Source linked

MeshDNS Hits 0.47ms Resolution on ESP8266 with Byzantine Fault Isolation

A cooperative DNS framework for resource-constrained IoT networks achieves sub-millisecond warm-cache resolution and isolates Byzantine faults using Ed25519-signed quorum voting on $2 microcontrollers.

meshdnsesp8266iotdnsbyzantine fault toleranceed25519

MeshDNS drops warm-cache DNS resolution to 0.47 milliseconds on commodity ESP8266 hardware—roughly a third of native mDNS's 1.39ms—while layering Ed25519-signed quorum voting to isolate Byzantine faults. That's not a theoretical simulation on a workstation; that's real silicon with sub-50 KB usable RAM and an 80 MHz clock.

Sub-millisecond DNS on a $2 Microcontroller

The paper's authors built MeshDNS for exactly the kind of network where a single DNS authority is a single point of failure: resource-constrained IoT deployments with unreliable links and adversarial peers. Each node maintains cache awareness using hash-based summaries, so the whole mesh knows who has what without flooding. Warm-cache queries resolve in under half a millisecond because the node just grabs the record locally; no round trips. That 0.47 ms number comes from an actual implementation on ESP8266—chips you can buy for a couple of bucks.

Trading Latency for Byzantine Safety on Cold Misses

Cold-cache misses are where the clever engineering shows up. Instead of trusting a single responder, MeshDNS requires identical answers from multiple admitted peers, each signed with Ed25519 signatures. The penalty is predictable: about 1.3 to 1.7 seconds per cold miss. That's a deliberate tradeoff—spend a couple of seconds to guarantee Byzantine fault isolation among peers, assuming physical hardware extraction is out of scope. In an environment where a compromised node could poison a whole network's name resolution, that 1.7 seconds is cheap insurance.

Scaling from 5 to 1,000 Nodes Without Central Authority

The team validated MeshDNS on a 5-node physical testbed and then ran discrete-event simulations scaling to 1,000 nodes. The results show resilient local name caches even under churn—nodes joining, leaving, or going silent. Because the framework uses shared-key admission and hash-based cache synopses, it doesn't need a central DNS resolver or a cloud gateway. Every node is both client and resolver.

What this enables next: persistent edge telemetry in sensor swarms where you can't trust every node but you can't afford a full-time DNS server either. MeshDNS is available now on GitHub under the artifact link, so anyone with a few ESP8266s and a soldering iron can start testing Byzantine-tolerant name resolution today.


Source: MeshDNS: A Cooperative DNS Resolution Framework for Resource-Constrained IoT Networks
Domain: arxiv.org

Read original source ->

External source stays available while the OJO article and comment thread stay local.

Comments load interactively on the live page.