91

SHARE
Research Highlight

Error-controlled Progressive Retrieval of Scientific Data

Topic:
Progressive data retrieval workflow: Given a requested tolerance 𝜏0, the retriever fetches data to form a low-resolution/precision representation; later, as the requested tolerance changes to 𝜏0, the retriever fetches additional data and add it with the existing data to form a high-resolution/precision representation
Progressive data retrieval workflow: Given a requested tolerance 𝜏0, the retriever fetches data to form a low-resolution/precision representation; later, as the requested tolerance changes to 𝜏0, the retriever fetches additional data and add it with the existing data to form a high-resolution/precision representation

The Science

Researchers at ORNL and Missouri S&T significantly expanded the original MultiGrid Adaptive Reduction of Data (MGARD) technique with resolution- and precision-based incremental refactoring whereby the data can be prioritized and progressively retrieved with fine-grained error bounds.

The Impact

By refactoring data into a resolution and precision hierarchy and storing it across the deep storage tiers, we can prioritize the data retrieval and minimize the I/O burden.