WEBVTT

0:00:03.180000 --> 0:00:07.360000
 Hello and welcome this video titled SDN's
 relationship to management control

0:00:07.360000 --> 0:00:09.740000
 and data planes.

0:00:09.740000 --> 0:00:14.260000
 So we first need to define
 what these planes are.

0:00:14.260000 --> 0:00:19.660000
 So everything that a networking device
 can do, a router, a switch, a firewall

0:00:19.660000 --> 0:00:24.440000
 can do, can be categorized as residing
 in one of these planes, the management

0:00:24.440000 --> 0:00:27.040000
 plane, the control plane
 and the data plane.

0:00:27.040000 --> 0:00:29.460000
 Sometimes the data plane is also
 called the forwarding plane.

0:00:29.460000 --> 0:00:30.860000
 They're one and the same.

0:00:30.860000 --> 0:00:35.160000
 So let's talk about what these planes
 are and specifically how they relate

0:00:35.160000 --> 0:00:38.660000
 to our controllers and our
 software divine networking.

0:00:38.660000 --> 0:00:40.980000
 So number one, the
 management plane.

0:00:40.980000 --> 0:00:45.520000
 So the management plane is responsible for
 giving you the network administrator

0:00:45.520000 --> 0:00:53.220000
 access to control and configure, monitor
 and troubleshoot the device.

0:00:53.220000 --> 0:00:57.940000
 So for example, as it says here, any
 feature or protocol that exists to

0:00:57.940000 --> 0:01:03.300000
 give you the human being, this ability
 resides in the management plane.

0:01:03.300000 --> 0:01:09.480000
 So when you are accessing the command
 line, you are accessing the management

0:01:09.480000 --> 0:01:16.400000
 plane. When you are accessing the web
 based GUI of a lightweight access

0:01:16.400000 --> 0:01:19.480000
 point controller, you are accessing
 the management plane.

0:01:19.480000 --> 0:01:25.240000
 When you are issuing telnet or SSH commands
 to remotely get into something,

0:01:25.240000 --> 0:01:27.460000
 that is the management plane.

0:01:27.460000 --> 0:01:32.600000
 When you are doing SNMP, that is the
 management plane, especially if you

0:01:32.600000 --> 0:01:38.080000
 are using SNMP to issue right operations
 to configure variables within

0:01:38.080000 --> 0:01:43.960000
 the MIB. So what relationship does software
 defined networking have with

0:01:43.960000 --> 0:01:47.620000
 this? Because typically, the management
 plane has involved one or two

0:01:47.620000 --> 0:01:51.640000
 things, either you connecting to the
 console port and configuring away,

0:01:51.640000 --> 0:01:57.500000
 or you remotely establishing a telnet
 SSH or maybe HTTP session and once

0:01:57.500000 --> 0:01:59.500000
 again, configuring away.

0:01:59.500000 --> 0:02:01.820000
 So how does SDN relate to this?

0:02:01.820000 --> 0:02:08.400000
 Well most SDN controllers rely on existing
 management plane mechanisms

0:02:08.400000 --> 0:02:15.260000
 such as what you are already
 familiar with, SSH, HTTPS.

0:02:15.260000 --> 0:02:20.200000
 Some new mechanism has been developed
 for new types of access.

0:02:20.200000 --> 0:02:25.540000
 You might hear of NetConf Yang
 or XML based commands.

0:02:25.540000 --> 0:02:36.160000
 So the idea behind this is, you know,
 if you are well versed in the Cisco

0:02:36.160000 --> 0:02:40.620000
 iOS command line and you know what the
 commands are to implement routing,

0:02:40.620000 --> 0:02:45.180000
 implement switching, implement
 spanning tree, that is great.

0:02:45.180000 --> 0:02:53.940000
 So imagine for a second that we built
 an SDN controller and let's say

0:02:53.940000 --> 0:02:59.540000
 it was all GUI based and so I went into
 a drop, so I am on the controller,

0:02:59.540000 --> 0:03:04.040000
 I went into a drop down box in the controller
 and that drop down box said

0:03:04.040000 --> 0:03:10.100000
 configure root bridge, drop down box,
 select switch, okay I see switch

0:03:10.100000 --> 0:03:13.680000
 one, switch two, okay I want switch
 three to be the root bridge and the

0:03:13.680000 --> 0:03:18.080000
 moment I selected switch three in the
 background, what that controller

0:03:18.080000 --> 0:03:22.480000
 did was it established an SSH session
 just like you would do in secure

0:03:22.480000 --> 0:03:27.980000
 CRT or hyper terminal and SSH session
 to that switch and then it sent

0:03:27.980000 --> 0:03:32.240000
 to that switch over that SSH session
 command you are used to spanning

0:03:32.240000 --> 0:03:36.140000
 dash tree, BLAN 2 root,
 something like that.

0:03:36.140000 --> 0:03:39.880000
 Well that would be great but what if
 you also want to integrate that same

0:03:39.880000 --> 0:03:45.200000
 SDN controller with a Juniper network,
 a Juniper switch or an ERISTA switch

0:03:45.200000 --> 0:03:48.040000
 or a HOA switch or something
 like that.

0:03:48.040000 --> 0:03:52.780000
 Well now you are asking the developer
 of that SDN controller to integrate

0:03:52.780000 --> 0:03:58.540000
 Juniper commands, ERISTA commands, HP
 commands, that is a tall order to

0:03:58.540000 --> 0:04:03.840000
 do all that. So instead some very smart
 people said you know what, we

0:04:03.840000 --> 0:04:08.160000
 should come up with some common language
 that can be used at the management

0:04:08.160000 --> 0:04:13.780000
 plane and this common language if vendors
 will support it, controllers

0:04:13.780000 --> 0:04:17.840000
 can use this common language to reach
 out and do basically the same things

0:04:17.840000 --> 0:04:22.700000
 that the CLI would do in Juniper or
 the CLI would do in Cisco and that

0:04:22.700000 --> 0:04:25.580000
 is basically what NetConf
 is all about.

0:04:25.580000 --> 0:04:29.800000
 NetConf is a way of let's coming up
 with a standardized way that we can

0:04:29.800000 --> 0:04:36.280000
 send structured defined data, defined
 commands to a device no matter what

0:04:36.280000 --> 0:04:39.440000
 that device is as long
 as it supports NetConf.

0:04:39.440000 --> 0:04:44.200000
 So now if I have a Cisco switch and a
 Juniper switch and an ERISTA switch

0:04:44.200000 --> 0:04:48.960000
 and they all say they are all manufactured
 to support NetConf then that

0:04:48.960000 --> 0:04:53.780000
 means that I can still connect to the
 console and issue CLI commands.

0:04:53.780000 --> 0:04:58.480000
 I can still remotely SSH or tell it
 into it via secure CRT and issue my

0:04:58.480000 --> 0:05:03.300000
 CLI commands but in addition to that
 now something like a controller can

0:05:03.300000 --> 0:05:07.740000
 send commands via a different structure,
 a different language called NetConf

0:05:07.740000 --> 0:05:12.340000
 and accomplish the exact same thing
 I was doing with the CLI and now we

0:05:12.340000 --> 0:05:15.340000
 don't have to worry about the differences
 between Juniper and ERISTA and

0:05:15.340000 --> 0:05:20.060000
 Cisco. That's sort of the idea behind
 NetConf Yang as well as XML.

0:05:20.060000 --> 0:05:24.020000
 All right so that was
 the management plane.

0:05:24.020000 --> 0:05:26.300000
 What is the data now we are skipping
 forward to the data plane otherwise

0:05:26.300000 --> 0:05:29.180000
 known as the forwarding plane.

0:05:29.180000 --> 0:05:36.980000
 So the data plane is those memory structures,
 those tables and databases

0:05:36.980000 --> 0:05:40.140000
 that actually act on data.

0:05:40.140000 --> 0:05:44.440000
 So when data comes into a router switch
 you have to ask yourself what's

0:05:44.440000 --> 0:05:45.940000
 going to look at that data.

0:05:45.940000 --> 0:05:47.920000
 Is it going to be
 the routing table?

0:05:47.920000 --> 0:05:50.100000
 Is it going to be the SEF tables?

0:05:50.100000 --> 0:05:55.960000
 Is it going to be some sort of security
 table that enforces access lists?

0:05:55.960000 --> 0:06:01.800000
 All that stuff that will actually look
 at the packet and make decisions

0:06:01.800000 --> 0:06:07.120000
 on the packet and decide where the packet
 goes that stuff resides in the

0:06:07.120000 --> 0:06:13.940000
 data plane. Your MAC address tables,
 your routing tables, even the cables

0:06:13.940000 --> 0:06:18.360000
 and NIC cards and packet buffers and
 queues, all that stuff touches your

0:06:18.360000 --> 0:06:22.320000
 packets, is used to forward your packets,
 that stuff resides in the data

0:06:22.320000 --> 0:06:31.440000
 plane. So like it says, MAC address tables
 typically have no MAC addresses,

0:06:31.440000 --> 0:06:33.880000
 they have to be learned
 by the switch.

0:06:33.880000 --> 0:06:35.260000
 Routing tables have no routes.

0:06:35.260000 --> 0:06:41.720000
 So the data plane is not responsible
 for the learning of this information.

0:06:41.720000 --> 0:06:47.040000
 So whatever features, whatever protocols
 are working in the data plane,

0:06:47.040000 --> 0:06:49.900000
 for example, Cisco
 Express forwarding.

0:06:49.900000 --> 0:06:55.300000
 If you look at the design documents
 and the technical documents of Cisco

0:06:55.300000 --> 0:07:00.260000
 Express forwarding, nowhere in there
 will it say, how does Cisco Express

0:07:00.260000 --> 0:07:04.340000
 forwarding learn that the route to the
 5050 network is reachable via that

0:07:04.340000 --> 0:07:07.620000
 way? And that the route to the 99 network
 is reachable via two different

0:07:07.620000 --> 0:07:09.620000
 paths, but that's the best one.

0:07:09.620000 --> 0:07:11.060000
 SEF doesn't do that.

0:07:11.060000 --> 0:07:15.020000
 SEF will take that information that
 was learned by something else.

0:07:15.020000 --> 0:07:18.020000
 You can put that information
 into a SEF table.

0:07:18.020000 --> 0:07:21.700000
 And now once it's in the SEF table
 as a packet comes in, Cisco Express

0:07:21.700000 --> 0:07:25.500000
 forwarding can look at that packet compared
 against the table and figure

0:07:25.500000 --> 0:07:26.880000
 out where it needs to go.

0:07:26.880000 --> 0:07:31.780000
 But SEF is not responsible for the
 initial population of the table.

0:07:31.780000 --> 0:07:33.540000
 That is the control plane.

0:07:33.540000 --> 0:07:35.040000
 We talked a little bit
 about this already.

0:07:35.040000 --> 0:07:36.880000
 The control plane does that.

0:07:36.880000 --> 0:07:40.360000
 So I don't think we need
 to go into that.

0:07:40.360000 --> 0:07:43.280000
 We've already discussed
 the control plane.

0:07:43.280000 --> 0:07:48.020000
 So also, let's take
 a step back here.

0:07:48.020000 --> 0:07:50.760000
 I don't think this
 is coming up here.

0:07:50.760000 --> 0:07:51.920000
 We didn't answer the question.

0:07:51.920000 --> 0:07:54.660000
 We answered the question, how does a software
 defined networking controller

0:07:54.660000 --> 0:07:56.480000
 deal with the management plane?

0:07:56.480000 --> 0:08:02.240000
 We said, well, it has the ability just
 like you to issue SSH CLI commands

0:08:02.240000 --> 0:08:06.940000
 or HTTP, but it also has the ability
 to use something called NetConf or

0:08:06.940000 --> 0:08:11.200000
 RESTConf, which is like a derivative
 of NetConf to reach in there.

0:08:11.200000 --> 0:08:14.480000
 So it can manage the device
 just like you can.

0:08:14.480000 --> 0:08:18.140000
 How does an SDN controller affect
 the forwarding plane?

0:08:18.140000 --> 0:08:21.380000
 Well, this goes back to the previous
 video about where we were talking

0:08:21.380000 --> 0:08:25.520000
 about the differences between the imperative
 approach and the declarative

0:08:25.520000 --> 0:08:32.160000
 approach. If I have an SDN environment
 that supports the imperative approach,

0:08:32.160000 --> 0:08:36.300000
 then that means my SDN controller actually
 has the ability to reach in

0:08:36.300000 --> 0:08:41.520000
 to the memory that's holding the routing
 table and put a route right in

0:08:41.520000 --> 0:08:45.400000
 there. It has the ability to reach
 in to the MAC address table and put

0:08:45.400000 --> 0:08:47.240000
 a MAC address right in there.

0:08:47.240000 --> 0:08:54.480000
 But typically speaking, when a packet
 comes into a router switch, does

0:08:54.480000 --> 0:08:56.640000
 the router switch even
 in that approach?

0:08:56.640000 --> 0:09:00.080000
 I just said, even in the imperative approach,
 does the router switch then

0:09:00.080000 --> 0:09:03.220000
 say, a controller, what
 do I do with this?

0:09:03.220000 --> 0:09:09.220000
 No, it doesn't. The router switch still
 has its own local forwarding plane,

0:09:09.220000 --> 0:09:11.280000
 its own local data plane.

0:09:11.280000 --> 0:09:16.280000
 Because remember, the controller
 was the control plane.

0:09:16.280000 --> 0:09:18.040000
 It populated that table.

0:09:18.040000 --> 0:09:21.520000
 But once it populated that table, the
 controller does not see the actual

0:09:21.520000 --> 0:09:26.160000
 packet. It does not see the actual Ethernet
 frame of your telnet session,

0:09:26.160000 --> 0:09:27.700000
 of your web browsing session.

0:09:27.700000 --> 0:09:28.940000
 The router does.

0:09:28.940000 --> 0:09:30.320000
 The switch does.

0:09:30.320000 --> 0:09:33.160000
 So as that packet or frame comes in,
 it's still going to be that router

0:09:33.160000 --> 0:09:39.100000
 switch's job to ultimately forward
 it based on that table.

0:09:39.100000 --> 0:09:45.560000
 So SDN, ideally in the imperative approach,
 will separate the control

0:09:45.560000 --> 0:09:50.160000
 and data plane. So in the imperative
 approach, it's called stateful SDN

0:09:50.160000 --> 0:09:54.200000
 because the controller itself
 knows everything.

0:09:54.200000 --> 0:09:55.480000
 It knows all the routes.

0:09:55.480000 --> 0:09:59.260000
 It knows the connections because it's
 got the brain of the routing protocol

0:09:59.260000 --> 0:10:03.360000
 in itself. It has the intelligence
 to learn MAC addresses and populate

0:10:03.360000 --> 0:10:06.720000
 that stuff itself.

0:10:06.720000 --> 0:10:10.040000
 It can directly program
 the data plane.

0:10:10.040000 --> 0:10:14.440000
 So some Cisco controllers
 can do that.

0:10:14.440000 --> 0:10:16.400000
 Some Cisco controllers cannot.

0:10:16.400000 --> 0:10:19.980000
 Other Cisco controllers do what's
 called the declarative approach.

0:10:19.980000 --> 0:10:20.780000
 We talked about this.

0:10:20.780000 --> 0:10:23.060000
 This is stateless SDN.

0:10:23.060000 --> 0:10:27.840000
 This is basically where the controller
 says, okay, router, you do routing

0:10:27.840000 --> 0:10:29.220000
 like you always have.

0:10:29.220000 --> 0:10:31.800000
 You learn your EIJRP neighbors.

0:10:31.800000 --> 0:10:34.080000
 You figure out how to learn
 your OSPF neighbors.

0:10:34.080000 --> 0:10:38.620000
 You know, you form your EIJRP topology
 table and you populate the routing

0:10:38.620000 --> 0:10:39.320000
 table with that.

0:10:39.320000 --> 0:10:40.800000
 You do all that stuff.

0:10:40.800000 --> 0:10:45.060000
 And then just report back to me what
 you've learned, what your tables

0:10:45.060000 --> 0:10:48.240000
 look like, what the state
 of your interfaces are.

0:10:48.240000 --> 0:10:52.140000
 So this is called stateless SDN.

0:10:52.140000 --> 0:10:55.420000
 Both the control and data planes reside
 within the networking devices

0:10:55.420000 --> 0:10:59.720000
 themselves. And the controller though
 can listen to an application running

0:10:59.720000 --> 0:11:06.260000
 on a server and then it can declare
 to the networking devices what it

0:11:06.260000 --> 0:11:10.020000
 thinks the overlay network
 needs to look like.

0:11:10.020000 --> 0:11:15.940000
 All right, and that concludes
 this particular session.

0:11:15.940000 --> 0:11:16.640000
 Thank you for watching.

