close up photography of black pen

Rural Mesh WiFi: Part 3 – Network Design

I recently had the pleasure of designing and installing a network for a family member who lives in a rural part of the country. This experience has inspired me to share the details so that other may benefit from my research and learnings.

If you missed the other posts regarding this project, feel free to check them out.

Project Planning

Here is the 100-foot view.

A wooded area with a farm house in the bottom, right corner and a shop in the top, left corner.

Fortunately, the property owner had already trenched a direct-burial ethernet cable between the house and the shop. All I had to do is plan the network equipment and convince the owner to spend the money on it.

Existing Equipment

It is always a good idea to consider using materials and equipment that you already have. For this project, there was a Google WiFi router in the house with two additional mesh access points. The underground ethernet cable fed into a commodity wireless router in the shop. Everyone had wireless internet. It worked.

The Problems

Because each router has a network address translation (NAT) feature, no devices in the shop could talk to any devices in the house. The only device in the shop that devices in the house could talk to was the shop router. This was not necessarily a problem in the beginning because the only goal was internet access. However, the owner wants to install surveillance cameras on the house and on the shop. This will require communication within the local area network.

The shop was what we call “double-NATed”. This is essentially a local network inside of a local network, and it is created by having the second router in the shop. For clarification, here’s what should have been done in the first place in the shop. The shop never needed a secondary router. The goal was to provide ethernet for a desktop PC and wireless access for phones. It was cheap and easy to use a router that was laying around to do this since most wireless routers have several LAN ports. Let’s look at the actual functionality that we need from the router: a) multiple LAN ports, and b) wireless access. Multiple LAN ports can be achieved with a network switch. Wireless access can be achieved by a wireless access point.

Previously, the internet uplink was in the house and fed to the shop. As you can see in the photo, the shop has a much wider field of view to the sky, so the Starlink dish was much better suited to be installed at the shop. While that configuration change is not a huge deal, it did mean that all the data would be going through the commodity router, and that most likely was a data bottleneck.

The third issue is that I have to configure, manage, and maintain this network. I am well-capable of working with different brands of consumer-grade routers, but the quality of the network and the amount of time we all spend working on it is greatly improved by using a solid, proven system.

The Solutions

Moving the internet uplink to the shop was going to be no big deal. Just install the internet source at the shop.

Once the internet source is in place, we need a router to assign IP addresses, perform network address translation, and route traffic to the appropriate devices. It is commonplace for consumer-grade routers to include a wireless access point that is integrated into the same case as the router. However, wireless network access is not a routing function. For design purposes, we consider wireless network access to be a different concern than network traffic routing.

In order to provide wireless access to the network, we needed to add a wireless access point.

We could power the wireless access point by wall-wart, but we know we want to add surveillance cameras in the future, so we opted to power the wireless access point with a power-over-ethernet (PoE) switch. The PoE switch supplies power to the access point using the same ethernet cable that provides data access. The same PoE switch can be used for surveillance cameras in the future.

The next step is to get the house connected to the network. We’ll use the existing direct-burial ethernet cable that runs between the shop and the house to accomplish this. On the shop side, the ethernet cable will connect to the PoE switch.

In the house, we will disconnect the existing Google WiFi mesh system since putting the device in bridge mode also disables its mesh capabilities. In its place, we will install a PoE switch just like we installed in the shop. The PoE switch gives us ethernet ports, as well as power. Just like in the shop, we plan to install cameras on the house that can be powered by the PoE switch.

Similarly to the shop, we need wireless network access in the house. We will achieve this by installing two wireless access points. One access point will go in the basement, and one will go on the main floor. Depending on the number of floors and layout of walls in a house, you may need fewer or more access points for adequate coverage.

The other items that will connect to the PoE switch include a small media server and an ethernet jack that had already been installed in one of the rooms in the house.

Note that most PoE switches are auto-negotiating, which means they will supply power to devices that need it, but they will not supply power to devices that don’t need it. Additionally, many PoE switches have a group of PoE ports and a group of ports that do not supply power. Be sure to use the appropriate port for your end device.

The only thing missing from the entire setup at this point is an extremely easy way for me to access the network and manage it centrally. For this, we’ll add a hardware controller with cloud access that I can log into from anywhere.

Assumptions

For everything we do, I assume gigabit speeds (1000 Mbps). Several of the legacy switches we replace were what they call “fast ethernet” or 10/100 Mbps devices. 100 Mbps is sufficient in many environments, but we didn’t want that limitation for our network.

All the devices have software-defined network (SDN) capabilities. This is what allows the centralized cloud management of the network. Not only do they need to be SDN-capable, but they must use the same SDN standards, which means they must all be the same brand.

When designing a network, you should consider the number of clients that will connect to your network. For example, we choose a 16-port switch for both the shop and the house. The switch gives us 8 standard gigabit ports and 8 PoE gigabit ports. That leaves plenty of room for our current and foreseeable future needs.

Conclusion

So far, we’ve dreamed up a pretty good network design based on the current and foreseeable requirements of the property owner and the network maintainer.

In our next post, we’ll take a look at the actual implementation details and the specific network equipment we chose, as well as the features of SDN.

Further Reading

For a very in-depth lesson in network design, check out Network Design Cookbook by Michel Thomatis.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.