tailscale / tailscale/tailscale

FR: Feature Request: Thin Edge Client for Embedded/IoT Devices (<256MB RAM)

Open
#18,013 3 comments 4 reactions 0 assignees View on GitHub
fr needs-triage
Dominant language
Go
Stars
36.5k
Forks
3.2k
Avg merge
2d 3h
Merged PRs (30d)
123

Description

### What are you trying to do?

Run Tailscale on embedded devices with <256MB RAM (routers, IoT gateways, industrial controllers, SBCs) where the current client's 72MB RSS footprint is too heavy.

**Current measurements (GL.iNet MT300N-V2, 128MB RAM, OpenWrt 24.10):**
```
VmRSS: 72,300 kB (56% of available RAM)
VmSize: 560,136 kB
Binary: ~25MB on disk (8MB compressed ipk)
```

On a 128MB device, Tailscale alone consumes **56% of RAM**, limiting what else can run. This is similar to #7272 where users hit OOM kills, but the root cause is baseline memory footprint.

**My use case:** Out-of-band management for remote lab infrastructure. The router provides rescue access to other devices when primary connectivity fails.

### How should we solve this?

Create a **"thin edge client"** variant optimized for resource-constrained gateways:

**Target:** <20MB RSS, <5MB binary

**Approach:**
- Leverage **kernel WireGuard** when available (vs userspace wireguard-go)
- Strip coordination-heavy features not needed for edge nodes
- Focus on: tunnel establishment, subnet routing, basic mesh participation
- Minimal dependencies

**Potential architecture:** Separate binary that handles data plane (tunneling) while delegating control plane to a more capable node or using simplified coordination protocol.

### What is the impact of not solving this?

**Blocked markets:**
- Millions of home routers (GL.iNet, Mikrotik, Ubiquiti EdgeRouter, TP-Link)
- IoT gateways with <256MB RAM
- Industrial edge controllers
- Legacy set-top boxes / Smart TVs
- Cost-sensitive deployments

**Current workarounds:**
1. Use Nebula instead (~5-10MB RSS) - loses Tailscale's ease-of-use
2. Use Headscale - same client footprint problem
3. Plain WireGuard - loses mesh coordination benefits

**Strategic impact:** These devices are ideal Tailscale nodes (always-on, gateway position, stable infrastructure) that drive network effects. Avery mentioned exploring machine-to-machine/IoT pricing in recent interviews - a thin client would enable this market.

### Anything else?

I've been a software engineer for 20+ years (founded an ISP, built traffic accounting systems, led teams at DAZN/Concur optimizing for resource-constrained platforms). Happy to contribute to this effort if it aligns with Tailscale's roadmap.

**References:**
- #7272 - OOM kills on 128MB OpenWRT devices (ongoing issue)
- [Recent Stratechery interview](https://stratechery.com/2025/an-interview-with-tailscale-co-founder-and-ceo-avery-pennarun/) where Avery discussed IoT/machine-to-machine market opportunity, which would benefit from smaller footprint edge clients

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.