SimTracker - Using the Web to track computer simulation results

Abstract

Large-scale computer simulations, a hallmark of computing at Lawrence Livermore National Laboratory (LLNL), often take days to run and can produce massive amounts of output. The typical environment of many LLNL scientists includes multiple hardware platforms, a large collection of eclectic software applications, data stored on many devices in many formats, and little standard metadata, which is accessible documentation about the data. The exploration of simulation results typically proceeds as a laborious process requiring knowledge of this complex environment and many application programs. We have addressed this problem by developing a web-based approach for exploring simulation results via the automatic generation of metadata summaries which provide convenient access to the data sets and associated analysis tools. In this paper we will describe the SimTracker tool for automatically generating metadata that serves as a quick overview and index to the archived results of simulations. The SimTracker application consists of two parts - a generation component and a viewing component. The generation component captures and generates calculation metadata from a simulation. These metadata include graphical snapshots from various stages of the run, pointers to the input and output files from the simulation, and assorted annotations describing the run. SimTracker generation can be done either during a simulation or afterwards. When integrated with a code system, SimTracker does its work on the fly, allowing the user to monitor a calculation while it is running. The viewing component of SimTracker provides a web-based mechanism for both quick perusing and careful analysis of simulation results. HTML is created on the fly from a series of Perl CGI scripts and metadata extracted from a database. A variety of views are provided, ranging from a high-level table of contents showing all of one's simulations, to an in-depth results page from which numeric values can be extracted and analysis tools can easily be launched. Annotations can be associated with a calculation at any time, allowing an end-user to customize the summary pages with titles, abstracts, and pointers to related information, for example. In this paper, we will present an overview of the design, implementation, and operational aspects of the SimTracker application. We will also discuss how it is being deployed in the environment of the Accelerated Strategic Computing Initiative [1]. SimTracker was designed as an extensible application that we are now adapting to use with several simulation codes.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 8 citations exist upstream.