mesh-egress-ab

name	mesh-egress-ab
description	Measure what a PMOVES exit node buys a household vs its raw local uplink (A/B), and map participating homes onto measured node capacity. Use for the Fordham Hill community-mesh pilot, exit-node speed/selection checks, and capacity planning.

mesh-egress-ab

Repeatable, node-agnostic egress measurement + capacity planning. Born from the Fordham Hill pilot: prove — with numbers, not claims — what routing a household through a community-run KVM exit node costs and buys versus its own local uplink, and size how many resident homes the fleet can carry.

Script: deploy/provision/mesh-egress-ab.sh (portable — curl + awk only for measure/capacity; adds the tailscale CLI for auto ab). Runs on Linux, macOS, and Android/Termux.

When to use

• Pilot demos / field data — capture a clean direct-vs-mesh A/B at a resident's location (e.g. SLATE on the Starlink Mini).
• Exit-node selection — which of kvm2 / kvm4-1 / kvm4-2 is fastest right now (results vary run-to-run; measure, don't assume).
• Capacity planning — map a participant count onto a node's measured uplink before onboarding homes.
• After any exit-node or Mullvad change — confirm egress still flows and where it exits.

How to invoke

# Auto A/B: direct vs every approved exit node, self-restoring (CLI node)
make -C pmoves mesh-egress-ab
#   or: bash deploy/provision/mesh-egress-ab.sh ab

# Portable measure of the CURRENT egress (Android/SLATE: toggle exit in the
# Tailscale app between runs, then diff the two snapshots)
bash deploy/provision/mesh-egress-ab.sh measure --label starlink-direct   --save direct.json
bash deploy/provision/mesh-egress-ab.sh measure --label starlink-via-kvm  --save viakvm.json
bash deploy/provision/mesh-egress-ab.sh compare direct.json viakvm.json

# Capacity: how many homes fit on a node, and does a participant count fit?
make -C pmoves mesh-capacity DOWN=845 HOMES=200

Output

measure reports public IP + org, download/upload Mbps, latency, jitter, TTFB. ab prints the same for direct + each exit node, then restores. capacity prints an oversubscription table (1:1 / 5:1 / 10:1 / 20:1 / 50:1) with the effective per-home budget and a FITS/needs-more-nodes verdict, anchored to the repo's own 10 Mbps/home budget.

Coordination (tooling · skills · agents · teams)

• Operator / resident node (SLATE, 4090, any home): runs measure/compare — no privileges needed.
• vps-deployer agent: runs ab from a fleet node and provisions/rebalances exit nodes via Hostinger + Tailscale MCP.
• Fleet team: pairs with fleet:status, fleet:acl-audit, and deploy/provision/exit-node-healthcheck.sh (L4 Mullvad egress/leak check).
• Capacity output feeds pmoves/docs/pilots/fordham-hill/ and the PMOVES-Wealth community ledger + Tokenism contribution model.

Citations

• deploy/provision/mesh-egress-ab.sh — the tool
• pmoves/docs/operations/MESH_EGRESS_AB_RUNBOOK.md — operator guide
• pmoves/docs/operations/FLEET_CAPACITY_ANALYSIS.md §6 — per-home budget basis
• pmoves/docs/operations/TAILSCALE_EXIT_NODE_RUNBOOK.md — exit-node lifecycle
• pmoves/mk/egress.mk — mesh-egress-ab / mesh-egress-measure / mesh-capacity targets