Grand Unified Testbed — one origin, four paths, one player

Grand Unified Testbed — one origin, four paths, one player An architecture diagram generated by Archify. MSP Media Origin · CMAF packaging · manifests · Architecture component · WP 2.1 · HNR MSP Media Origin CMAF packaging · manifests WP 2.1 · HNR Public CDN edge · CloudFront · GCore gCDN · Architecture component · Baseline Public CDN edge CloudFront · GCore gCDN Baseline DTH broadcast · scheduled via FTA API · TBC · Architecture component DTH broadcast scheduled via FTA API · TBC DTH far edge cache · broadcast receive · Architecture component · EBU-DTH DTH far edge cache broadcast receive EBU-DTH VSAT backhaul · QoS + slice via CAMARA · Architecture component VSAT backhaul QoS + slice via CAMARA Far edge in 5G core · private network · Architecture component · ViaSat DTV Far edge in 5G core private network ViaSat DTV Satellite pre-population · scheduled via Fenix API v3.3.0 · Architecture component Satellite pre-population scheduled via Fenix API v3.3.0 Far edge smart cache · Varnish + nginx on disk · Architecture component · Arctic Space DTE Far edge smart cache Varnish + nginx on disk Arctic Space DTE GCore mini CDN · terrestrial fallback · Architecture component GCore mini CDN terrestrial fallback RAI origin server · partner-operated · Architecture component RAI origin server partner-operated Reference Player · Test Harness · emits CMCD / CMSD · Architecture component · HNR Reference Player Test Harness · emits CMCD / CMSD HNR unicast, no orchestration segments for carousel segments on demand segments ahead of playout on cache miss origin fetch

Read it this way

  • • One MSP Media Origin on the left, one Reference Player on the right, four delivery paths between them.
  • • Rows are the four testbeds, ordered by how much machinery each adds to the Baseline.
  • • Only the middle column changes per testbed. The origin and the player are the same software everywhere.

Baseline has no middle column

  • • Baseline reaches the player with no transmission scheduling at all — that is what makes it the control case.
  • • The gap is deliberate, not an unfilled cell.
  • • Baseline is the control case every other row is measured against.

What this licenses

  • • Functional results transfer across rows, because the shared components are the same software.
  • • Performance results do not: each row sits on different infrastructure with its own limiting factors.
  • • What decides which content moves is the Supply & Demand Control diagram, not this one.