Parallel Works ACTIVATE as a DOE Integrated Research Infrastructure (IRI) facility, and ACTIVATE as a consumer of IRI facilities. This repository holds the design, the reference implementation, and the deployment kits that connect ACTIVATE (commercial platform and the High Security Platform) with the American Science Cloud, the IRI Facility API v2, and the Genesis Mission platform.
| Directory | What it is |
|---|---|
docs/INTEGRATION_PLAN.md |
The integration plan: surfaces, resource and identity model, facility readiness, secure-enclave path, verification status, phased schedule |
docs/ARCHITECTURE.md |
Architecture with diagrams: the two modes and the gateway, request path, identity, job and filesystem sequences, routing, edge kit, resource model, status ledger, topology dashboard |
docs/DEPLOYMENT.md |
Choosing a shape, federation and edge deployment, gateway configuration, platform workflows, prototype runbook |
activate-iri/ |
IRI Facility API v2 adapters over ACTIVATE. Federation mode publishes everything the platform operates; edge mode runs on the login node of a single existing cluster. Tests, offline demo, edge install kit, federation manifests |
activate-iri-connector/ |
ACTIVATE workflow that submits PSI/J jobs to any IRI facility; cross-facility status board |
amscrot-activate/ |
AmSCROT service client so AmSC orchestration can drive ACTIVATE natively |
prototype/ |
The running prototype: facility file, scripts to run and publish the endpoint through pw endpoints, and a live smoke test |
iri-topology/ |
Facilities topology dashboard: aggregates IRI endpoints into the status-monitor topology format and serves the map |
flowchart TB
AMSC["AmSC orchestrator<br/>AmSC-IRO, AmSCROT, Airflow"]
USER["ACTIVATE user<br/>iri-job workflow, iri command"]
EP["activate-iri endpoint<br/>one IRI Facility API v2 for everything ACTIVATE reaches,<br/>with other facilities consolidated behind it"]
SYS["Systems ACTIVATE reaches, no IRI software on them<br/>lab clusters, existing and NeoCloud clusters,<br/>elastic cloud Slurm, partner sites via the edge kit"]
LABS["Facilities with their own IRI endpoints<br/>ALCF, NERSC, OLCF, ESnet"]
AMSC -- "mode 1: AmSC calls in<br/>Keycard or ACTIVATE credential" --> EP
USER -- "mode 2: any facility<br/>through one URL" --> EP
USER -. "or directly" .-> LABS
EP -- "jobs and file operations run through<br/>the ACTIVATE API: workflow runs, SSH, or local" --> SYS
EP -- "gateway: forwarded with<br/>the caller's facility token" --> LABS
Mode 1: AmSC reaches every system ACTIVATE operates through one endpoint, and the endpoint runs the work through the ACTIVATE API, so the systems need no IRI software. Mode 2: ACTIVATE users call any IRI facility from their account, through the same endpoint or directly. Gateway: facilities that run their own IRI endpoint are consolidated behind the PW endpoint, so a client needs one URL.
Run the reference implementation offline, no platform account needed:
cd activate-iri && make venv && make test && make demo
Every endpoint serves its own console at /console/: what it has stood up (facilities behind it, resources, open incidents) and a panel that drives the IRI API with the viewer's ACTIVATE credential, following a job from submission through the ACTIVATE workflow run to the cluster and back. The prototype's console is at https://activate-iri.activate.pw/console/; prototype/demo_a30.sh is the terminal version of the same demo against the lab A30 GPU server.
Current prototype status is recorded in prototype/NOTES.md.
License: MIT. Copyright (c) 2026 Parallel Works, Inc.