In a digital world, everything we do is connected by networks. Yet, while networks are more important than ever, designing, configuring, and managing a network has never been more complex.
Something needs to change.
The Big Network mission is to make connectivity and networking simple to set up and maintain - by anyone. We are changing the way networks are built and managed.
In 2012, I built a wireless gigabit Internet Service Provider (ISP) called Really Fast in Seattle, Washington. We provided customers with fast Internet services with simple pricing and great customer support. We ran a reliable network for our customers by using fixed wireless, fiber and diverse pathed internet in combination with network tunneling to make the network resilient to maintenance events and outages.
At Really Fast, our target customers were tenants in multi-dwelling and multi-tenant buildings. To deliver services, we used multiple wireless backhaul radios and fiber optics lines that fed back to our data centers. When we could, we created diverse paths from our datacenters to the buildings we serviced. When we could not, we had to improvise, and that’s where we used network tunneling as part of our strategy.
In locations where we could not build our own diverse paths, we built devices that could run static overlay IP tunnels over the top of a 3rd party bandwidth connection. These tunnels would point back to a concentrator in our datacenter which would hand off to our core. We would run these tunnels over Cable, DSL, or whatever we could get. We built tunnels with GRE, OpenVPN, IPSEC, SoftEther and other tunneling techniques. We were able to carry our network, OSPF, and iBGP routes over the tunnels. This worked great for handling maintenance events and outages on our primary paths.
We used tunneling technology to build redundancy in places that would otherwise be single homed to the Internet. This solution kept us out of trouble during any network disruptions and kept our customers incredibly happy with our services. We sold the company in 2015 and it became part of Wave Broadband’s Gigabit offering.
In 2018 I co-founded and helped launch another Gigabit ISP by the name of Honest Networks in Manhattan, New York. Same premise: simple, reliable, and fast Internet service with great customer service. It is now one of the highest rated residential ISPs in New York.
At Honest, we would deliver Internet over wireless backhaul technology along with fiber. This wireless technology can do 10Gbps at low latency, however, it is has the risk of being affected by rain fade. When heavy storms would come into New York, we needed a fast way to recover from dropped packets that would be un-noticeable by customers. We again implemented tunneling as a failover method for radios until dual path physical infrastructure could be installed. The tunneling strategy again worked great. As with ReallyFast, we used tunneling technology to keep our network online and our customers happy.
Yet there were always problems with tunnels: the speed via tunnel was never good enough to be a primary link, there were various imperfections in using tunnels, especially behind Network Address Translation (NAT). We could not support advanced services or common network protocols through several types of tunnels. As both companies grew we started to get requests for different services: Customers wanted access to the cloud; they wanted to link branch locations together; They wanted to cross connect to their datacenter. The tunnels were good enough for general Internet traffic, but could never meet the reliability requirements or flexibility required for advanced services.
It is worth touching on the topic of tunnels and NATs for a moment: Tunnels need at least one site of the tunnel endpoint to be generally discoverable on the Internet. You generally need some globally available IP address for tunnel clients to connect to, creating a discoverable and therefore DoS’able (Denial of Service) point on your network. If you’re behind NAT, you need to forward ports. If you’re behind a Dynamic IP address, you need something like Dynamic DNS to keep the endpoint known to clients. This is all overhead, complexity, and operational burden for the benefit of establishing backup paths for a network to use!
In addition to various problems with tunnels themselves, we spent many hours laying out tunnels, configuring tunnels, and monitoring their status. Fully meshed tunnel design doesn’t scale linearly with endpoints / locations, so the overhead expands rapidly. Here’s some math based upon number of locations:
Setting up tunnels manually for every connection was painful.
In between selling my ISP in 2015 and starting another in 2018 I moved from Seattle, my home of 34 years, to NYC with the intention to start a new ISP and having a new life adventure. I spent time exploring technologies, building ideas and consulting. I also spent alot of time studying how networking was changing as we knew it. Intent based networks, mesh networks, software defined networks, sd-wan and sd-lan, edge and zero-trust to name a few. I was building new contacts and was exploring developments with Software Defined Wide Area Networking (SD-WAN) and SDN. From what I saw SD-WAN was just an overlay network that used multiple ISPs to provide bulletproof connectivity for service delivery. We had already partially done pieces of this at Really Fast and Honest.net with our tunnelling. It was time to expand the vision of overlay networks.
I became obsessed with the idea of networks that could be consistent over the top of any network and managed centrally. I decided to leave Honest.net to build a new company focused entirely on Software Defined Overlay Networking and building a technology that could easily connecting everything and be deployed anywhere. I wanted to build a platform where anyone could simply provision secure, virtual, over the top networks and services at the click of a button.
I knew we would need to have the following properties: