Kesheng Wu

Query-Driven Visualization and Analysis

Author(s): Ruebel, Oliver | Abstract: This report focuses on an approach to high performance visualization and analysis, termed query-driven visualization and analysis (QDV). QDV aims to reduce the amount of data that needs to be processed by the visualization, analysis, and rendering pipelines. The goal of the data reduction process is to separate out data that is scientifically interesting and to focus visualization, analysis, and rendering on that interesting subset. The premise is that for any given visualization or analysis task, the data subset of interest is much smaller than the larger, complete data set. This strategy---extracting smaller data subsets of interest and focusing of the visualization processing on these subsets---is complementary to the approach of increasing the capacity of the visualization, analysis, and rendering pipelines through parallelism. This report discusses the fundamental concepts in QDV, their relationship to different stages in the visualization and analysis pipelines, and presents QDVs application to problems in diverse areas, ranging from forensic cybersecurity to high energy physics.

Towards Autonomic Science Infrastructure: Architecture, Limitations, and Open Issues

Scientific computing systems are becoming increasingly complex and indeed are close to reaching a critical limit in manageability when using current human-in-the-loop techniques. In order to address this problem, autonomic, goal-driven management actions based on machine learning must be applied end to end across the scientific computing landscape. Even though researchers proposed architectures and design choices for autonomic computing systems more than a decade ago, practical realization of such systems has been limited, especially in scientific computing environments. Growing interest and recent developments in machine learning have spurred proposals to apply machine learning for goal-based optimization of computing systems in an autonomous fashion. We review recent work that uses machine learning algorithms to improve computer system performance, identify gaps and open issues. We propose a hierarchical architecture that builds on the earlier proposals for autonomic computing systems to realize an autonomous science infrastructure.

Identifying Anomalous File Transfer Events in LCLS Workflow

This short paper reports our on-going work to study and identify anomalous file transfers for a large scientific facility known as Linac Coherent Light Source (LCLS). We identify the anomalies based on the statistical models extracted from the recent observations of the file transfer events. This data-driven approach could be used in different use cases to identify unusual events. More specifically, we propose two different identification strategies based on the different properties of the observed file transfers. Because these methods capture key aspects of the two different segments of the data transfer pipeline, they are able to make accurate identifications for their respective workflow components. The current anomaly detection algorithms only make use of the file sizes as the primary feature. We anticipate that integrating more information will improve the prediction accuracy. Additional work is planned to validate the identification algorithms on more data and in different use cases.