Follow me:
Listen on:

Day Two Cloud 109: PacketFabric Wants To Make Networking As Easy As Cloud (Sponsored)

Episode 109

Play episode

PacketFabric lets you provision point-to-point and hybrid cloud connectivity as a service. Built on a private fiber network, the company’s goal is to let you set up WAN networking as if it was software. The company has colocation facilities around the world, as well as connections into Internet exchanges, and cloud and SaaS providers.

PacketFabric is our sponsor for today’s Day Two Cloud episode. Our guest is Anna Claiborne, Co-Founder, CTO and CPO at PacketFabric.

We discuss:

  • Standing up a WAN circuit with PacketFabric
  • Connecting hybrid and multi-cloud networks
  • PacketFabric’s underlay, including dark fiber and DWDM equipment
  • PacketFabric vs. Direct Connect and ExpressRoute
  • Using PacketFabric’s Cloud Routers
  • Protocol support
  • Multi-tenancy in PacketFabric’s network
  • More

Show Links:

PacketFabric Blog

@packetfabric – PacketFabric on Twitter

PacketFabric on LinkedIn

PacketFabric on YouTube



[00:00:04.440] – Ethan
Welcome to Day Two Cloud and man, we got a show for you. We’re going deep in infrastructure as code, dealing with networking, cloud networking, connecting cloud environments, and we’re doing it with sponsor PacketFabric. Our guest today is Anna Claiborne founder. A lot of this is is her thing, her idea, her passion. And Ned, this was one heck of an interview.

[00:00:24.930] – Ned
She really nerded it out and went deep on some of the networking technology. I’m glad you could follow her, because a few times I was like, I’m a little lost here. But then we came back up for air. And the main thing was the tech is cool, right? But the way that they present it to you is way simpler than previous networks you might have had to manage and connect. So that was the main takeaway for me is yes, the cool stuff is there.

[00:00:47.970] – Ned
No, you don’t have to deal with it.

[00:00:49.730] – Ethan
So enjoy this interview with Anna Claiborne. She’s the co-founder, CTO and CPO at PacketFabric. And again, we are going to nerd out a lot, but don’t get lost in that because at the end of the day, the whole point of this is how easy it is to build a multi-cloud WAN that interconnects all the things with PacketFabric. Well, Anna, welcome to Day Two Cloud. We are very glad to have you on the show today.

[00:01:14.640] – Ethan
And you’re a founder. That means this whole PacketFabric thing. It’s at least partially your fault. So let’s go back in time. Set us up Anna. What was the year and what problems were you facing that drove you to start PacketFabric?

[00:01:27.670] – Anna
Well, the the year was twenty fifteen back a long, long time ago in a galaxy far away. And and the the problem was, is it’s kind of a this has been the problem with networking, not just in twenty fifteen, but pretty much forever is that it is it’s a very long process. It’s not an API driven thing, it’s not a modern thing. Network technology is rooted back at the very start of the Internet and it really hasn’t really hasn’t evolved much from there.

[00:02:02.710] – Anna
And so the thing the inspiration for starting PacketFabric was how do we make a network look and feel like software? How do we make it like cloud compute? Because that’s the evolution that we just got done watching back from starting in like twenty ten. I was watching Cloud grow up and going from racking and stacking servers to all of a sudden that you have a UI where you can click to provision or you have an API where you can spin up a thousand servers.

[00:02:33.220] – Anna
And so the basic problem statement was how do we make network like that? How do we do that thing?

[00:02:40.390] – Ethan
If we’re talking twenty ten to twenty fifteen timeframe software defined networking was, we can say, making some inroads. And so maybe there was some inspiration there.

[00:02:49.090] – Anna
There was. But software defined networking really. I mean that term has been I feel like it’s been so abused, it’s been so abused and turned into so many different things. And it was a it was originally meant to be more on like the LAN side for actually like how do you how do you adjust your traffic dynamically according to traffic patterns and things like that? That was I was more the original inspiration for for SDN. And then it just got kind of used everywhere for everything.

[00:03:25.270] – Ned
Right. If we’re talking about that twenty fifteen time frame, I’m thinking I was probably working on some Microsoft Azure deployments at that time and spinning up network stuff inside Azure was amazing. All I had to do was submit an arm template and boom, I had all the Vnets I wanted all the subnets. I wanted every address space.

[00:03:40.990] – Ned
And one project I specifically remember working on was, OK, we have all that, but we want to set up a direct connect, an express route connection to Azure and that took two months. Yeah, I could set up all the networking I wanted in Azure in two minutes, but setting up that circuit two months. So yeah, I agree with that. Kind of sucked.

[00:03:59.770] – Anna
Yeah. Yeah. And that’s like but networking is always the long pole in the tent. Right. Whatever you’re doing it’s always like oh we have to wait for the network. Oh we have to get more network and it’s always this, it’s always this huge time investment both on like the sales side even to get a price for how much does a circuit cost it? It’s a minimum time investment of two months to go in and find a salesperson and negotiate that.

[00:04:29.140] – Anna
And it’s like it’s complete insanity that something still takes that long in this day and age. You know, they’re faxing you a pricelist.

[00:04:39.340] – Ethan
Yeah, the LAN stuff is bad and the WAN stuff that we’re talking about right even worse. And of course, they want you to sign a contract and the lead times and all the rest of it. So I think we’re getting a sense of the problem statement here Anna that you were wanted to address. How can we make network function more like Cloud? OK, there’s the set up, so lay it on us. What is it? PacketFabric in a nutshell, what is PacketFabric.

[00:05:01.240] – Anna
So what is PacketFabric? We are doing for networking what the big cloud providers did for compute and that is the entirety of the elevator pitch and that right now we’re primarily focused on the WAN. So that’s point to point circuits, metro ethernet type circuits equivalence private, like there’s a flavor of private peering in there too, cloud connectivity. So both like hybrid cloud connectivity and multi-cloud connectivity. So making that whole universe of different WAN connectivity, no matter what it is you need to connect to, whether that’s colocation and cloud or cloud and cloud or off a branch office and cloud making that all accessible via an API and a UI so that you can interact with it programmatically or you can go in and tap some buttons and have that connectivity.

[00:05:54.640] – Anna
You know, actually up and running in a minute or you can increase the capacity in a minute. Whatever it is that you need to do, it’s instant.

[00:06:04.220] – Ethan
So you know what’s what’s wrong with this, model Anna? Is I’m really going to miss calling a WAN provider and getting the onboarding process where I’m on with some tech who has no idea about anything about my circuit and has no idea why it won’t come up. And I’m on hold for hours. I’m going to miss that tremendously.

[00:06:21.280] – Anna
Well, what about the camaraderie that you gain when you get to have the really big provisioning calls? Because then. That one didn’t work, so then you get to get a lot more people on call from both sides so that everybody can hang out and chat about what’s not working, that’s.

[00:06:38.830] – Ethan
Well, yeah, yeah. That camaraderie, it’s I mean, that’s important to build that in the industry. Actually, with all of that, you do find that there’s people working around the corner that you didn’t know.

[00:06:46.840] – Ethan
It’s a small world after all, and all that stuff. But yeah, I don’t need that much more camaraderie anyway, OK? There’s so much implied in what you said Anna about how I can use PacketFabric to provision a circuit. And we’re going to get into some more details here, but. I was doing homework on PacketFabric and reading on your site, and one of the things you say over and over again, carrier class, OK, what is carrier class mean? Why do I care about that?

[00:07:12.510] – Anna
Well, like we talked about before, with the whole SDN term being somewhat used and abused is that there’s a whole suite of over the top sort of WAN solutions available today where you can do things really modern things like build VPN tunnels over the Internet. And by modern, I mean not modern, like people building IPSec tunnels over the Internet for ages, but it has been repackaged a lot lately and and sold as it’s a brand new solution that is going to give you this whole wide, wonderful world when in reality you still are at the you you still are at the whim of the Internet and subject to everything that the Internet is subject to.

[00:08:01.170] – Anna
DDOS attacks, man in the middle attacks, dramatic changes in latency, outages. All these things still are are still a big impact, whereas and everyone knows this is that that’s why you go to a private line, private circuit for the applications that are really critical or to move traffic between different collocations facilities on your backbone or to connect to a cloud provider directly. You move to a private circuit, too, so that you can have that low latency, dependable latency connection.

[00:08:33.960] – Anna
And so that’s what carrier class really means, is that we’ve built our network from the ground up. We are not the Internet and we’re not relying on that. We have our own dark fiber. We run our own DWDM equipment. We run our own Ethernet network on top of that to provide services and we built all the software for it. So it’s really the the term is imparting that we are a full fledged telecom network, not not some software that runs something over the Internet.

[00:09:06.870] – Ethan
OK, you built your own thing from the ground up then. Does that mean you’re going to give me some kind of an SLA guarantee?

[00:09:13.050] – Anna
Yes, we have we have an excellent SLA that’s actually available on our website. You can go there. You don’t have to talk to anyone to get it. You just have to go to fPacketFabric dot com and click the SLA link and you can read all about it.

[00:09:27.120] – Ned
And I imagine building a network up from the ground that gives you an opportunity to make a lot of smart decisions and make things a little bit easier for folks who want to consume a circuit from PacketFabric. What does it look like from the consumer standpoint? If I go to your website, can I just spin up a circuit sight unseen or is there a longer, more involved process that I would typically expect?

[00:09:50.340] – Anna
Nope, you can build a point to point one hundred gig from L.A. to New York right now, and it will be ready in one minute.

[00:09:57.000] – Ned

[00:09:59.430] – Anna
If you do or you if you wanted to do a couple of hundred gigs, you could build a multi hundred gig connection over to London. You could build a multiple hundred gigs in Europe. You can do the same in the US and you could build an entire backbone, continent spanning backbone.

[00:10:17.250] – Ned
OK, but I would have to sign up for like a year at least. Right, to get to have that amount of bandwidth allocated to me?

[00:10:24.510] – Anna
Nope. It’s a month to month term like any modern SaaS type thing you would expect. It is a month to month term if you don’t need it next month.

[00:10:37.280] – Ned
Do use it next month.

[00:10:38.480] – Anna
Don’t use it next month.

[00:10:39.890] – Ned
That is very different than my two month experience. So thank you for that.

[00:10:44.680] – Anna
Yes, yes. And I mean, that’s like that’s one of the things that has that I think makes everybody. I don’t know, it’s just one of the antiquated carryovers from telecom is that everything’s locked in four year, multi year. I mean, there’s five year agreements that happen. And that’s not that’s not a modern, modern, agile feeling thing. When you’re signing an agreement for five years, how does anyone really know what their network is going to look like in five years or three years? I mean.

[00:11:20.610] – Ethan
If they say they do, they’re lying. Yeah, no.

[00:11:23.860] – Anna
Yeah, so, I mean, things change so fast, especially with everything with migrations to cloud and moving applications to cloud and trying to figure out what that right balance is between, you know, what do I have in colocation? What do I have in Amazon? What do I have in Azure? What do I have in Google? Like, there’s a lot of dynamic activity happening right now. And to have your network be the thing that takes six months to adjust in between all these changes is just unworkable.

[00:11:53.410] – Ethan
You lead me to the next question here, because the elephant in the room is I need to have my stuff somewhere that PacketFabric is. That is you’re probably not if I have some on Prem facility in, I don’t know, New York somewhere, you may or may not be there, you’re probably not. So what is it that I can actually connect together with PacketFabric? You’ve mentioned some of it, you know, multi-cloud stuff and walk us through how that works.

[00:12:18.160] – Anna
Where we are right now as we are in hundred and ninety sites across Australia, the US and Europe. So those are multitenant colocation facilities. And we’re also connected to all of the major cloud on ramp’s in Australia, the United States and Europe in all those locations. So essentially that means we’re available with every major cloud provider that’s Amazon, Azure or Google, Oracle, IBM. We also have several of the alternative clouds with us, like Vultr and Linode too.

[00:12:49.540] – Ethan
Vultr yay!

[00:12:51.250] – Anna
Yeah, yeah. I love Vultr Linode, Both those are fabulous. And so that means that you can access us from any of those multitenant colocation facilities, any of those cloud providers. We also have deployment’s with some large enterprises where we go in into their private data centers and pop those. And so not only do we bring the whole ecosystem to them, but also they get then access to all these things that they never had the ability to get to before, like connect directly to SaaS providers, like if they’re using Microsoft Office, plug directly into into that data center, they.

[00:13:31.070] – Ethan
That have that kind of stuff is so handy. I worked at a fintech at one point and we had a pop right on site from three carriers were in there. And that was awesome for getting good latency and good performance to certain of our customers. Yeah, that’s that’s cool stuff.

[00:13:46.840] – Anna
Yeah. And this is and this is bringing not just three carriers, this is bringing dozens of carriers and within one PacketFabric pop brings in dozens of carriers, all the major clouds, a bunch of SaaS companies, a bunch of security companies, you know, our entire ecosystem. So it really.

[00:14:06.430] – Ethan
That’s all dark fiber that you’ve lit to connect these facilities.

[00:14:10.540] – Anna
In the metros. It’s all dark. We run all of us. So all of our own dark on the long haul spans, we typically do spectrum so that we’re actually getting to a specific wavelength that way, yeah we’re getting a light wave to do the really long spans. So those are cross country spans. And then there’s some there’s some little corner cases where we’re using just like traditional 100 gig waves from from service providers because we can’t get either of the aforementioned two preferred things.

[00:14:45.520] – Ned
So you mentioned you have connection into all the major clouds. And typically, if I wanted to connect in to a major cloud, you know, if I was using AWS, I would provision a direct connect circuit into my existing network. What would the advantage be of using PacketFabric to get that connectivity into AWS as opposed to going and getting the circuit myself?

[00:15:07.670] – Anna
So I don’t know how I don’t know how much you’ve done this, but it like every cloud providers, whether it’s direct, connect, express route, every cloud provider has a different program for this. They have different pricing. They have different APIs, a different UI. And it’s also a if you’re doing it yourself, it is a slow process. You have to go. You have to get the LOA from them. You have to run the cross, connect. The big clouds, make it abundantly clear that they don’t prefer to deal with individual customers and will make that experience less desirable.

[00:15:46.100] – Anna
So you’re usually talking weeks, which which is fast for the networking world. Right. You know, you’re talking weeks to work through these issues and finally get your circuit set up and running. Where with PacketFabric. You literally just go to the web interface or the API. And again, 60 seconds later, you have your ten gigs of express route that you can start using right now. So there’s a huge time to market component. And then also just dealing with, again, if you’re going to be sizing or changing that circuit up and down or if you’re dealing with multiple clouds, it’s really nice to have one interface, one user interface, one API to deal with all those multiple clouds instead of having to remember the caveats and the the quirkiness and the unique snow-flakiness of every one of those things that you’re dealing with. So that’s one of the things that I’ve heard from customers that especially in the multi-cloud environments, they really like the fact that they are able to use single API, single UI to cover everything that they’re doing.

[00:16:47.240] – Ned
OK. OK. And would I ever use one of the other things you can do is you can have direct connect to multiple AWS regions, so maybe I have one to US east and then one over and EU. Is that something you can set up for me as opposed to me trying to manage that mess as well?

[00:17:06.110] – Anna
It depends on the situation. It depends on what you’re trying to accomplish. You can use fabric network to use that to go over to Europe and then actually on board to the cloud network there. And it really depends on what sort of latency goals you’re trying to achieve, because a lot of times the one of the primary drivers, and we actually have a cool video that shows this, too, is for Direct Connect. People just think about it in terms of like latency.

[00:17:30.020] – Anna
Oh, I want a low latency connection. But the thing that is not thought about a lot is the fact of that predictable latency is what yields better throughput for TCP base transfers. Right.

[00:17:41.780] – Anna
So, you know, when latency changes your constant, that’s going to affect your actual TCP throughput. And the longer that latency is, the worse the throughput gets. But especially so when it’s short and stable. That’s when you get the best throughput and when you’re doing especially if you’re trying to run applications across clouds or if you’re running a hybrid cloud type applications where you have something in a colocation and something in a cloud, that throughput of data becomes really important.

[00:18:10.880] – Ned
Is there anything you can do about the egress fees on AWS?

[00:18:14.420] – Anna
Yeah, actually, actually, actually I can. So the great thing about the egress fees is that with every cloud provider, they are substantially less over their direct connectivity product than they are over the Internet. So, yeah, so not only do you get like these better data transfer speeds, but you’re paying less ultimately for the data transfer. So it’s cheaper and faster.

[00:18:40.310] – Ned
Woah, I didn’t think that was allowed. Isn’t there some sort of like pyramid where you’re not allowed to do this?

[00:18:44.920] – Anna
I know.

[00:18:46.520] – Ethan
So Anna, I want to get under the hood here. What is this thing that I’m buying from PacketFabric? Describe the services it like. Is it L3 VPN? Is it VPLS? What’s happening down there with all this automation and magic?

[00:18:58.070] – Anna
So it depends on the product. Let’s start with a point to point right. So a point point is just a it’s a martini pseudowire.

[00:19:04.520] – Ethan
So that would be L2 then pseudowire. Looks to me like a switchport.

[00:19:10.850] – Anna
Yes. And you can pass pretty much. I mean, there’s like a couple of really like crazy edge case exceptions, but any layer 2 protocol, anything that you could pass over a martini psuedowire, you can pass over that. And so it’s great for that’s something that’s very comparable to buying a wave or any other telecom type circuit that you would buy. That is the just connect point A to point B, and I want to throw a bunch of junk over it.

[00:19:37.700] – Ethan
OK, so that’s an option. Give me give me some other options.

[00:19:42.740] – Anna
So that’s one virtual circuit. So the point to point is one virtual circuit, right? It’s a point. It’s a portal and A side portal on the B side, one virtual circuit between it. But then we also have and that’s what’s commonly referred to in MEF definitions of EPL, Ethernet private line. And then we also have an EVPL based product, virtual private line.

[00:20:04.790] – Anna
So you can do multiple connections from one port. And so what that means that you can do is if you have one port in say, a colocation facility, you can use one virtual circuit to connect to Amazon. You can use one virtual circuit to connect to your other colocation facility, whether that’s in the metro or across the country. And then you can use another virtual circuit to connect to Azure.

[00:20:30.830] – Ethan
You’re saying virtual circuits. So not a big broadcast domain like a VPLS, it doesn’t sound like.

[00:20:36.050] – Anna
No. No, it’s the virtual circuit itself. The easiest way to think of it, it’s just it’s a VLAN, right? So you’re getting yeah. You’re getting just the VLAN. Every virtual circuit is on its own isolated VLAN. And those VLANs also don’t have to match coming in and going out because we’re we’re popping the tag and pushing on a new one so you can do your own independent VLAN schemes in between your internal network and what’s going over PacketFabric too, like you can match them. Not match em.

[00:21:03.140] – Ethan
Just Q in Q?

[00:21:04.160] – Anna
We do Q in Q as well. And so that was initially developed because of Microsoft ExpressRoute actually, you know I was talking about each of the cloud providers being their own special snowflake. Well, Microsoft did express route in a way that is not like any of the other cloud providers. And so you have to use S tag on it. And so we developed Q in Q functionality for that. And that’s part of our ENNI product lines that we try to stick to the MEF definitions, even though the MEF definitions drive me crazy.

[00:21:37.320] – Anna
It’s like the great thing about standards is that there’s so many to choose from, like my favorite line ever. But anyway, we decided, you know, we use MEF because it’s generally well understood.

[00:21:49.830] – Ethan
Yeah, MEF for the in case there’s Day Two Cloud listeners who don’t live in the networking space. Metro Ethernet Forum, right Anna?

[00:21:56.100] – Anna
Yes, yes. OK, one of the multitude of standards bodies out there in networking.

[00:22:00.720] – Ethan
Right. So many. So you’ve described a point to point service to point to multipoint service, which are both layered 2-ish is there a straight up layer three service?

[00:22:09.250] – Anna
Yes. The layer three service is our cloud router service and we’re running a VIRF for you. So now we have you have your own isolated independent router that’s running and you can connect that BGP to Cloud, colocation, and new product. We’re coming out with that. It’s not out. It’s going to be out in two weeks. I know, because earlier you said, what does this do for the branch offices that were jumping on the IPSec bandwagon? And so we will have a IPSec VPN connectivity, too.

[00:22:41.730] – Anna
So you can now connect your cloud, your branch, your colocation all together, layer three. And it’s got a beautiful, of course, UI and API on it. So you don’t even have to worry about the details of BGP anymore. You just you’re going to fill in a couple of inputs. You know, it’s like two screens and you’re going to hit connect and then it’s going to go ahead and build that BGP session with the cloud or the router. Your colocation can do the IPSec tunnel to route to your branch office and then all those things will be connected. Layer three.

[00:23:12.840] – Ethan
Boy there’s there’s there’s a lot here. Now, if I got a layer 2 service from you, well, I got it back up and ask a question first. You talked about a port. So if I get a layer 2 service of some flavor from PacketFabric, well, there is no port. I mean, I’m right. So what am I plugging into? How am I connecting?

[00:23:28.020] – Anna
So for for the layer 2 services, those all all require an actual port. So your your in a it requires a port somewhere. I should say that not necessarily yours, but you are going to ultimately be terminating that to a port.

[00:23:45.030] – Anna
So it’s say that you’re in Equinix San Jose for example, you have some colocation area, you have a rack there and Equinix San Jose and you’re like, well I need to connect to Google and I want to connect to Linode. And I also want to get some application from Microsoft. And so you would order a port with PacketFabric, you would run a cross connect to that port, and then you could build all of those layer 2 services. So you would build a connection to Google, you would build a connection to Microsoft.

[00:24:17.370] – Anna
And then what did I even say was the third one? I can’t remember. Now, you would build three different layers 2 connections on three different VLANs from that port and have all that.

[00:24:29.340] – Anna
And that’s really centered more around the colocation universe, the cloud router, the layer three products that’s centered more around the cloud native, branch office universe. Those are all just BGP based. We already have all the underlying connectivity set up to the cloud providers everywhere. Like we’ve already we’ve already done the hard part.

[00:24:51.150] – Anna
We’ve already run the cross connect. We’ve already done all that. So all you’re doing is building BGP session over to the cloud providers and getting getting the routes, the IP addresses that you’re actually going to use from the cloud and that’s it. Then you have a path there and same thing with connecting branches over IPSec.

[00:25:09.990] – Ethan
So just like Serverless is made up of servers , it turns out networking with PacketFabric is made up of ports. There’s ports involved.

[00:25:17.220] Hashtag networkless. Yes. I love it.

[00:25:21.390] – Ethan
I’m assuming if it’s a layer 2 service, you don’t care what I push over that thing. So I could do IPv4 traditionally, of course, but also IPv6.

[00:25:30.150] – Anna

[00:25:30.600] – Ethan
So so let’s extend that then. L2 straightforward enough. But what about L3. The cloud router service and BGP is IPv6 supported?

[00:25:38.220] – Anna
It is supported with some provider. I think Google has full v6 support.

[00:25:44.730] – Ethan
It gets sketchy when you get into the cloud service and what they support in it and what they mean by support, you know.

[00:25:50.410] – Anna
Yes, it’s always in quotations, always in air quotes of yeah. And again, back to the snowflake. Thing is, this is why it’s kind of nice to have somebody to deal with that for you because, well, we will allow v6 where it is allowed and supported and not allow it where it is not allowed and supported.

[00:26:09.780] – Anna
And I don’t like off the top of my head, I can’t remember which clouds it is, but there is a huge difference. I mean, there’s a huge range of IPv6 support in there.

[00:26:19.470] – Ethan
Yeah, we did a show on on some of that with and with an AWS networking human not all that long ago. Goes back a few months Ned. But that was that was with Nick Matthews. I think we talked about where that was heading. But from a PacketFabric perspective, just let alone the endpoints and what you’re connecting to. But from a PacketFabric perspective, if it’s supported on the other end, you support it too. Yes.

[00:26:44.430] – Anna
Yep. Yeah, yeah, yeah.

[00:26:46.230] – Ethan
OK. That’s that’s that’s kind of big stuff.

[00:26:51.690] – Ned
If you’re so obviously you’re running on some sort of shared systems, if I’m getting a port from you, other ports are dedicated to other tenants, but is it always a single physical port is dedicated to a specific tenant or is it sometimes ports are shared across tenants? Like what’s your approach to handling handling multi tenancy in your environment?

[00:27:12.780] – Anna
So some ports are like all of the customer owned ports are single tenant. Obviously because they are customer owned. We do have multiple tenancy on the ports that face cloud providers. You can do like we have two flavors of that. So we call it hosted and dedicated, dedicated means that the port facing the cloud provider is entirely dedicated to you and you alone. It provides several different advantages, like you can run a ton of VLANs over that. If you have one hundred VPCs, which we do see that you use case where people have hundreds of VPCs and you need a separate VLAN to go into every single VPC you can have a dedicated hundred gig port into your cloud provider and that four hundred different VLANs to go to each VPC.

[00:28:01.050] – Anna
So that’s a great option for that. But multi tenancy comes in, is in our hosted product and that’s where you’re just getting a single virtual circuit. AKA, a single VLAN into your into your cloud provider and we run those. So we have a shared port facing the cloud provider and multiple customers are run across that and everyone’s isolated in their own VLAN. So it’s still very private. Safe to share all that. But that’s the that’s the biggest case of multi tenancy.

[00:28:28.530] – Anna
And so that applies for both the layer 2 like the hybrid cloud product and then also the layer three cloud router product. You would be going across one of those multitenant ports.

[00:28:37.950] – Ethan
So if I if I have an auditor come in the door, they can have all this information from PacketFabric and understand exactly how the multitenancy is being delivered and the different security mechanisms are in place, all of which some very standard and very normal for any kind of a service provider network. So. We had talked about latency earlier and that came up, you know, if I needed to craft a path that had, you know, predictable characteristics for TCP transport.

[00:29:04.030] – Ethan
So, OK, does that mean the link you provide because you’ve just got so much bandwidth? I know it’s going to be predictable or does that mean, hey, if I set TCP values and hand you my very precious traffic that I need handled oh so specially you’re going to honor that and do magic to to to carry in a special way.

[00:29:23.960] – Anna
So, no, we’re not going to do and we’re not going to do any sort of inspection like that and honor and and we’re not going to look at your traffic in that way. We are. We’re going to introduce some features here later on this year around QoS so that you can set probably something. This is still something that we’re working on. But it’s a it’s a common request that we’ve heard like a tri color marking type QoS support. So that’s that would be the type of special traffic handling that we would do.

[00:30:01.400] – Ethan
And could you actually do that? I mean, so if that’s a feature that’s been added, you know, let’s say I hand you EF marked traffic or, you know, AF41 for video or whatever the value is, you could prescribe a path that’s going to give me low loss, low jitter kind of a thing.

[00:30:19.970] – Anna
Ew, now you’re getting now you’re getting into the really into the really crazy stuff. So this is some of the stuff that we are thinking about.

[00:30:27.890] – Anna
And I don’t want to go too deep into it because this is a lot of our special this is a lot of our special sauce here. But obviously, since we’ve you know, we designed we control our entire network. We write our own software with something like, say, segment routing. We can start doing these types of things to get really creative, like if you like, if say that you only want your traffic to run through a particular country, you can start to do things like that or say that you want some of your traffic to take a very particular path according to latency flavor. We can start to do things like that.

[00:31:04.190] – Ethan
But you need a controller and you need something that knows the condition of the global network end to end. Kind of a big deal. Yeah.

[00:31:11.890] – Anna
Uh. Yeah, yeah, yeah, yeah.

[00:31:14.270] – Ethan
And if you say SR, are we talking SR-MPLS in your case or SR v6?

[00:31:19.640] – Anna
We’re talking SR-MPLS in our case.

[00:31:21.590] – Ethan
Ok that, that makes sense from the other things you just said about the network.

[00:31:24.890] – Anna

[00:31:25.970] – Ethan
So PCE controller shove a bunch of tags on there based on dynamic changing conditions. Oh that’s a lot. That’s a lot. That’s a lot of automation, especially if you got a whole bunch of people taking you up on that tricolor offer.

[00:31:39.500] – Anna
We we have we have a lot of really smart people in our R&D group.

[00:31:44.240] – Ethan
So so the question then is, do you really need to offer that or is your backbone just so massive with bandwidth that offering the services is in a sense pointless because you’ve just got that much pipe to move traffic around predictively?

[00:31:59.300] – Anna
I mean, it’s both right. Because, you know, on one hand, we do have we do have a huge amount of bandwidth. You know, it’s like forty terabytes now. It’s probably more you know, it’s more than the last time that I even looked, though. We do have just a huge amount of bandwidth available. And we also like the way we build the network is that we have a minimum of two diverse paths out of every single colocation facility that we’re in.

[00:32:24.540] – Anna
So and in most cases, it’s many more can be up to five different fiber paths out of there. So fiber cuts and things like that, it’s just, you know, it’s a BFD timer. 15ms away from a re-route. And so it’s like you don’t mind a little bit of variation and you’re not particularly choosy over if there’s not something you need to avoid with your network traffic.

[00:32:49.770] – Ethan

[00:32:50.210] – Anna
Then yeah, the network works flawlessly, beautifully. We really don’t go down. It takes a huge, huge event, multiple, multiple fiber cuts for anything bad to happen. But there are obviously cases who are more selective about where their traffic goes or they need a they need a fixed latency, especially in like video or games. And there’s a lot of use cases out there like that.

[00:33:14.150] – Ethan
Yeah, but that is pretty specialized. You know, those customers that need that, I get that. But some of them are big enough that maybe you do want to offer that feature because they’re going to pay for it. So.

[00:33:25.070] – Anna
Yes. Yeah, yeah. When you need something like that, you need it. So yeah, there’s, there’s a good case to do to to do creative things like that.

[00:33:34.250] – Ethan
And you did answer the other part of the question, which is physical redundancy. You’ve got multiple egress points coming out of each of your POPs. So you could do the traffic engineering if you if you really wanted to add that.

[00:33:45.260] – Anna
Yeah, yeah. We we absolutely could. And we’ve kind of been blessed in a way, I guess, in that we are so overprovisioned with capacity is that we can get a little lazy and just sort of shove everything over the most optimal path all the time. But that is we do have like we have some really pretty sweet, like in-house written capacity management tools that go through and look at the entire network at all times. And we can run we can run failure scenarios actually, like taking down individual links and see how that will impact everything. So we have really cool stuff available. But in order to we haven’t really gotten to exercise it to its fullest yet.

[00:34:24.110] – Ethan
Are you doing any fancy routing today or traffic engineering beyond the modeling you were just describing or you just doing, you know, straight up iBGP, BGP on top MPLS?

[00:34:35.180] – Anna
Yep. Yep. Just yeah. BGP and MPLS. Nothing. Nothing. Nothing incredibly fancy there yet. We are running SR on all of our edges. Like it is like we are running it and we’re starting to experiment with a lot of the stuff that we just talked about so.

[00:34:52.400] – Ethan
Oh, well if you have I mean that’s a big step if you’ve already gotten that far to have SR. OK, interesting. This isn’t yeah. This isn’t pie in the sky. It’s more of the just the though.

[00:35:03.540] – Anna
Yeah. Yeah. Definitely not pie in the sky. I mean it’s something that we’ve been seriously like in terms of evolution of our network. Like we move incredibly fast, like I’m actually like, you know, most most service providers you still have, you know, Cisco GSR running in there from like nineteen ninety or whenever they were last manufactured. But we use multiple vendors for everything. So because it’s just a software integrate, every new vendor is just software integration away.

[00:35:34.190] – Ethan
Oh you so fancy. I know you said something so big there we just. Oh another vendor is just an integration away that says so much about your platform on the back end and what you’ve done to facilitate that, because multi vendor in that environment is really hard because of the processes and the automation typically tied to a vendor, maybe two vendors. If you’re just like, yeah, we got an abstraction layer pretty much. Did we just do whatever?

[00:35:59.060] – Anna
Yeah, we do. We have a super elegant abstraction layer and all we’re doing is plugging in and we’re just plug it in on the back end the exact RPC commands or API or whatever or whatever the vendor is that they actually support. We’re just plugging that in. And so it really is just a module where it’s trivial, really.

[00:36:21.770] – Ned
It’s trivial, she says.

[00:36:24.410] – Anna
It’s our automation, though, is what gives our customers the supreme benefit of having that same automation that they can put to work for them. Like that’s sort of the big drawback of networking services, right? That there’s no API that exists where you can play around with things and spin things up and down. You know, your your DevOps crew can’t put something in Ansible or terraform to do that until now.

[00:36:49.520] – Anna
That’s like, and that was one of the huge benefits of cloud computing, is that it became more of a playground too where you could experiment with things and try things. And now you can do that with the network through through PacketFabric. You have an API to start playing with and experimenting with and doing some of these things that were previously pie in the sky, like bandwidth on demand. You do need bandwidth up and down, you know, to a cloud provider, for example, like that’s now possible.

[00:37:16.610] – Ned
One thing I want to ask about is we’ve talked about how there’s some automation available for for your services. It’s not just go into the UI, clicky, clicky, and it’s not necessarily CLI. So what are my automation options? Do I do this through Python? Is there something for TerraForm? Give me give me some options here.

[00:37:34.490] – Anna
Yeah. So that’s, that’s the great thing. That’s a great thing about a nice REST based API right is the you have all the options. So since we do a lot of python and we do a lot of python internally, we have a Python code examples available on GitHub. We don’t we don’t quite have a full SDK yet. We like to, but we don’t. So we have those code examples available. It’s pretty easy to plug into for the for the SRE DevOps crew, it’s pretty easy to plug the API into Ansible and terraform.

[00:38:14.030] – Anna
Oh, that’s the cool thing we have coming up. We actually have an API cookbook. We have a Python API cookbook coming out that’s going to walk through a ton of different examples of how to do things. And it’s in this really cool Google tool that I’m blanking on the name of now, but basically has all the code and you can run it and step through it actually like the code in a Web browser, and it’s super cool.

[00:38:39.260] – Ned
OK, so lots of different options if I just want to go directly against the API, I’ve got that available to me, but I’ve got some other options as well. And it sounds like you’re building out quite a catalog. I’d love to see that SDK at some point, but that’s a big investment.

[00:38:55.370] – Anna
Yes, it is. And I mean, that’s that’s the goal, right? Ultimately, we have the same the whole same suite of tools available, like any SaaS provider’s API. You go there, you’re just clicking on it. You know, you’re just looking at different different flavors of code examples and in your language of choice.

[00:39:12.830] – Anna
And that’s what the network becomes, is it’s the whole you know, the whole network to code movement. And this is this is I mean, this is it. You know, if you want your WAN to be code and this is how it happens.

[00:39:29.090] – Ned
Yeah. But I mean, there’s a big move to to have like an automation pipeline that’s responsible for deploying and maintaining your network environment and your cloud environments. That’s like this. And keeping everything in source control and then doing it through something like GitOps, I know I’m throwing a lot of buzz words out here.

[00:39:47.460] – Anna
It’s like the whole CI, you know, everything’s everything’s CI/CD now, right in your CI/CD pipeline. And now for the first for the first time, your network can be part of that CI/CD pipeline. Like it’s it’s not controlled by people with fax machines. It is an API. You can you can integrate that wherever you want.

[00:40:10.780] – Ethan
You taking me back to the days when I used to work for an ISP and I did a lot of domain registrar stuff and I was having to fax things on company letterhead, overdue network solutions to migrate a domain and stuff like that. Oh, it was it was it sucked back then and that was almost 20 years ago.

[00:40:24.840] – Anna
Like my first job in high school was maintaining frame relay, like all the frame relay and ISDN lines for a bank which had a ton of branches. And so yeah, that was like, oh man, was on the phone so, so much with them and faxing things and. Yeah, yeah.

[00:40:46.260] – Ethan
Well Anna, Ned and I think we’ve gotten through our questions that we had for you and had a fantastic conversation. Is there anything about PacketFabric that we didn’t know to ask that you want to brag about or that you think is super cool and you want to highlight.

[00:40:59.460] – Anna
Do we have another hour? I like I could go on. I mean, I can talk forever about our internal technology because it really is. It really is that cool.

[00:41:12.840] – Ethan
Let’s do a forward looking thing then. Is there a roadmap of features that are coming, something like what’s the next few things that are coming that you’re super excited about?

[00:41:21.650] – Anna
So there’s a lot that I’m excited about and there’s also a lot that that our marketing department will probably kill me. But the one that I can say is the one that I can I did already mention it before and I’ll mention it again, because it is it is imminent here in the next few weeks is our VPN, IPSec VPN connectivity, you know, bringing that to cloud router. So, you know, it’s your basic nice site to site VPN so you can bring all of your branches into the connectivity ecosystem to do things like connect the cloud to your branch and at least the best way possible.

[00:42:03.140] – Anna
So you’re offloading that traffic off the Internet, off that IP, IPSec VPN as quick as possible, getting onto private backbone and directly into the cloud provider. So, you know, you’re you are stabilizing the connection and reducing the cost of the connection for those branch offices all at once. So I think that’s pretty cool.

[00:42:21.620] – Ethan
Excellent. And we’re recording this. Yeah. This is this the show. If you’re listening to this, it should have come out the fourth of August, and that means very soon that that service is going to be ready and available to you. So Anna if people want to find out more about PacketFabric, where would you send them?

[00:42:37.910] – Anna
Dub, dub, dub dot PacketFabric dotcom

[00:42:41.750] – Ethan
PacketFabric, dot com. OK, and now if I’m a if I’m a little company and is PacketFabric just not for me, it’s really just for big companies. Or can anybody use PacketFabric? Give me the range of appropriate customers.

[00:42:53.930] – Anna
That I mean that’s one of the fabulous things about network, just like compute, it applies to everybody. You know, there there is no there is no such thing as too small use case or too big a use case.

[00:43:05.060] – Anna
And whether you just if you need 50 megs to Amazon or you need a terabyte of backbone connectivity for spanning Europe and US, both those use cases equally valid for our network, and they’ll both come with a nice API where you can do everything really fast.

[00:43:27.110] – Ethan
And you personally Anna, are you too busy coding or are you public on the Internet where people could follow you on Twitter or something? Maybe you’ve got a blog or a book you want to talk about.

[00:43:35.790] – Anna
I wish I had time for that also. I also wish that I was still coding. It’s been like about a year, I think, since I’ve done any coding, unfortunately. But no, I don’t really have like a like pretty much everything is right now. I’m investing heavily in creating the next generation of PacketFabric products so that we can do something even more revolutionary for the network and bring it out of the Dark Ages.

[00:44:08.780] – Ethan
Anna Claiborne founder at PacketFabric, thank you very much for being our guest today on Day Two Cloud. If you’re listening, PacketFabric is all over the interweb, you can find them again at PacketFabric dotcom. They got a blog. They’re on Twitter at PacketFabric. You can find them on LinkedIn. They’ve got a YouTube channel. Just Google the things you know how to find them from there. And our thanks to PacketFabric for sponsoring Day Two Cloud virtual high fives to you for tuning in.

[00:44:30.560] – Ethan
If you visit PacketFabric tell them you heard about it on Day Two Cloud part of the Packet Pushers podcast network. If you have suggestions for future shows, Ned and I are listening. You can hit either of us up on Twitter at Day Two Cloud show or fill out the form on Ned’s fancy and newly redesigned website Ned in the cloud dot com. And if you like, engineering oriented shows like this one, because we know you do visit packet pushers, dot net, subscribe all of our podcasts, newsletters and websites are there.

[00:44:57.710] – Ethan
It is nerdy content designed for your professional career development. And until then, just remember that cloud is what happens while it is making other plans.

More from this show

Day Two Cloud 147: Google Cloud Is Not Just For Devs

Today on Day Two Cloud we peel back the curtains on Google Cloud with a GCP insider to find out how Google Cloud differentiates itself, its embrace of a multi-cloud approach, and more. Our guest is Richard Seroter, Director of Outbound Product Management...

Episode 109