Co-planning pad for cowmesh.net <-> tomesh.net

This started off from an email conversation to share knowledge between the two mesh networks, with a additional objective to gather ideas about the appropriate hardware for cowmesh to purchase in the near future.

Reference image: [uploading is not working - so this doc]: j.mp/cowplansaux

Summary

cowmesh

  • TP-Link WDR 4300 routers (no longer able to purchase locally)
  • Raspberry Pi (runs Community Radio + other local services)
  • TP-Link signal booster
  • Ubiquiti rocket (can only flash older models)
  • solar charge setup
  • Commotion mesh tool

  • Our earlier setup had 6 node points covering upto 3km radius. TP-Link had a good coverage upto 200mts radius.
    With gradual usage some of the routers and PIs became non functional/bricked due to many technical challenges.
    Now we are left with 2 routers and also support to commotion stopped.

  • Internet access point is far away but not a priority for the network, local services is the goal
  • Target 10 km links as Community Radio programs are run from far apart villages

tomesh

Familiarity with:

  • Raspberry Pi, Espressobin, and other single-boards (see prototype)
  • HPM5G LibreRouter radio (on Espressobin)
  • Meraki MR16
  • LibreRouter
  • General openWRT concepts and setup and hardware
  • MikroTik devices running stock from meshnet @ DWeb Camp
  • cjdns, yggdrasil, babeld, batman-adv
  • ipfs, dat, ssb
  • no production mesh network in the city

Some recommendations received

Possible approaches

  1. Openwrt routers, use any proprietary point-to-point radio in bridge mode for mesh links (this is what ben did at dweb camp, an espressobin was used as the router but we should not use it for a production network due to SD corruption risk on power cycles)

  2. Plan for LibreRouter, because this is a path to:

    • Reliable supply chain of hackable hardware (commercial hardware seems generally less available in project region)
    • Big global community with shared goals behind project
    • Internationalization efforts could help with more community participation
    • Compatible with other hardware as well

    This means near-term choices should be compatible with LibreMesh on LibreRouter.

  3. Make Raspberry Pi routers too and use its Ethernet port to wire up any appropriate radio in bridge mode (or more than one radio through a simple vlan-tagging switch), experimental with SD corruption risk but potentially very self-configuring and integrated

Possible device configurations

  • Meraki MR16, market is flooded with used $6 devices, make good dual-band access points or openwrt mesh router in the future.

    • Retail for $400+
    • Used pricepoint very low due to cost prohibitive licensing model
    • Initial flashing requires the devices to to opened
      • Becomes openwrt device
      • Detailed instructions avaialble
      • Make it into a openwrt device
    • Todo:
      • it seems to run LibreMesh, Yurko can validate?
      • can it be purchased at $6 at cowmesh?
        • Avaialblity in India
        • Shipping to india
        • May require relay via North America

    Raspberry Pi + Meraki MR16 running LibreMesh (2.4G radio as access point, 5G radio as mesh link)

    • 5G radio would need to use a directional antenna via uFL connector for distance
    • range won't be great.
    • itial flashing requires knowledge
    • devices need a power source
      • 12V DC
        or
      • standard POE Injector
        (approach #2)

  • Raspberry Pi + vlan switch + MikroTik sxtsq_5_ac as in (approach #3)

Q & A

So If I understand correctly from what you mentioned is, you have installed Toronto Mesh on espressobin board and created a mesh setup.

It works but that is not a typical setup. Toronto Mesh isn't any particular setup, we have a bunch of tools, one that can be installed on espressobin is called prototype.

a. do you see anything we need to worry wrt to these n0-n1-n3- radios as a mesh layer

The range and hardware availability

Backhaul bandwith - make sure you have enough. When you "run out" n2 may need to be a seperat radio for each leg of the link.

b. do you see anything we need to watch out for when we assume we can hang the p1,p2,.. to the mesh

DHCP and moving nodes -> changing IPs, longer discussion, but I think Yggdrasil would be useful here for stable IP assignments

When dealing "short distance" mesh (especailly omni directional) you have to be carefull with saturating the wireless spectrum. When possible try to keep neighbouring or overlapping wireless signals on differnt frequencies.

c. do you know anyone who has been deploying services on a mesh, where an internet DNS cannot be assumed - and do you see how we can manage discovery of nodes and services

Toronto Mesh deals with this a lot, we have no DNS. Currently working on some ideas/theories on how to propegate announcements.

d. is it advisable to run DNS or distributed DNS - what are the best practices (although we do have a low-literacy context here)

Depends on mesh protocol (Layer 2 vs. Layer 3 mesh, can we automate DNS configuring). How many nodes do we expect?

e. do we know of off-the-shelf tools that will help us with these, including the captive portal needs.

Yes I believe so

Select a repo