WEBVTT

0:00:03.180000 --> 0:00:06.520000
 Hello and welcome to this video in which
 we're going to continue our discussion

0:00:06.520000 --> 0:00:10.940000
 of network automation and programmability
 topics by talking about controller

0:00:10.940000 --> 0:00:13.500000
 based SDN architectures.

0:00:13.500000 --> 0:00:17.400000
 Really here I'm just going to be introducing
 to you really fast three

0:00:17.400000 --> 0:00:21.520000
 concepts that you should know which
 are things called underlay network,

0:00:21.520000 --> 0:00:24.640000
 overlay network and SDN fabric.

0:00:24.640000 --> 0:00:27.640000
 You're going to see these terms in a lot
 of documentation, a lot of configuration

0:00:27.640000 --> 0:00:30.020000
 guides, a lot of white papers.

0:00:30.020000 --> 0:00:33.920000
 So let's talk about what do
 these terms actually mean.

0:00:33.920000 --> 0:00:39.720000
 Okay, let's start talking about
 underlay and overlay network.

0:00:39.720000 --> 0:00:48.140000
 So when you implement an SDN controller
 into your network, the idea is

0:00:48.140000 --> 0:00:53.380000
 you still need the network at somehow
 either the controller will program

0:00:53.380000 --> 0:00:57.960000
 it itself or you will program it, but
 the network still under fundamentally

0:00:57.960000 --> 0:01:00.040000
 needs to know routes.

0:01:00.040000 --> 0:01:02.920000
 It needs to know paths, it needs to
 know MAC addresses, it needs to know

0:01:02.920000 --> 0:01:06.340000
 how to get data from
 point A to point B.

0:01:06.340000 --> 0:01:09.660000
 So all that stuff that you're used to
 for the last 20 years about making

0:01:09.660000 --> 0:01:14.180000
 the routes happen, opening up the paths
 so packets can get to their destination,

0:01:14.180000 --> 0:01:15.800000
 all that still has to happen.

0:01:15.800000 --> 0:01:18.320000
 And that's what's called
 the underlay network.

0:01:18.320000 --> 0:01:22.760000
 It's the set of protocols and features
 that you are already used to to

0:01:22.760000 --> 0:01:26.840000
 establish full IP reachability
 between the endpoints.

0:01:26.840000 --> 0:01:31.140000
 Now, in this type of environment though
 where we're implementing a controller,

0:01:31.140000 --> 0:01:32.680000
 there's one key difference.

0:01:32.680000 --> 0:01:38.160000
 So for the longest time networks used
 to be designed like this, where

0:01:38.160000 --> 0:01:46.540000
 you had your access layer switches, okay,
 so maybe you switch one, switch

0:01:46.540000 --> 0:01:54.420000
 two, and switch three, and these access
 layer switches would just be pure

0:01:54.420000 --> 0:01:56.280000
 layer two switches.

0:01:56.280000 --> 0:02:00.360000
 So they might have some VLANs and some
 ports on their VLANs like we see

0:02:00.360000 --> 0:02:07.220000
 right here. And then in the three layer
 design model, they would connect

0:02:07.220000 --> 0:02:11.780000
 up to a couple of distribution switches
 like maybe distribution switch

0:02:11.780000 --> 0:02:18.020000
 one, distribution switch two, has
 some redundant links between them.

0:02:18.020000 --> 0:02:21.960000
 And the distribution switches is where
 your layer three would happen.

0:02:21.960000 --> 0:02:25.720000
 This is where your interface VLANs
 would be right interface VLAN one,

0:02:25.720000 --> 0:02:26.960000
 two and three, one, two and three.


0:02:26.960000 --> 0:02:30.280000
 So this is where your routing would
 take place to route traffic between

0:02:30.280000 --> 0:02:38.220000
 the VLANs. And then between buildings
 in your campus or something.

0:02:38.220000 --> 0:02:44.000000
 Well, in this new methodology, the
 idea is we're going to bring layer

0:02:44.000000 --> 0:02:48.080000
 three all the way down to
 the access layer device.

0:02:48.080000 --> 0:02:57.520000
 So when you plug in your PC or laptop
 to the switch port, either that

0:02:57.520000 --> 0:03:00.180000
 switch port will have an IP
 address on it right there.

0:03:00.180000 --> 0:03:01.740000
 So it will not be a
 switch port anymore.

0:03:01.740000 --> 0:03:06.660000
 It'll be a routed port or there'll be
 an SVI and interface VLAN on that

0:03:06.660000 --> 0:03:10.720000
 access layer switch that will
 do the routing for you.

0:03:10.720000 --> 0:03:14.600000
 But the most of the documentation you'll
 read will say that the individual

0:03:14.600000 --> 0:03:17.980000
 ports on a switch are now
 layer three things.

0:03:17.980000 --> 0:03:20.220000
 They all have IP addresses.

0:03:20.220000 --> 0:03:24.240000
 So that's why it says right here, all
 links are typically configured as

0:03:24.240000 --> 0:03:26.500000
 layer three point to point.

0:03:26.500000 --> 0:03:29.840000
 So VLANs pretty much the
 thing of the past.

0:03:29.840000 --> 0:03:34.160000
 Every switch to switch connection, IP
 address, IP address, point to point

0:03:34.160000 --> 0:03:36.080000
 connection, we're using a subnet.

0:03:36.080000 --> 0:03:39.640000
 Every connection from a switch to router,
 IP address, IP address, we're

0:03:39.640000 --> 0:03:43.440000
 using a subnet. You can start to see here
 how IPV6 would be really important

0:03:43.440000 --> 0:03:46.640000
 because we're burning up a
 lot of subnets to do this.

0:03:46.640000 --> 0:03:52.740000
 So you design your network that way and
 then use common industry standard

0:03:52.740000 --> 0:03:57.900000
 routing protocols like OSPF or ISIS
 to build that reachability to make

0:03:57.900000 --> 0:04:01.340000
 sure that every network is
 known to every device.

0:04:01.340000 --> 0:04:03.280000
 That's your underlay network.

0:04:03.280000 --> 0:04:06.780000
 And like we said, this
 is nothing new.

0:04:06.780000 --> 0:04:09.460000
 Network engineers have been building
 and maintaining underlay networks

0:04:09.460000 --> 0:04:13.000000
 for years. We just haven't
 called them by that name.

0:04:13.000000 --> 0:04:16.740000
 Now, this is where we get to something
 new, the overlay network.

0:04:16.740000 --> 0:04:23.080000
 So in my last video, I was talking
 about how the routers and switches

0:04:23.080000 --> 0:04:24.980000
 had their own control planes.

0:04:24.980000 --> 0:04:28.980000
 They had their own routing protocols, their
 own ways of learning MAC addresses.

0:04:28.980000 --> 0:04:33.080000
 And so they built their reachability
 by themselves.

0:04:33.080000 --> 0:04:37.120000
 And then they would report back
 to the controller what they had.

0:04:37.120000 --> 0:04:40.940000
 So the controller would have the full
 visibility of the topology, know

0:04:40.940000 --> 0:04:44.880000
 what links were congested, possibly
 know the delay from this point to

0:04:44.880000 --> 0:04:46.860000
 the so the controller
 would know everything.

0:04:46.860000 --> 0:04:53.420000
 Then some application would talk to
 that controller and say, hey, I need

0:04:53.420000 --> 0:04:55.800000
 right now the network
 to look like this.

0:04:55.800000 --> 0:05:00.100000
 I need a path from here to here that has
 these characteristics, this quality

0:05:00.100000 --> 0:05:05.720000
 of service. And maybe I need this path
 separate from every other path.

0:05:05.720000 --> 0:05:08.800000
 I want to make sure this path is dedicated
 for me and nobody else can

0:05:08.800000 --> 0:05:12.180000
 use it. And then we said, okay, so now
 the controller is going to reach

0:05:12.180000 --> 0:05:17.740000
 out to the routers and switches with
 the requirements that the application

0:05:17.740000 --> 0:05:20.660000
 needs and say, this is
 what I need from you.

0:05:20.660000 --> 0:05:23.700000
 And then the routers and switches will
 dynamically program themselves

0:05:23.700000 --> 0:05:25.760000
 to meet those requirements.

0:05:25.760000 --> 0:05:30.300000
 All right. So as the routers and switches
 are doing that, as they're changing

0:05:30.300000 --> 0:05:35.240000
 the network, like by tweaking certain
 routes dynamically, adding access

0:05:35.240000 --> 0:05:40.460000
 list dynamically, maybe implementing
 virtual routing and forwarding VRF

0:05:40.460000 --> 0:05:45.120000
 instances dynamically, all that stuff
 that's taking place on top of the

0:05:45.120000 --> 0:05:48.780000
 underlay. So the application
 can do what it needs to do.

0:05:48.780000 --> 0:05:52.040000
 That's what we call
 our overlay network.

0:05:52.040000 --> 0:05:54.040000
 So it's a virtual network.

0:05:54.040000 --> 0:05:57.900000
 A lot of times it's temporary in nature,
 not there permanently, is temporary

0:05:57.900000 --> 0:06:01.640000
 there for the application to use that are
 created by a software and implemented

0:06:01.640000 --> 0:06:03.840000
 by the underlay network.

0:06:03.840000 --> 0:06:10.220000
 So practically, the main technologies
 are used to build the overlay network,

0:06:10.220000 --> 0:06:17.880000
 our virtual routing and forwarding
 instances VRFs, MPLS VPs and VXLAN

0:06:17.880000 --> 0:06:19.560000
 tunneling technology.

0:06:19.560000 --> 0:06:22.880000
 Now, let me stop and pause
 here for a second.

0:06:22.880000 --> 0:06:28.320000
 For the purposes of the CCNA exam, they
 don't expect you to know the details

0:06:28.320000 --> 0:06:35.140000
 of how things like MPLS VPNs and VXLAN
 work, or even VRFs necessarily.

0:06:35.140000 --> 0:06:37.200000
 They don't expect you to know
 how to configure that stuff.

0:06:37.200000 --> 0:06:40.960000
 But they do want you to know the differences
 between an underlay network

0:06:40.960000 --> 0:06:46.740000
 and overlay network and that the overlay
 network is made, it's created

0:06:46.740000 --> 0:06:49.720000
 primarily by these
 three protocols.

0:06:49.720000 --> 0:06:53.060000
 So just be aware that these three features
 and protocols, there's other

0:06:53.060000 --> 0:06:58.080000
 things too, but these are the main ones
 that are used to create that dynamic

0:06:58.080000 --> 0:07:02.400000
 overlay network that our application
 needs to do its thing.

0:07:02.400000 --> 0:07:09.960000
 And then the last term you might see
 is something called the fabric.

0:07:09.960000 --> 0:07:15.160000
 So the term the fabric is just like
 it says here, a term that aggregates

0:07:15.160000 --> 0:07:19.400000
 both the underlay and the
 overlay networks together.

0:07:19.400000 --> 0:07:26.060000
 So it's typically used to describe
 only those devices in the underlay

0:07:26.060000 --> 0:07:30.520000
 network that can be programmed and
 controlled by the SDN controller.

0:07:30.520000 --> 0:07:34.580000
 So let me just draw something here
 that might help explain this.

0:07:34.580000 --> 0:07:45.440000
 Let's say that this represents a set
 of routers and switches that are

0:07:45.440000 --> 0:07:48.440000
 not able to work with
 the SDN controller.

0:07:48.440000 --> 0:07:54.100000
 Maybe they can be learned by the controller,
 maybe they can report their

0:07:54.100000 --> 0:08:00.780000
 statistics to the controller, but they
 don't have any ability to be programmed

0:08:00.780000 --> 0:08:02.340000
 by the controller.

0:08:02.340000 --> 0:08:05.660000
 So if the controller says, hey, I need
 you to make the network look like

0:08:05.660000 --> 0:08:08.660000
 this with these requirements, these
 devices, they don't know how to do

0:08:08.660000 --> 0:08:12.900000
 that. These are legacy devices that
 can't be programmed that way.

0:08:12.900000 --> 0:08:20.580000
 And then we have other devices, routers
 and switches that they can.

0:08:20.580000 --> 0:08:26.080000
 So these pink things, and typically in
 this environment, I probably shouldn't

0:08:26.080000 --> 0:08:30.600000
 have drawn as many, but we have a full
 mesh where everything is able to

0:08:30.600000 --> 0:08:36.060000
 talk to everything else,
 something like that.

0:08:36.060000 --> 0:08:40.500000
 And all these links are point
 to point layer three links.

0:08:40.500000 --> 0:08:42.920000
 So we have IP addresses
 on each thing.

0:08:42.920000 --> 0:08:48.660000
 So it's point to point, IP end to end,
 and the key differentiating factor

0:08:48.660000 --> 0:08:51.980000
 is these devices not only can report
 their statistics and information

0:08:51.980000 --> 0:08:56.400000
 back to the controller, they can
 be programmed by the controller.

0:08:56.400000 --> 0:09:01.100000
 They can be dynamically programmed
 with MPLS VPN and VRFs and VX land

0:09:01.100000 --> 0:09:06.880000
 tunnels to bring up these temporary
 overlay networks.

0:09:06.880000 --> 0:09:11.660000
 So when we use the term the switch, the
 fabric or the SDN fabric, typically

0:09:11.660000 --> 0:09:15.000000
 what we're talking about is
 this set of devices here.

0:09:15.000000 --> 0:09:19.200000
 Usually these devices are not included
 in that terminology of the fabric

0:09:19.200000 --> 0:09:24.160000
 because they can't be controlled
 by the SDN controller.

0:09:24.160000 --> 0:09:27.960000
 So that's what the fabric
 is really talking about.

0:09:27.960000 --> 0:09:33.220000
 A full mesh of devices with multiple equal
 cost paths between their destinations

0:09:33.220000 --> 0:09:38.180000
 and that can be programmed
 by the SDN controller.

0:09:38.180000 --> 0:09:39.920000
 So that's it for this video.

0:09:39.920000 --> 0:09:42.660000
 Hopefully it's a little bit clear now
 in these terms of the differences

0:09:42.660000 --> 0:09:47.680000
 between the underlay network, the
 overlay network, and the fabric.

