24 September 2015- BREAKOUT 6 - 15:30
Large datasets are increasingly being generated from scientific instruments, numerical analysis or large-scale simulations across many fields including astronomy, geosciences and neuroscience. These datasets are on the order of terabytes and growing to petabytes. Transporting such datasets is very costly both in terms of network bandwidth and time. Increasingly, the need has arisen for active data repositories that combine large data storage with computational resources for active querying, visualization and analysis of the data. Typical use cases involve requesting a specific representation (projected slice, rendering or image format) of a subset or the entire volumetric dataset. This can be used to provide interactive web viewing of an overview of the dataset or through multiscale and multiresolution navigation viewing and retrieving a full resolution subset of the data. The dataset may be accessed and viewed in parallel by many users or processes simultaneously. More advanced requests may involve sophisticated subsetting, extracting subvolumes along a path, for example. Remote feature extraction and analysis is an additional key use case. The ability to deploy a feature extraction service close to the data, allows the extraction of lighter weight properties, using machine vision, pattern recognition or other analytics, from the full volume dataset. These extracted features are significantly smaller than the originating dataset and can be returned as independent data objects. The goal of this bird of a feather would be to identify common needs and shared requirements for a volumetric imaging service that could apply across scientific domains. In the event of significant convergence, efforts can be focused on a common implementation and data formats.
Contact Person: Sean Hill