IP‑ST // Live Benchmark
Eurovision‑Style Live Benchmark — RB 1.4
Results without run manifests + raw logs are non‑comparable.
Dare to test it? This might surprise you.
If you build live streaming systems, you already know the hardest content isn't talking heads — it's high‑entropy live production: strobes, smoke, fast pans, sudden lighting changes, confetti‑like detail. This page publishes RB 1.4, a reproducible benchmark protocol that pits the industry baseline for low-latency streaming — Low‑Latency DASH (LL‑DASH) with chunked CMAF, played via dash.js (DASH‑IF reference player) — against IP‑ST, a frameless, asynchronous architecture that converges a scene from independent state datapoints.
This is not a demo. It's a protocol designed to produce hard numbers: startup, interruptions, traffic smoothness, and stability under high‑entropy scenes.
Ready to challenge your assumptions?
No hype. Run it. Log it. Compare it.
If you're confident in the baseline — you should be excited to test this.
Evidence beats opinions. RB 1.4 is designed for that.
Why "Eurovision‑style" content breaks streaming first
Modern ABR streaming works by encoding multiple representations and delivering content in segments/chunks described by a manifest (MPD). The client continuously adapts what it downloads based on network conditions and buffer health.
Low‑latency DASH using chunked CMAF reduces end‑to‑end delay by exposing sub‑segment chunks (so the client can access media data before a full segment completes). This improves latency, but it also tightens the system's margin for error when network conditions fluctuate.
In other words: when the show becomes chaos, segmented delivery and adaptation decisions get stress‑tested first.
Baseline vs IP‑ST — two fundamentally different paradigms
Control / Baseline
LL‑DASH + chunked CMAF + dash.js
- Protocol: Low‑Latency DASH (LL‑DASH)
- Packaging: chunked CMAF
- Player: dash.js (DASH‑IF reference implementation)
- Delivery: MPD manifest + segmented/chunked ABR switching
Intentionally hard to dispute: best-practice approach to low-latency HTTP streaming with a widely recognised reference player.
Treatment
IP‑ST — frameless asynchronous state streaming
- No frames. No segments. Continuous state convergence.
- Client maintains a Dynamic State Buffer — scene converged from independent datapoints (ID+S).
- Transport: designed for non-blocking delivery (QUIC: UDP-based, no head-of-line blocking between streams).
What we measure (not what we "promise")
RB 1.4 is built around measurable, comparable outcomes — even though the paradigms differ.
Startup (first useful output)
- Baseline: TTFF — time to first frame
- IP‑ST: TTFS — time to first state
Interruptions (service instability)
- Baseline: stall count + stall duration (rebuffer/spinner events)
- IP‑ST: state starvation time (dropped points/ms and starvation windows)
Traffic smoothness (network reality)
- Throughput variance (1s sampling)
- Peak‑to‑average throughput ratio
Baseline adaptation stress (segmented ABR only)
- ABR switch frequency and oscillation patterns — because ABR clients switch representations based on measured conditions.
Physical marker proof (cross‑paradigm validation)
- Optional visual event marker (e.g., timecode/strobe at t = 20s) to enable motion‑to‑photon verification and visible artifact inspection across both systems.
⚠ Important: Comparable results require evidence
Results without run manifests and raw logs are non‑comparable. RB 1.4 is designed to be reproducible. If you want your results to be considered reliable and comparable, include the full evidence set so conformance to the published protocol can be checked.
Required evidence artifacts:
- Run manifests (*_manifest.json) for every run
- Raw logs (baseline and IP‑ST JSONL)
- Player/client versions and hashes
- Network profile parameters used
- Output summaries (summary.csv, summary_by_profile.csv)
Official validation requires a conformance review against the original RB 1.4 plan and parameters.
RB 1.4 — Public Runbook Summary (5 steps)
RB 1.4 is designed to be reproducible in your lab or cloud environment.
Prepare two clients
Baseline: dash.js LL‑DASH player. IP‑ST: DSR client (frameless state renderer).
Use two rights‑cleared clips
High‑entropy live production style. Easy/stable scene.
Apply controlled network profiles
Good broadband. Typical mobile. Congested mobile (+ jitter).
Run the A/B matrix
Minimum recommendation: 120 runs (2 clips × 3 profiles × 2 systems × 10 repetitions).
Generate evidence artifacts
summary.csv, raw JSONL logs, run manifests, and optionally a short screen capture around the marker event.
If you believe segmented ABR is already optimal, RB 1.4 is your chance to prove it.
Results without run manifests and raw logs are non‑comparable. Official RB 1.4 validation requires a conformance review against the published protocol. Submit the full evidence set with your results.
Benchmark Kit — What's Inside
The kit is public. Register below to receive the download link and your run‑ID prefix. No manual approval required for the kit — just leave your details so results remain traceable.
RB 1.4 Full Runbook (PDF)
Frozen baseline definition, network profiles, run matrix, and reporting requirements.
Baseline LL‑DASH Harness (dash.js + collector)
dash.js player page with log collector hooks for the LL‑DASH baseline.
IP‑ST Log Normalizer (DSR JSONL → canonical schema)
Converts DSR JSONL output into the canonical schema for cross-paradigm comparison.
Analysis Template (Python + CSV + charts)
Python scripts, summary.csv schema, and chart definitions for result analysis.
Sample Logs (synthetic)
Validate your analysis pipeline without PoC access — reference JSONL logs.
Evidence Pack Template
DD‑friendly folder structure for sharing results with partners and investors.
PoC Endpoint Access — Separate Process
PoC MPD URLs, IP‑ST endpoints, access tokens and infrastructure details are not part of the public kit. They are delivered only via manual email approval with time-limited credentials.
To request PoC access: email ari.jarvinen@a-nodes.com with your run‑ID prefix after downloading the kit.
Download Benchmark Kit
Leave your details — you'll receive the download link and a run‑ID prefix immediately after submission. Results stay traceable this way.
⚠ Ground rule
Do not publish PoC endpoints, tokens, or access links. Publish your results — not access details.
IP Notice — Patent Application Filed
A patent application covering key aspects of IP‑ST is filed and pending. This benchmark protocol is published to enable external validation and to make the comparison methodology explicit.
Ready to run it?
Dare to test it? If you run RB 1.4, you'll have something rare in this space: a baseline‑respectful, evidence‑first comparison between segmented low‑latency ABR streaming and a frameless asynchronous state architecture.
Need to test against a live PoC endpoint? Request PoC Endpoint Access (separate manual approval — download the kit first).