HN846: How Meter Built a Topology Design Tool to Speed Deployments from Hours to Seconds (Sponsored)

Ethan Banks (0:00)

Welcome to Heavy Networking from The Packet Pushers. If you know OSPF pretty well, and have been meaning to learn ISIS for years because I know OSPF, how hard can ISIS be? But you still haven't gotten around to it? Yeah, you found your tribe. I'm Ethan Banks along with co-host Drew, Conray Murray, follow us on LinkedIn and talk to us via the Packet Pushers community Slack group, which you can join for free via packetpushers.net/community.

On today's episode, we are chatting with some developers from Meter about an internal tool that they built to generate network topologies for customer designs. And you might like to build such a tool yourself, or perhaps you're getting involved in network automation, and you'd like to learn a few things from folks who are proper developers. Joining us today are Alex La Fazia and Daniel Kaczmarczyk from our sponsor Meter. Alex and Daniel both worked on this project, and they have different angles at which they came from it and ways that they interacted with each other. And we're gonna chat with Alex and Daniel today about this.

And Alex, the first question goes to you. If you would explain this internal tool that you guys built in a nutshell, what's it do? Why did you need it?

Alex La Fazia (1:04)

Yeah, so really, Meter, you're supposed to just give us an address, and we're gonna build you a reliable network.

Ethan Banks (1:11)

An address like a physical street address.

Alex La Fazia (1:13)

Yes, that's correct. That's the goal. So with that, we have to come up with the design. That's one of the first things we have to do. That's every device, connection, SFP, how the MDFs connect with the IDFs. That all has to be designed and configured. And we can't really have that go wrong. So, you know, a wrong label or a wrong cable can cause some pretty bad issues out in the deployment. And at scale, those problems grow.

We initially started out with just normal, lucid diagrams. We had a team of folks building topologies by hand.

That, I mean, some of them, some of the large campuses, college campuses, took nearly 100 hours to develop. And it still was taking place with hand building topologies. So with that, that means human error is going to be involved. As we're coming out with the new Gen 1 equipment, one of the things that I was put on, coming in as a network engineer, was to standardize these topologies for the next generation of equipment. As we're doing that, we're seeing kind of a deterministic way of building these topologies. The meter topologies, I should say, because we're a full stack solution, we get to dictate that topology. We are after all like a utility.

So we built a tool. That tool turned around those topologies that took 100 hours until like seconds with just a few inputs.

Ethan Banks (2:50)

A couple of words you said there I wanted to key on. One was at scale. So what kind of scale are you dealing with?

Alex La Fazia (2:58)

We are scaling to like, we need to deploy hundreds of networks in a given date. Even let's say we have to build a lot of topologies and they all have to be very accurate and standardize.

Drew Conry-Murray (3:17)

And the equipment we're talking about, switches, APs, other things, cabling, you said?

Alex La Fazia (3:23)

Cabling, we need to know basically this is, we need to build a topology with this tool that is going to build the rest of the dashboard and be handed off for not just our internal teams, our support teams, but the installers. So that's everything that they need to build that network from the ground up.

Drew Conry-Murray (3:42)

Okay.

Ethan Banks (3:43)

So this isn't a Vizio diagram. There's a lot more to this. This is a bill of materials. This is cables and labels. This is everything required from soup to nuts to complete the installation and have at the end of it a functional network.

Alex La Fazia (3:56)

And to kind of give, I guess, a little bit more into that is when we build this topology with these high level inputs, we don't just get a topology with those connections. We literally get the virtual devices in the dashboard, their labels, the port labels, how they're going to connect, the rack elevations, everything gets created from this one input.

Daniel Kaczmarczyk (4:18)

And just to say what Alex means when he says one input, we are talking about simply describing the number of switches and number of IDFs and what is an IDF. It's a very small amount of information you have to provide.