DB
Back homeIndustry

Media & streaming

Playback QoE telemetry, live engagement counters, and session state on one engine — line-rate ingest with current-state reads.

The problem

Why this is hard today

A streaming platform watches millions of concurrent sessions: playback telemetry — rebuffering, bitrate, startup time — streams in continuously, engagement counters move every second, and the questions that matter are about the live audience right now.

The typical stack splits QoE telemetry into a stream processor, live counters into a cache, and session history into a warehouse — three systems that drift, and a viewer-experience dashboard that lags the incident it is meant to catch.

The fit is one engine that ingests QoE events at line rate, maintains per-session and per-title current state as O(1) reads, and rolls up live engagement continuously in the same SQL.

Where NYXDB fits

Use-case journeys

QoE observability

Continuous transforms turn raw playback telemetry into live per-title QoE tables.

Observability & logs

Live engagement dashboards

Dashboards read pre-aggregated keyed tables and subscribe to live counters.

Real-time dashboards

Playback time series

High-rate ingest and windowed downsampling of QoE metrics for trend analysis.

Time series

Architecture

How NYXDB fits streaming telemetry

Playback events append durably at line rate under the memory governor; keyed tables hold the latest state per session and title; continuous transforms maintain the live engagement tables dashboards read as ordinary tables.

  1. 01

    Line-rate ingest

    Shard-owned, lock-free write paths, WAL-backed, governed under pressure.

  2. 02

    Session state

    Keyed tables keep the latest per (session, title) — O(1) reads.

  3. 03

    Engagement rollups

    tumble/hop windows maintained continuously by transforms (ADR-080).

  4. 04

    Live streams

    STREAM SELECT pushes QoE and engagement deltas to dashboards as they commit.

Real SQL

Representative queries

Sessions currently rebuffering
SELECT session_id, title_id, rebuffer_ratio
FROM session_state
WHERE rebuffer_ratio > 0.02;

Every statement follows the engine’s own test SQL shapes. See the SQL reference for full syntax.

Capabilities

What you get

Line-rate ingest

Lock-free shard-owned write paths.

O(1) session state

Latest per session and title, instantly.

Windowed engagement

tumble/hop maintained continuously.

Live push

STREAM SELECT to QoE dashboards.

Proof

Measured on the vetted benchmark lane

~2.56µs

event → engagement-table propagation

flat, empty to a 2M-row tail

vetted table

View benchmark268ns

per-session state count, flat 1k–16k keys

O(1) maintained counter

vetted table

View benchmark1.53M rows/s

durable QoE ingest, single core

1M-row batches, best durable mode

vetted table

View benchmark

Measured on AMD EPYC 9554P 64-Core Processor (64t), 1505GB RAM, 6x NVMe — measured on server-class hardware. Release build, median of 5 (re-captured stragglers: median of 3), commit-pinned (7945ae33, 2026-07-09). Ingest figures are engine-side. See the full benchmark suite.

Learn more

Related documentation

Keyed tables

Current state per session.

Transforms

Continuous transforms and STREAM SELECT.

Memory governor

Backpressure under load.

Telemetry your streaming ops can query

Explore observability or run a node.

Get started Read the docs