.

Job Scheduling Strategies for Parallel Processing, 2009

In parallel systems, similar jobs tend to arrive within bursty periods. This fact leads to the existence of the locality phenomenon, a persistent similarity between nearby jobs, in real parallel computer workloads. This important phenomenon deserves to be taken into account and used as a characteristic of any workload model. Regrettably, this property has received little if any attention of researchers and synthetic workloads used for performance evaluation to date often do not have locality. With respect to this research trend, Feitelson has suggested a general repetition approach to model locality in synthetic workloads . Using this approach, Li et al. recently introduced a new method for modeling temporal locality in workload attributes such as run time and memory . However, with the assumption that each job in the synthetic workload requires a single processor, the parallelism has not been taken into account in their study. In this paper, we propose a new model for parallel computer workloads based on their result. In our research, we firstly improve their model to control locality of a run time process better and then model the parallelism. The key idea for modeling the parallelism is to control the cross-correlation between the run time and the number of processors. Experimental results show that not only the cross-correlation is controlled well by our model, but also the marginal distribution can be fitted nicely. Furthermore, the locality feature is also obtained in our model.

2010 IEEE International Conference on Computer Design, 2010

A new generation of high-performance engines now combine graphics-oriented parallel processors with a cache architecture. In order to meet this new trend, new highlyparallel workloads are being developed. However, it is often difficult to predict how a given application would perform on a given architecture. This paper provides a new model capturing the behavior of such parallel workloads on different multi-core architectures. Specifically, we provide a simple analytical model, which, for a given application, describes its performance and power as a function of the number of threads it runs in parallel, on a range of architectures. We use our model (backed by simulations) to study both synthetic workloads and real ones from the PARSEC suite. Our findings recognize distinctly different behavior patterns for different application families and architectures.