TinyMoQ — Ultra-Low-Latency Live Streaming CDN

How it works

On-demand relays, just-in-time

No always-on edge fleet to pay for. A relay is allocated the instant a stream goes live and reclaimed automatically when it ends.

Single region — go live → spin up → connect

Publisher goes live → orchestrator allocates a micro-relay in ~35 ms → viewers connect over MoQ / QUIC.

Global, multi-region — on-demand edge pull

A distant viewer is routed to the nearest relay, which automatically pulls the origin stream — served on demand, no pre-warmed edge required.

Why TinyMoQ

A CDN that behaves like the workload

Traditional CDN edges are heavy and always-on. TinyMoQ is elastic & cost effective.

⚡

Instant, on-demand relays

Micro-relays boot and start serving in ~35 ms using ~19 MB each. We spin one up the moment a stream goes live and scale to zero when idle.

🚀

Sub-second live latency

MoQ over QUIC / WebTransport replaces the multi-second delay of HLS/DASH. New viewers receive the catalog and first media sub-second.

💸

Pay only when streaming

Capacity-based orchestration spins relays up just-in-time and reclaims them automatically. No idle fleet to pay for.

🌍

Global, multi-region

Independent relay clusters per region with on-demand edge pull. A viewer anywhere is routed to the nearest relay, which pulls the origin automatically.

🔐

Secure by default

Every stream is gated by short-lived signed tokens — for publishers and viewers alike.

📈

Massive density

A single node runs thousands of relays' worth of headroom — engineered to sustain ~7,500 concurrent viewers (~15 Gbps).

The difference

Always-on edge vs. micro-relay

Traditional CDN edge

FootprintHeavy, always-on

StartupPre-provisioned

Idle costPaid 24/7

Live latencyMulti-second (HLS/DASH)

TinyMoQ micro-relay

Footprint~19 MB, purpose-built

Startup~35 ms, on demand

Idle costScales to zero

Live latencySub-second (MoQ/QUIC)

Footprint ladder

A MoQ relay across deployment models

The same relay, four platforms. Disk comes from the actual binaries on the box; boot and memory are measured. Only the platform underneath changes.

Model	Disk image	Memory	Boot: OS	Boot: app	Cold → serving
Container Docker / Kubernetes	~60–120 MB	~30–50 MB	0	~ms	~tens of ms
microVM + full Linux fly.io / Firecracker	~100–150 MB	~50–100 MB	~100–125 ms	~ms	~125–175 ms
Unikraft binary-compat mode	~40–55 MB	~63 MB	~38–39 ms	~43 ms	~82 ms
Unikernel, native TinyMoQ	11.5 MB	~19 MB	~15 ms	~1 ms	~35 ms

See the full breakdown →

Live orchestration

Watch your fleet scale up and down in real time

Every relay is allocated, monitored, and reclaimed automatically — visible the moment it happens.

tinymoq.com/fleet

TinyMoQ Fleet

2 active · 3,142 conns · live

Active nodes

Draining

Warming

Connections

3,142 / 6,000

Egress

6.3 Gbps

Fleet load

52%

MAX_USERS2000 HIGH_WATER0.75 TARGET_FILL0.6 MIN_NODES0 MAX_NODES25

Live broadcasts

Broadcast	Viewers	Tracks	Age
live/keynote-2026	1,820	2	14m
live/match-feed	1,322	2	47m

Stream routing (broadcast → relay)

Broadcast	Relay	Conns	Egress
live/keynote-2026	relay-fra-2	1,820	3.6 Gbps
live/match-feed	relay-iad-1	1,322	2.7 Gbps

Relays

#	Endpoint	State	Load	Conns	CPU	RSS	Egress	FDS	Age
node1	relay-iad-1	active	0.66	1,322 / 2,000	11%	19.4 MB	2.7 Gbps	84	47m
node2	relay-fra-2	active	0.46	1,820 / 4,000	14%	21.1 MB	3.6 Gbps	96	14m

updated 1:49:33 PM · polling /status every 1.5s

Real-time fleet dashboard.

Why MoQ · why now

Built natively for the future of live media

Media over QUIC (MoQ) is the emerging standard for low-latency live streaming. Built on QUIC and WebTransport, it collapses the multi-second delays of segmented formats like HLS and DASH into a continuous, sub-second flow of media — with the congestion control, multiplexing, and connection migration that QUIC brings for free.

The catch: MoQ relays have historically meant always-on infrastructure. TinyMoQ is a production CDN engineered for the standard from the ground up — an elastic fleet of lightweight micro-relays that exist only while a stream needs them, then disappear.

The result is a CDN that delivers the latency of MoQ with the economics of serverless: spin up in milliseconds, serve at the edge, scale to zero.

Relays that spin up in milliseconds and scale to zero.