WEBVTT

0:00:03.280000 --> 0:00:05.440000
 Hello and welcome to this video.

0:00:05.440000 --> 0:00:09.480000
 In this video, I'm going to talk about software
-defined networking's relationship

0:00:09.480000 --> 0:00:14.740000
 to the management, control,
 and data planes.

0:00:14.740000 --> 0:00:17.940000
 So we're going to start by defining what
 is the management plane, in case

0:00:17.940000 --> 0:00:22.800000
 you've never heard of that before, and
 how does SDN affect it, if at all?

0:00:22.800000 --> 0:00:26.220000
 We're going to define what the data plane
 is, how the data plane is typically

0:00:26.220000 --> 0:00:30.920000
 populated. We're going to talk about
 the control plane, and how SDN can

0:00:30.920000 --> 0:00:34.640000
 affect both the control
 and the data plane.

0:00:34.640000 --> 0:00:38.160000
 So let's start by talking about
 the management plane.

0:00:38.160000 --> 0:00:43.360000
 So all the things, actually before
 we talk about the management plane,

0:00:43.360000 --> 0:00:45.060000
 let's do a high-level
 overview here.

0:00:45.060000 --> 0:00:48.600000
 All the things that a networking device
 can do, and when I say a networking

0:00:48.600000 --> 0:00:52.520000
 device, I mean a router, a switch,
 a firewall, anything that a packet

0:00:52.520000 --> 0:00:54.880000
 or a frame can flow through.

0:00:54.880000 --> 0:00:58.820000
 All the things that a networking device
 can do can fall into one of three

0:00:58.820000 --> 0:01:05.020000
 logical places. The management plane,
 the control plane, or the data plane.

0:01:05.020000 --> 0:01:07.660000
 So let's talk about all three of these
 things, and see how they relate

0:01:07.660000 --> 0:01:10.140000
 to software-defined networking.

0:01:10.140000 --> 0:01:12.760000
 So let's start with the
 management plane.

0:01:12.760000 --> 0:01:16.900000
 So as it says, the management plane
 is responsible for giving you, the

0:01:16.900000 --> 0:01:20.860000
 network administrator, access
 to control the device.

0:01:20.860000 --> 0:01:25.240000
 So when you are configuring the device,
 whatever is giving you the ability

0:01:25.240000 --> 0:01:30.620000
 to configure it, that feature or protocol
 or language resides in the management

0:01:30.620000 --> 0:01:35.100000
 plane. When you're monitoring it, once
 again, when you're troubleshooting

0:01:35.100000 --> 0:01:39.020000
 it, whatever is giving you the ability
 to do that resides in the management

0:01:39.020000 --> 0:01:44.320000
 plane. So any feature or protocol that
 gives you the ability to configure,

0:01:44.320000 --> 0:01:48.700000
 monitor, or troubleshoot the device
 resides in the management plane.

0:01:48.700000 --> 0:01:51.440000
 So you are managing the box.

0:01:51.440000 --> 0:01:58.560000
 So whether you're doing console access,
 telnet, SSH, possibly accessing

0:01:58.560000 --> 0:02:05.220000
 it via HTTP or HTTPS, or SMMP, these
 are very common protocols used to

0:02:05.220000 --> 0:02:10.320000
 manage your boxes, your networking devices,
 and all these protocols reside

0:02:10.320000 --> 0:02:15.500000
 in the management plane.

0:02:15.500000 --> 0:02:20.980000
 Now, if we implement software-defined
 networking into our mix here, does

0:02:20.980000 --> 0:02:23.440000
 that affect how we
 manage our boxes?

0:02:23.440000 --> 0:02:25.320000
 Does it affect the
 management plane?

0:02:25.320000 --> 0:02:33.060000
 Well, it can. Many SDN controllers rely
 on existing management plane mechanisms.

0:02:33.060000 --> 0:02:38.180000
 For example, before an SDN controller
 is introduced into your network,

0:02:38.180000 --> 0:02:47.020000
 your management plane or an SSH or
 a secure CRT session and telnetting

0:02:47.020000 --> 0:02:52.720000
 or SSHing into an individual router switch
 and then issuing your CLI commands

0:02:52.720000 --> 0:02:55.140000
 over that SSH session.

0:02:55.140000 --> 0:02:59.960000
 Now when we implement an SDN controller
 or a network automation tool into

0:02:59.960000 --> 0:03:06.640000
 the mix, that same SSH session can
 be initiated to the individual box,

0:03:06.640000 --> 0:03:10.680000
 but no longer be initiated by you, it's
 being initiated by the controller.

0:03:10.680000 --> 0:03:14.100000
 So that controller can automatically,
 maybe you're not even sitting at

0:03:14.100000 --> 0:03:20.320000
 the desk, maybe you have a policy that
 says at 5 o'clock PM today, initiate

0:03:20.320000 --> 0:03:25.260000
 an SSH session into every single device
 and issue a show command, like

0:03:25.260000 --> 0:03:30.160000
 show processes CPU, then collect all
 the results, display it graphically

0:03:30.160000 --> 0:03:31.360000
 in my controller.

0:03:31.360000 --> 0:03:36.380000
 So that controller hasn't really affected
 the management plane, hasn't

0:03:36.380000 --> 0:03:40.680000
 changed the management plane, it's just
 using the existing protocols that

0:03:40.680000 --> 0:03:45.020000
 were already there that you were manually
 using before, it's using those

0:03:45.020000 --> 0:03:48.000000
 same protocols to reach into
 your boxes and manage them.

0:03:48.000000 --> 0:03:52.900000
 Now there are some new mechanisms that
 have been developed for controllers

0:03:52.900000 --> 0:03:59.260000
 to get into the management plane, such
 as NetConf Yang and XML based commands.

0:03:59.260000 --> 0:04:04.720000
 Not a lot of devices support this yet,
 but this is another way that software

0:04:04.720000 --> 0:04:09.060000
 defined networks that controllers can
 reach into your box and manage them,

0:04:09.060000 --> 0:04:14.440000
 instead of just issuing a standard command
 line driven process over SSH.

0:04:14.440000 --> 0:04:18.300000
 So the moral of the story here is the
 management plane hasn't changed

0:04:18.300000 --> 0:04:21.560000
 a whole lot from the
 perspective of SDN.

0:04:21.560000 --> 0:04:25.440000
 You still have to go into your router
 switch, you still have to program

0:04:25.440000 --> 0:04:31.100000
 it with a username and a password, maybe
 give it some basic SSH credentials,

0:04:31.100000 --> 0:04:35.440000
 some basic SNMP credentials, and then
 the controller can access the box

0:04:35.440000 --> 0:04:40.180000
 using those credentials much
 like you would do manually.

0:04:40.180000 --> 0:04:44.740000
 Now let's move over to the data
 plane for a second, what is it?

0:04:44.740000 --> 0:04:49.780000
 The data plane is responsible for transporting
 data through a network,

0:04:49.780000 --> 0:04:55.780000
 this is also frequently called the forwarding
 plane, so any feature, protocol,

0:04:55.780000 --> 0:05:02.520000
 entity that exists to give a device this
 ability resides in the data plane.

0:05:02.520000 --> 0:05:08.040000
 So anything that a packet would touch,
 as a packet is flowing through

0:05:08.040000 --> 0:05:13.840000
 your firewall, your router, your switch,
 if a packet touches it, if that

0:05:13.840000 --> 0:05:19.640000
 packet is looking at something to decide
 where should I go, what should

0:05:19.640000 --> 0:05:26.240000
 happen to me while I'm in this box, that
 device, that thing resides within

0:05:26.240000 --> 0:05:31.840000
 the data plane. So we're talking about
 Mac address tables, routing tables,

0:05:31.840000 --> 0:05:34.640000
 or in the case of Cef forwarding tables
 like your forwarding information

0:05:34.640000 --> 0:05:42.320000
 base, your cables, your Nix, your
 packet buffers and queues.

0:05:42.320000 --> 0:05:46.600000
 So all of these things have to exist,
 in the case of your Mac address

0:05:46.600000 --> 0:05:49.740000
 tables and your routing tables, they
 all have to be programmed with some

0:05:49.740000 --> 0:05:53.480000
 sort of forwarding information so they
 actually know what to do with the

0:05:53.480000 --> 0:05:55.980000
 packet once a packet comes in.

0:05:55.980000 --> 0:05:59.280000
 Now how does that information
 get in there?

0:05:59.280000 --> 0:06:02.500000
 Because after all, typically those
 memory structures, your Mac address

0:06:02.500000 --> 0:06:06.180000
 table, your routing table, start out
 as empty, they don't have anything

0:06:06.180000 --> 0:06:10.500000
 in there. Mac addresses tables don't have
 any Mac addresses, routing tables

0:06:10.500000 --> 0:06:11.900000
 don't have any routes.

0:06:11.900000 --> 0:06:16.600000
 Well, the data plane is not responsible
 for learning stuff.

0:06:16.600000 --> 0:06:17.920000
 It's not responsible for that.

0:06:17.920000 --> 0:06:21.740000
 That is the job of our third plane
 called the control plane.

0:06:21.740000 --> 0:06:26.860000
 Whatever the control plane learns that
 is relevant to packet handling

0:06:26.860000 --> 0:06:32.340000
 is downloaded into the data plane.


0:06:32.340000 --> 0:06:35.880000
 Now sometimes when you look at a data
 plane structure, for example, let's

0:06:35.880000 --> 0:06:38.340000
 take a routing table on
 a router as an example.

0:06:38.340000 --> 0:06:42.700000
 A routing table is an example of something
 that resides in the data plane.

0:06:42.700000 --> 0:06:47.140000
 When a packet comes in, it's going to
 be compared against entries in the

0:06:47.140000 --> 0:06:50.520000
 routing table. And if there's a matching
 entry, we know what to do with

0:06:50.520000 --> 0:06:54.520000
 the packet. If there's not a matching
 entry, we drop the packet so that

0:06:54.520000 --> 0:06:57.780000
 routing table is affecting
 the data.

0:06:57.780000 --> 0:06:59.700000
 So it resides in the data plane.

0:06:59.700000 --> 0:07:02.480000
 But if you look at a routing table,
 if you do, for example, a show IP

0:07:02.480000 --> 0:07:07.400000
 route command on a Cisco router and look
 at a single line, a single entry,

0:07:07.400000 --> 0:07:12.200000
 you're actually going to see some stuff
 in there that was populated because

0:07:12.200000 --> 0:07:17.120000
 of the control plane, but isn't
 really part of the data plane.

0:07:17.120000 --> 0:07:20.500000
 For example, if I look at a routing entry,
 one of the first things you'll

0:07:20.500000 --> 0:07:25.420000
 show me is a little character that will
 indicate how that route was learned.

0:07:25.420000 --> 0:07:26.820000
 Was it learned via RIP?

0:07:26.820000 --> 0:07:29.520000
 Was it learned via EIGRP or BGP?

0:07:29.520000 --> 0:07:32.180000
 Now, how a route was learned?

0:07:32.180000 --> 0:07:35.360000
 Does that really impact the
 usefulness of that route?

0:07:35.360000 --> 0:07:38.560000
 When a packet comes in,
 does that packet care?

0:07:38.560000 --> 0:07:41.260000
 If the route was learned
 via OSPF or RIP?

0:07:41.260000 --> 0:07:45.620000
 No. All it cares that there's a line and
 there an entry it can match against.

0:07:45.620000 --> 0:07:49.540000
 So although that indicator of how the
 route was learned is part of the

0:07:49.540000 --> 0:07:52.840000
 routing table, it's not really
 part of the data plane.

0:07:52.840000 --> 0:07:56.400000
 It doesn't directly impact
 the forwarding of a packet.

0:07:56.400000 --> 0:07:59.720000
 If we continue along that routing entry,
 there's other stuff in there.

0:07:59.720000 --> 0:08:03.580000
 For example, the administrative distance,
 the routing metric, the RIP

0:08:03.580000 --> 0:08:07.780000
 hop calendar, the OSPF cost,
 that's part of that route.

0:08:07.780000 --> 0:08:12.060000
 But once again, that is not directly
 relevant to a packet that's looking

0:08:12.060000 --> 0:08:18.200000
 at that entry. All the packet cares
 about is the subnet, the mask, the

0:08:18.200000 --> 0:08:20.800000
 egress interface, and the IP
 address to the next hop.

0:08:20.800000 --> 0:08:26.300000
 That's it. Any other piece of information
 that were added onto that route,

0:08:26.300000 --> 0:08:30.180000
 the control plane put those piece of information
 there, but it's not directly

0:08:30.180000 --> 0:08:32.440000
 relevant to the data plane.

0:08:32.440000 --> 0:08:35.800000
 So let's talk about
 the control plane.

0:08:35.800000 --> 0:08:40.720000
 The control plane is responsible for
 teaching the device the information

0:08:40.720000 --> 0:08:44.920000
 it needs to put into
 the data plane.

0:08:44.920000 --> 0:08:49.200000
 So any feature protocol that exists
 to provide this ability resides in

0:08:49.200000 --> 0:08:49.920000
 the control plane.

0:08:49.920000 --> 0:08:54.320000
 This would be your dynamic routing protocols
 or maybe your static routes.

0:08:54.320000 --> 0:08:58.800000
 It'd be the process of dynamic
 MAC address learning.

0:08:58.800000 --> 0:09:00.540000
 Now a lot of people just
 stop right there.

0:09:00.540000 --> 0:09:03.960000
 They think, okay, well my routing
 protocol will put my routes in.

0:09:03.960000 --> 0:09:08.460000
 My dynamic MAC learning and however that
 logic works will put MAC addresses

0:09:08.460000 --> 0:09:10.100000
 in the MAC address table.

0:09:10.100000 --> 0:09:11.200000
 The tables are built.

0:09:11.200000 --> 0:09:13.340000
 That's it. I'm done.

0:09:13.340000 --> 0:09:14.840000
 Not necessarily.

0:09:14.840000 --> 0:09:21.940000
 Because anything that impacts packet
 handling, not just where do I forward

0:09:21.940000 --> 0:09:25.780000
 the packet, but for example,
 should I forward the packet?

0:09:25.780000 --> 0:09:29.860000
 There might be some data structure that
 when a packet comes in, that data

0:09:29.860000 --> 0:09:34.580000
 structure says, oh, I match that
 packet, kill it, drop the packet.

0:09:34.580000 --> 0:09:36.960000
 That definitely impacted
 the data, right?

0:09:36.960000 --> 0:09:38.320000
 It wasn't allowed to be forward.

0:09:38.320000 --> 0:09:43.460000
 It was killed. Well, some control plane
 logic, for example, interactions

0:09:43.460000 --> 0:09:47.940000
 between a network device and a AAA server,
 those interactions would learn

0:09:47.940000 --> 0:09:52.820000
 stuff, learn security policies that
 would directly impact the forwarding

0:09:52.820000 --> 0:09:54.380000
 plane or the data plane.

0:09:54.380000 --> 0:09:58.160000
 In this case saying, kill
 the packet, drop it.

0:09:58.160000 --> 0:10:02.180000
 Even DHCP transactions, right?

0:10:02.180000 --> 0:10:04.900000
 In your laptop, okay?

0:10:04.900000 --> 0:10:10.180000
 One thing that resides in the data plane
 is your ARP table, your ARP table,

0:10:10.180000 --> 0:10:13.900000
 and your laptop has to have an
 IP address to do anything.

0:10:13.900000 --> 0:10:18.760000
 Well, the process of learning
 an IP address is DHCP.

0:10:18.760000 --> 0:10:21.940000
 That logic is part of
 the control plane.

0:10:21.940000 --> 0:10:26.800000
 Now how does SDN affect or
 impact all of this stuff?

0:10:26.800000 --> 0:10:32.360000
 Well, sometimes SDN gives you the ability
 to separate the control plane

0:10:32.360000 --> 0:10:33.480000
 from the data plane.

0:10:33.480000 --> 0:10:38.060000
 This is what we call the imperative approach
 of SDN, which is also called

0:10:38.060000 --> 0:10:43.160000
 stateful SDN, where all functions of
 the control plane, your ability to

0:10:43.160000 --> 0:10:47.220000
 learn routing information, your ability
 to learn what MAC addresses exist

0:10:47.220000 --> 0:10:47.980000
 in your network.

0:10:47.980000 --> 0:10:52.720000
 Instead of having that logic reside
 on the individual networking boxes

0:10:52.720000 --> 0:10:57.620000
 like your routers and switches, that
 logic is removed from the box and

0:10:57.620000 --> 0:10:59.700000
 put into a controller instead.

0:10:59.700000 --> 0:11:04.080000
 The controller is responsible for learning
 where everything is and how

0:11:04.080000 --> 0:11:05.540000
 to get to things.

0:11:05.540000 --> 0:11:09.300000
 That would be called stateful SDN
 or the imperative approach.

0:11:09.300000 --> 0:11:13.320000
 So in this case, the controller can
 directly program the data plane of

0:11:13.320000 --> 0:11:18.740000
 devices. Now, this sounds kind of neat
 and interesting in theory, and

0:11:18.740000 --> 0:11:22.720000
 there are some controllers out there
 that do this, but there are some

0:11:22.720000 --> 0:11:25.180000
 definite risks to this approach.

0:11:25.180000 --> 0:11:28.340000
 For example, what if I have a whole
 bunch of routers and switches that

0:11:28.340000 --> 0:11:31.680000
 I bought, and these devices
 have no brain, right?

0:11:31.680000 --> 0:11:33.600000
 They don't have the
 ability to run OSPF.

0:11:33.600000 --> 0:11:36.160000
 They don't have the ability
 to do dynamic MAC learning.

0:11:36.160000 --> 0:11:38.000000
 They just have empty tables.

0:11:38.000000 --> 0:11:41.160000
 All they can do is talk back to the
 controller, and the controller has

0:11:41.160000 --> 0:11:41.920000
 to tell them everything.

0:11:41.920000 --> 0:11:45.140000
 They have to populate their tables
 and tell them what to do.

0:11:45.140000 --> 0:11:49.760000
 Well, what happens if that controller
 becomes unreachable, or the controller

0:11:49.760000 --> 0:11:54.420000
 goes down? Now we have a big problem
 on our hands, and that is one thing

0:11:54.420000 --> 0:11:57.360000
 that could happen with the imperative
 approach because the controller

0:11:57.360000 --> 0:11:59.640000
 is the brains of everything.

0:11:59.640000 --> 0:12:04.480000
 Or what if the controller exists, but
 there's a bottleneck getting to

0:12:04.480000 --> 0:12:08.660000
 him? We're sending him so many new packets
 saying, hey, what do I do with

0:12:08.660000 --> 0:12:09.960000
 this one? What do I
 do with this one?

0:12:09.960000 --> 0:12:11.100000
 What do I do with this one?

0:12:11.100000 --> 0:12:14.260000
 That there's a bottleneck getting that
 information to the controller and

0:12:14.260000 --> 0:12:17.120000
 the controller getting information
 back out to the device and saying,

0:12:17.120000 --> 0:12:19.560000
 hey, program your table this way.

0:12:19.560000 --> 0:12:24.040000
 So that's another definite risk of
 having the imperative approach.

0:12:24.040000 --> 0:12:27.920000
 Then we also have the declarative approach,
 which is called stateless

0:12:27.920000 --> 0:12:32.980000
 SDN. Where the control plane and the
 data planes, they still reside in

0:12:32.980000 --> 0:12:35.100000
 the individual routers
 and switches.

0:12:35.100000 --> 0:12:40.760000
 The controller, however, declares how
 it wishes the network to function.

0:12:40.760000 --> 0:12:44.840000
 The controller says, hey, if you ever
 see this type of traffic, implement

0:12:44.840000 --> 0:12:46.600000
 this security policy.

0:12:46.600000 --> 0:12:49.120000
 Implement this quality
 of service policy.

0:12:49.120000 --> 0:12:53.740000
 Implement this route if you happen to
 see this packet with this criteria.

0:12:53.740000 --> 0:12:58.640000
 So on the declarative approach, the
 networking devices would translate

0:12:58.640000 --> 0:13:03.680000
 the requests of the controller what the
 controller would like to see into

0:13:03.680000 --> 0:13:04.460000
 their own community.

0:13:04.460000 --> 0:13:07.800000
 And then it would allow the commands
 to program their own forwarding path

0:13:07.800000 --> 0:13:14.260000
 or data path. So that
 concludes this video.

0:13:14.260000 --> 0:13:15.900000
 And I hope you found it useful.

0:13:15.900000 --> 0:13:16.620000
 Thank you for watching.

