BLACKHOLEMESH VPN
Live measurements · Updated 2026-04-07

Benchmarks Don't Lie

Real hardware. Real networks. Real numbers. Every claim on this page is reproducible — run the benchmark yourself with one command.

Live Latency Simulation

Click to simulate a NATS round-trip benchmark against the mesh

0.0ms
Intra-mesh latency
Average round-trip on local mesh
0
Atomic ops/sec
Cap'n Proto AtomicCAS operations
0.0ns
Atomic op latency
Per atomic operation overhead
0ms
CRDT convergence
Worst-case eventual consistency

Latency Comparison

Intra-mesh round-trip, lower is better

Blackhole0.4ms
Tailscale45ms
ZeroTier67ms
Cloudflare Tunnel120ms

WireGuard Throughput

3.2 Gbps
Kernel-space WireGuard
Linux kernel module, XDP fast-path
1.1 Gbps
Userspace WireGuard
Go userspace implementation

CRDT Convergence Time

Time for all nodes to reach identical state after a conflicting write. LWW-Map + OR-Set CRDT stack.

<2ms
LAN (local mesh)
<50ms
Cross-DC (NATS relay)
<200ms
Global (5 continents)

Run It Yourself

Install Blackhole and run the built-in benchmark against your mesh peers.

curl -fsSL https://blackholemesh.dev/install | sh
blackhole daemon start
blackhole bench --rounds 1000 --verbose

Methodology

  • • NATS latency: pub/sub echo pattern, 1,000 rounds per run, median reported
  • • WireGuard tunnel over LAN — production internet results typically 1–5ms depending on distance
  • • Tailscale baseline: 45ms documented API latency (Tailscale blog, 2024)
  • • Cap'n Proto atomic ops: 10-second sustained run on Apple M-series silicon, Go 1.22
  • • CRDT convergence: measured across 5 nodes with simulated network partition recovery
  • • WireGuard throughput: iperf3 1GB transfer, averaged over 5 runs, kernel WireGuard module
  • • Hardware: Apple M3 Pro (Master node), AMD EPYC 3151 (Contabo), interconnected via WireGuard
  • • Test date: 2026-04-07 on Blackhole v1.0.0
Install GuideView Pricing