Skip to content

UC2 — Geospatial + EHT

Lead: C.K. Chan  ·  WBS: 5.0  ·  DOI deliverable: Month 17

Story

The Event Horizon Telescope (EHT) collaboration produces petabytes of correlator output per observing campaign. Production-scale processing (100 PB pipeline) is deferred to the MESA follow-on operations proposal. The prototype demonstrates the same architecture at sub-PB scale.

In parallel, geospatial analytics workloads — fire-perimeter mapping, crop classification, drought monitoring — share enough of the storage and compute pattern with EHT that they can be exercised on the same prototype.

What the prototype demonstrates

  1. Lakehouse scaling on multi-TB Parquet datasets with sub-second queries.
  2. Federated storage across CyVerse, TACC Corral, and OSN — the same federated read path that the production EHT pipeline would use at 100 PB.
  3. Agentic orchestration drives a multi-stage pipeline: ingest → calibrate → image → catalog → publish.

Production deferral

The full 100 PB EHT pipeline is deferred to the follow-on Category I/II operations proposal. The sub-PB prototype validates the architecture under the prototype scope.

Deliverables (Month 17)

  • Reproducible workflow with DOI.
  • Empirical performance numbers feeding the hardware-spec dataset (WBS 8.0).

Status

Draft — content matures through Phase 2.