WEBVTT

0:00:02.960000 --> 0:00:09.380000
 Welcome to this video, which I've titled
 Controller-based SDN architectures.

0:00:09.380000 --> 0:00:12.580000
 We have three main topics that we're
 going to go through fairly quickly

0:00:12.580000 --> 0:00:16.500000
 in this video. We're going to talk
 about what is an underlay network,

0:00:16.500000 --> 0:00:22.260000
 what is an overlay network, and what
 is this thing called an SDN fabric.

0:00:22.260000 --> 0:00:26.420000
 So let's talk about these from a standpoint
 of architectural elements.

0:00:26.420000 --> 0:00:29.700000
 We're going to talk about underlay
 networks, overlay networks, and the

0:00:29.700000 --> 0:00:31.800000
 SDN fabric. So what are they?

0:00:31.800000 --> 0:00:35.120000
 Let's start with the first
 one, the underlay network.

0:00:35.120000 --> 0:00:43.120000
 So the underlay network is basically,
 like it says here, the protocols

0:00:43.120000 --> 0:00:50.300000
 and features established to have full IP
 reachability between your endpoints.

0:00:50.300000 --> 0:00:56.640000
 So, fancy way of saying, you've physically
 connected your routers and

0:00:56.640000 --> 0:01:01.100000
 switches together, so those routers and
 switches, they're physical links,

0:01:01.100000 --> 0:01:04.500000
 they're their NIC cards, that's
 part of the underlay network.

0:01:04.500000 --> 0:01:07.480000
 That's all your layer one and your
 layer two if we're talking the OSI

0:01:07.480000 --> 0:01:12.340000
 model here. Then your underlay network
 is also responsible for giving

0:01:12.340000 --> 0:01:15.180000
 IP reachability end to end.

0:01:15.180000 --> 0:01:19.260000
 So whatever that is, whether you're configuring
 EIGRP, OSPF, maybe static

0:01:19.260000 --> 0:01:23.840000
 routes, RIP, so you've got some routing
 protocol now running on all your

0:01:23.840000 --> 0:01:29.340000
 layer three switches and your routers
 to provide end-for-end reachability.

0:01:29.340000 --> 0:01:33.620000
 So once that is done and from any point
 on the network you can ping any

0:01:33.620000 --> 0:01:38.620000
 other point, you have now created
 your underlay network.

0:01:38.620000 --> 0:01:43.960000
 So from a software-defined networking,
 so I'm an SDN perspective, an underlay

0:01:43.960000 --> 0:01:47.740000
 network typically has certain
 characteristics.

0:01:47.740000 --> 0:01:49.620000
 What are those characteristics?

0:01:49.620000 --> 0:01:54.140000
 Well, number one, all links typically
 configured as layer three point

0:01:54.140000 --> 0:01:57.920000
 to point. So if you studied switching,
 you've probably learned that you've

0:01:57.920000 --> 0:02:01.300000
 probably seen a lot of topology diagrams,
 they have three, four, five

0:02:01.300000 --> 0:02:03.440000
 switches and they're
 all layer two.

0:02:03.440000 --> 0:02:07.140000
 All those switches are maybe in some
 VLANs with VLAN trunks and ether

0:02:07.140000 --> 0:02:10.580000
 channels and maybe there's really not
 any layer three in those switches,

0:02:10.580000 --> 0:02:13.100000
 maybe the first layer three
 is a router up here.

0:02:13.100000 --> 0:02:16.540000
 So you've got four or five switches
 all doing layer two and this is the

0:02:16.540000 --> 0:02:24.460000
 very first doing routing.

0:02:24.460000 --> 0:02:28.400000
 Those links connecting switch one to
 switch two, those would not be VLAN

0:02:28.400000 --> 0:02:33.160000
 trunk links, those would be layer
 three point to point routed links.

0:02:33.160000 --> 0:02:36.660000
 So every single link between a router
 and a router or router and switch

0:02:36.660000 --> 0:02:41.680000
 or a switch and a switch would be a
 layer three routed point to point

0:02:41.680000 --> 0:02:46.080000
 link. That is a typical design consideration
 when doing an underlay network

0:02:46.080000 --> 0:02:49.460000
 in software-defined networking.

0:02:49.460000 --> 0:02:53.200000
 You would use some common industry
 standard routing protocols.

0:02:53.200000 --> 0:02:58.640000
 Most underlay networks utilize either
 OSPF or ISIS because they're very

0:02:58.640000 --> 0:03:03.240000
 well known, pretty much any vendor's
 equipment supports them.

0:03:03.240000 --> 0:03:06.360000
 We're not really going to look at proprietary
 routing protocols here when

0:03:06.360000 --> 0:03:08.080000
 building our underlay network.

0:03:08.080000 --> 0:03:14.580000
 So the reality is you as a network
 engineer, you've been building and

0:03:14.580000 --> 0:03:17.080000
 maintaining underlay
 networks for years.

0:03:17.080000 --> 0:03:19.840000
 You probably just didn't call it that
 but that's what an underlay network

0:03:19.840000 --> 0:03:25.720000
 is. Now what's this thing
 called an overlay network?

0:03:25.720000 --> 0:03:31.320000
 An overlay network is a virtual network
 that you're creating on top of

0:03:31.320000 --> 0:03:35.420000
 or over your underlay network.

0:03:35.420000 --> 0:03:40.760000
 So your underlay network is providing your
 physical connectivity, is providing

0:03:40.760000 --> 0:03:43.420000
 the basic IP reachability.

0:03:43.420000 --> 0:03:47.500000
 An overlay network is on top of that
 and is practically implemented via

0:03:47.500000 --> 0:03:52.140000
 VRFs, MPLS VPNs, VXLAN
 or something else.

0:03:52.140000 --> 0:03:54.840000
 So let's just draw something here real
 quick just to demonstrate this

0:03:54.840000 --> 0:03:57.500000
 and hopefully bring this home.

0:03:57.500000 --> 0:04:02.940000
 Let's say that this is
 our network right here.

0:04:02.940000 --> 0:04:08.980000
 So maybe these boxes are
 layer three switches.

0:04:08.980000 --> 0:04:20.580000
 And here we have some routers and here
 we have a couple of other layer

0:04:20.580000 --> 0:04:29.280000
 three switches. So what I'm drawing
 right now is my underlay network.

0:04:29.280000 --> 0:04:34.440000
 And typically in an underlay network we
 want as much redundancy as possible.

0:04:34.440000 --> 0:04:37.680000
 Multiple redundant links
 between devices.

0:04:37.680000 --> 0:04:44.520000
 So something like this and
 we'll say R1, R2 and R3.

0:04:44.520000 --> 0:04:52.940000
 And let's just put one more layer
 three switch right here.

0:04:52.940000 --> 0:04:57.580000
 Okay. Now as part of building out the
 underlay network we said that every

0:04:57.580000 --> 0:05:03.060000
 single link is a point
-to-point routed link.

0:05:03.060000 --> 0:05:08.240000
 So this is going to have most
 likely a slash 30 subnet.

0:05:08.240000 --> 0:05:11.260000
 This right here is going to
 have a slash 30 subnet.

0:05:11.260000 --> 0:05:23.500000
 Every single link is going to be
 underlay network is finished.

0:05:23.500000 --> 0:05:28.460000
 Now I've got full IP reachability from
 every point to every other point

0:05:28.460000 --> 0:05:32.280000
 on this network.

0:05:32.280000 --> 0:05:34.820000
 Okay. Now what is the
 overlay network?

0:05:34.820000 --> 0:05:41.100000
 All right. So let's say that
 we have this design.

0:05:41.100000 --> 0:05:51.240000
 Okay. Let's say that in the future
 I'm just going to put this here as

0:05:51.240000 --> 0:05:54.060000
 a payroll server one.

0:05:54.060000 --> 0:05:57.400000
 And this over here might
 be payroll server two.

0:05:57.400000 --> 0:06:00.780000
 I'm sort of anticipating that those payroll
 servers are going to be right

0:06:00.780000 --> 0:06:06.920000
 there. And what I want to have happen
 is the moment those payroll servers

0:06:06.920000 --> 0:06:12.920000
 actually exist and connect they
 are going to be authenticated.

0:06:12.920000 --> 0:06:15.960000
 So they're going to have to prove themselves
 to the network by offering

0:06:15.960000 --> 0:06:19.280000
 digital certificates or credentials
 or something.

0:06:19.280000 --> 0:06:22.680000
 So we know who and what they are that
 their payroll servers as in this

0:06:22.680000 --> 0:06:27.440000
 example. And once we know their payroll
 servers the network are SDN controller

0:06:27.440000 --> 0:06:32.860000
 I should say will already have a policy
 in place that says hey whenever

0:06:32.860000 --> 0:06:38.020000
 payroll talks to each other
 they should take this path.

0:06:38.020000 --> 0:06:39.540000
 They should go across here.

0:06:39.540000 --> 0:06:44.000000
 They should go across
 here, here, and here.

0:06:44.000000 --> 0:06:48.620000
 So payroll should not be allowed to
 go on any other links beside that.

0:06:48.620000 --> 0:06:53.980000
 Furthermore if we have another department
 that comes online at some point.

0:06:53.980000 --> 0:07:00.700000
 Maybe HR. So here's HR right here.


0:07:00.700000 --> 0:07:04.960000
 Maybe HR right here.

0:07:04.960000 --> 0:07:09.060000
 Well if they ever connect once they're
 authenticated and we verify who

0:07:09.060000 --> 0:07:19.760000
 they are. Well their traffic
 should take this path.

0:07:19.760000 --> 0:07:24.060000
 So what we're building out here is
 what's called the overlay network.

0:07:24.060000 --> 0:07:28.840000
 In other words once payroll connects the
 software defined networking controller

0:07:28.840000 --> 0:07:34.760000
 is going to send down some policies to
 these devices to this switch right

0:07:34.760000 --> 0:07:40.500000
 here, this switch right here, this guy,
 this guy, and this guy ensuring

0:07:40.500000 --> 0:07:44.840000
 that those links exist, ensuring those
 paths exist, and also putting in

0:07:44.840000 --> 0:07:46.860000
 maybe some access lists.

0:07:46.860000 --> 0:07:50.220000
 Saying okay when P1 and P2 talk to
 each other these are the only links

0:07:50.220000 --> 0:07:51.420000
 they're allowed to take.

0:07:51.420000 --> 0:07:54.480000
 They're not allowed to
 take any other path.

0:07:54.480000 --> 0:08:04.440000
 So we have just built a path and the
 same thing is going to happen when

0:08:04.440000 --> 0:08:09.020000
 HR connects. When HR connects we're
 going to build path that's going to

0:08:09.020000 --> 0:08:11.980000
 be enforced via access lists.

0:08:11.980000 --> 0:08:15.480000
 Maybe certain quality of service will
 be implemented on that path to ensure

0:08:15.480000 --> 0:08:21.300000
 certain bandwidth and delay
 requirements are met.

0:08:21.300000 --> 0:08:26.080000
 And now any packets from HR will
 have to go across this path.

0:08:26.080000 --> 0:08:28.480000
 This is considered
 overlay networks.

0:08:28.480000 --> 0:08:30.780000
 We now have two overlay networks.

0:08:30.780000 --> 0:08:34.180000
 We have the green overlay network for
 payroll and we have the red overlay

0:08:34.180000 --> 0:08:37.080000
 network for human resources.

0:08:37.080000 --> 0:08:42.080000
 Now those overlay networks couldn't even
 happen unless the underlay network

0:08:42.080000 --> 0:08:43.680000
 was there to support it.

0:08:43.680000 --> 0:08:47.220000
 If those routers and switches weren't
 already there, if they weren't already

0:08:47.220000 --> 0:08:50.460000
 cabled up, if they didn't already have
 routing protocols to give end to

0:08:50.460000 --> 0:08:54.400000
 end IP reachability, these red and
 green paths would be meaningless.

0:08:54.400000 --> 0:08:56.660000
 There wouldn't be anything
 to carry them.

0:08:56.660000 --> 0:09:03.260000
 So the overlay network is a dynamic
 path that we program on top of our

0:09:03.260000 --> 0:09:08.420000
 supporting structure which
 is the underlay network.

0:09:08.420000 --> 0:09:10.880000
 And these overlay networks
 can grow and shrink.

0:09:10.880000 --> 0:09:16.400000
 They can change just depending on what
 our policy is at any given time.

0:09:16.400000 --> 0:09:27.080000
 And this is the primary job of the now
 there's one other term I want you

0:09:27.080000 --> 0:09:29.560000
 to be familiar with.

0:09:29.560000 --> 0:09:33.880000
 So we just talked about underlay and
 overlay which is something called

0:09:33.880000 --> 0:09:39.820000
 the SDN fabric. Now this is just a fancy
 term for the physical infrastructure

0:09:39.820000 --> 0:09:43.740000
 that's used to build
 the underlay network.

0:09:43.740000 --> 0:09:47.620000
 So the fabric is composed of the actual
 switches, routers, cables, and

0:09:47.620000 --> 0:09:54.840000
 internal switching paths used
 to build the underlay network.

0:09:54.840000 --> 0:09:58.800000
 It's typically used to describe only
 those devices in the underlay network

0:09:58.800000 --> 0:10:04.060000
 that can be programmed or controlled
 by the SDN controller.

0:10:04.060000 --> 0:10:07.840000
 Now if you've watched any other videos,
 you've gotten a taste of the fact

0:10:07.840000 --> 0:10:12.600000
 that when the SDN controller is speaking
 to my networking devices and

0:10:12.600000 --> 0:10:18.040000
 controlling them, there's a variety
 of languages used to support that.

0:10:18.040000 --> 0:10:22.820000
 It might be something like simply in
 a CLI session over SSH which pretty

0:10:22.820000 --> 0:10:28.540000
 much any device supports or might be
 more programmatic like net conf yang

0:10:28.540000 --> 0:10:33.540000
 or rest conf yang which only certain
 router switches support.

0:10:33.540000 --> 0:10:39.000000
 So whatever the mechanism, whatever
 the language or protocol that the

0:10:39.000000 --> 0:10:45.480000
 SDN controller and our automation tools
 are using to build those overlay

0:10:45.480000 --> 0:10:50.560000
 networks, well our underlay devices
 have to support that.

0:10:50.560000 --> 0:10:52.800000
 Now that's not always possible.

0:10:52.800000 --> 0:10:58.140000
 Sometimes we have an existing network
 that's already in place that when

0:10:58.140000 --> 0:11:01.220000
 we get our controller in there, if our
 controller is speaking for example,

0:11:01.220000 --> 0:11:06.420000
 net conf yang as a native language, maybe
 our existing devices don't support

0:11:06.420000 --> 0:11:09.960000
 that. And so instead what we decide to
 do is well we're going to buy some

0:11:09.960000 --> 0:11:15.700000
 new devices at the edge that do support
 net conf yang that natively speak

0:11:15.700000 --> 0:11:19.240000
 that language but our sort of interior
 of our existing network doesn't

0:11:19.240000 --> 0:11:23.520000
 support that. Well a lot of times this
 term the fabric will be relating

0:11:23.520000 --> 0:11:28.500000
 to those new devices we just bought
 that speak the same language as the

0:11:28.500000 --> 0:11:32.400000
 controller. So the controller doesn't
 really have any direct influence

0:11:32.400000 --> 0:11:37.120000
 or impact over those middle devices
 necessarily but it does on the outer

0:11:37.120000 --> 0:11:42.760000
 devices. Now in an ideal world every single
 device you have, every firewall,

0:11:42.760000 --> 0:11:47.520000
 every router, every switch would support
 the exact same language as the

0:11:47.520000 --> 0:11:51.920000
 controller. So the controller could
 reach out and command and control

0:11:51.920000 --> 0:11:53.800000
 all of those devices.

0:11:53.800000 --> 0:11:58.000000
 That's not always possible but
 that is an ideal situation.

0:11:58.000000 --> 0:12:05.340000
 So a lot of times in documents and papers
 when they have a little device

0:12:05.340000 --> 0:12:10.680000
 can be controlled by
 your SDN controller.

0:12:10.680000 --> 0:12:14.980000
 And the last thing I want to mention
 here is that in an ideal world your

0:12:14.980000 --> 0:12:17.480000
 fabric would be fully meshed.

0:12:17.480000 --> 0:12:20.580000
 There'd be multiple connections to
 get everywhere, there'd be multiple

0:12:20.580000 --> 0:12:24.560000
 equal cost paths to get between destinations
 so you could do, you could

0:12:24.560000 --> 0:12:26.220000
 take advantage of load balancing.

0:12:26.220000 --> 0:12:32.400000
 That's not always practically possible
 but this term the fabric a lot

0:12:32.400000 --> 0:12:34.040000
 of times assumes that.

0:12:34.040000 --> 0:12:38.020000
 So the term, an SDN fabric a lot of times
 assumes that you've got a network

0:12:38.020000 --> 0:12:42.500000
 where every single device supports whatever
 the language is that the controller

0:12:42.500000 --> 0:12:48.300000
 is using to control them and that we have
 multiple equal cost paths between

0:12:48.300000 --> 0:12:52.040000
 destinations. Even if that's not necessarily
 true that's a lot of the

0:12:52.040000 --> 0:12:54.660000
 assumption that goes behind this.

0:12:54.660000 --> 0:12:58.020000
 So that brings us to the
 end of this video.

0:12:58.020000 --> 0:12:58.820000
 Thank you for watching.

