HomeVideos

Build a Complete Free CCNA Home Lab in 2026 With No Gear

Now Playing

Build a Complete Free CCNA Home Lab in 2026 With No Gear

Transcript

607 segments

0:00

Then it says I'll create the full lab

0:02

topology in CML. It figured out the data

0:05

model that it needed to use to create

0:07

that and then it pushed that out to CML

0:09

with this configuration, roughly this

0:11

topology. And if I go back to Firefox, I

0:14

have a three EBGP. I didn't even ask for

0:17

EBGP. It just I said BGP and it said I

0:20

will do EBGP. And I don't want to take

0:22

Dalton's thunder, but we've got

0:23

something else coming

0:25

that will make this even cooler for our

0:27

wireless engineers.

0:29

>> is amazing. So Justin, if you see this

0:30

video, dude,

0:32

really impressed, man. You have now made

0:35

the lives of network engineers so much

0:37

easier. Okay, Joe, this is fantastic. Is

0:38

there some more stuff you can show us?

0:40

>> Absolutely. I can show you not only

0:42

building the lab, but building a

0:44

scenario and getting some coaching tips.

0:46

>> And one of the really cool things about

0:48

this is whenever you drag this client

0:50

across the canvas, it actually changes

0:52

the signal strength based on how far it

0:54

is on the screen from the AP. So it

0:57

really kind of mimics what it would look

0:58

like if you were physically moving away

0:59

from the AP.

1:00

>> Everyone, David Bombal back with two

1:01

very special guests. Joe, great to have

1:03

you back on the show.

1:04

>> It's great to be here.

1:05

>> Dalton, great to have you on the show.

1:06

>> Thanks. Thanks for having me.

1:07

>> So it's your first time, so perhaps you

1:08

can tell us a bit a bit about yourself.

1:10

>> Sure. Yeah, my name is Dalton Ortega.

1:12

I'm the product manager for CML and I've

1:14

been with the group since about, yeah, 4

1:15

years ago. It'll be 4 years ago later

1:17

this month.

1:17

>> Joe, you're really well known. Come on,

1:19

you've got to do this, dude.

1:19

>> Oh.

1:20

>> Come on, show us that.

1:22

>> Yes, I've collected quite a bit.

1:24

>> You've done a lot. So, I mean, you're

1:26

very well known. Just for people who

1:27

don't know who you are, though.

1:28

>> Uh yes, normally you and I speak about

1:31

the the network at Cisco Live Europe,

1:33

but this time I'm here doing Cisco

1:35

Modeling Labs um at Cisco. I actually

1:37

turned 28 at Cisco Monday. Uh so I've

1:39

done a lot. I started in TAC. Uh now I

1:42

work in our Learn with Cisco um group

1:45

and I do I work with the Cisco Modeling

1:47

Labs team. I work with our Continuing

1:49

Education program. I do standards work

1:52

at the IETF, and I'm just an all-around

1:54

network geek.

1:55

>> You're the guy I love to talk to, you,

1:56

man. So, really looking forward to this.

1:58

CML, there's been some updates, so

2:00

perhaps you guys can bring us up to

2:01

speed. Tell us what's what you guys are

2:03

doing and why the product is just

2:05

getting better and better.

2:06

>> You know what, Dalton? Why don't you, as

2:08

the product manager, yeah.

2:09

>> absolutely. So, we just released our

2:11

latest version, CML 2.10, a couple weeks

2:13

ago, and it's a pretty big one,

2:14

honestly. Yeah, there's some really cool

2:16

things we're going to show off today,

2:17

like MCP server that's been integrated

2:19

into CML. I'm going to show off a very

2:20

new wireless fabric that we have in CML,

2:23

essentially end-to-end wireless without

2:24

needing real radios. And there's a lot

2:26

of other features, too, like RADIUS

2:28

authentication, XRD, like a Docker

2:31

containerized version of a service grade

2:33

service provider routers.

2:35

A lot more other things, too, like

2:36

packet capture streaming, be able to

2:37

send your packet capture directly to

2:39

Wireshark, rather than downloading a

2:40

pcap and exporting, so a lot of fun

2:42

stuff.

2:43

>> That's great.

2:44

So, Joe, you got to you got to demo for

2:45

us, I hope.

2:46

>> I do. So, um

2:49

it wouldn't be Cisco Live in 2026 if we

2:52

didn't talk about AI. And so, CML is no

2:55

exception. We are on the AI bandwagon,

2:57

and Dalton mentioned MCP, so model

3:00

context protocol. It's one way that all

3:03

products, all applica- all things can

3:06

connect to generative AI. Um and one of

3:09

the challenges people have always had

3:11

with CML, there's been a few, but one of

3:13

the big challenges is how do I get going

3:15

quickly? We

3:17

all the time, do you have sample labs to

3:18

this and that and the other thing? And

3:21

but while we do have some static sample

3:23

labs, it's kind of fun to just think,

3:25

what do I want to learn today? Or what

3:27

do I want to play with today? And use

3:29

your natural language, whatever that may

3:32

be, English, Spanish, French, whatever,

3:34

and say, "CML, make me a lab." And that

3:37

is what this MCP server allows us to do.

3:39

Um so, here I've got my my CML, you see

3:42

that? My mouse is moving, nothing up my

3:44

sleeve, just a long badge. Um this is

3:47

I've got a lot of labs in here. I

3:48

collected a lot of craft over the years.

3:50

I've been using CML as a fanboy for a

3:53

while, but I want to create a new lab

3:55

here. I'm going to do it fairly quickly.

3:57

So, I'll come over here to cursor. So,

3:59

this is cursor. It's like a VS code.

4:02

But, what I want to point out is even

4:03

though I'm going to use cursor for this

4:05

demo, I could do this with Claude

4:06

desktop. I could do this with VS code. I

4:09

could do this with LM Studio or Ollama

4:12

and just local models if I have my own

4:14

GPU. But, I'm going to ask it, "Hey, can

4:16

you build a CML lab with three IOL

4:20

routers and a simple BGP config?" And

4:23

I'm going to send that in. And what what

4:25

happens here is MCP exposes a list of

4:29

what are called tools. And tools are

4:30

nothing more than functions. In this

4:33

case, they happen to be Python

4:34

functions.

4:35

And CML was written with API first. So,

4:39

everything I do in CML, so if I come

4:41

back over here to Firefox and I like go

4:43

down here to a lab like this Wi-Fi lab

4:46

and I click and I move around, all of

4:48

these things are actually making API

4:50

calls in the background. So, what these

4:52

tools are going to do is allow the LLM,

4:55

which is right now thinking about what

4:57

I'm asking. So, the model I'm using in

5:00

the background here is is Claude 46

5:02

Sonnet. It's thinking about what I want.

5:04

It says, "Ah, you asked for IOL routers.

5:07

So, I'm going to look and see what node

5:09

definitions." So, it went out and called

5:11

the tool to get the node definitions out

5:14

of CML. And this is simply the the LLM

5:17

translates my natural language request

5:20

into functions that it needs to call and

5:22

it figures out what arguments to pass to

5:24

those functions. So, all of this gets

5:27

done automatically. And then it says,

5:29

"Haha, I thought for 42 seconds." And I

5:32

said, "The IOL node definition is

5:33

available at this.

5:35

I'm going to use this. Now, I'll create

5:37

the lab with three IOL routers. You

5:40

asked for a simple BGP config, Joe. So,

5:42

let me come up with this IP address. You

5:45

didn't give me any IPs, so I'll just

5:47

invent something.

5:48

Um then it says I'll create the full lab

5:50

topology in CML.

5:52

Let me click on that. I'll show you it

5:55

figured out the data model that it

5:57

needed to use to create that. And you

5:58

can see that all here. And then it

6:00

pushed that out to CML with this

6:02

configuration, roughly this topology.

6:05

And if I go back to Firefox,

6:08

back to my lab list, scroll all the way

6:10

down, I have a three EBGP. I didn't even

6:14

ask for EBGP. It just I said BGP and it

6:17

said I will do EBGP.

6:19

>> So, Joe, this is great, but you know,

6:21

being the cynical old network engineer

6:22

that I am, can you just show us a

6:23

console and do a show run or something

6:25

so we can see that it actually did some

6:26

BGP?

6:27

>> Not yet. I will show you that I'm going

6:29

to I'm going to do this in two phases.

6:31

Once, I'm going to This is the day zero

6:33

config, but it didn't notice it didn't

6:35

start the lab. I didn't ask for the lab

6:37

to be started. So, let me come over here

6:39

and I'll say

6:41

great.

6:41

>> Okay.

6:42

>> Uh

6:44

this will start really quick because I

6:46

did IOL.

6:47

And look, I'm being very glib with my

6:50

I'm not saying please start this lab,

6:52

use this ID. It will say, "Okay, I need

6:55

to figure out to start a lab. Great,

6:56

I've got this."

6:58

My tool here for starting a lab um

7:01

takes a lab ID. It needs that, but

7:03

there's this optional wait for

7:04

convergence argument. I didn't say start

7:07

the lab and wait for it to be up. It

7:08

just figured you probably want me to

7:10

wait and tell you that the lab is up.

7:12

But now if I go back here, you can see

7:14

the devices are starting.

7:15

>> Oh, I see.

7:16

>> Now I can give you that console you

7:18

asked for.

7:19

I'm sure the devices are coming up.

7:22

Should be pretty quick to get the uh

7:23

prompts on router three here. Still

7:26

booting. There we go. So, router three.

7:28

So, you wanted a show run?

7:29

>> Now I can see BGP there, so yeah.

7:31

>> Yep. So, there's the show run.

7:32

>> I can trust you, Joe, but I'm you know

7:35

>> you can trust the config is there, but

7:36

let's see what show

7:38

show IP BGP summary says.

7:42

Yeah. Says I got some peers, so there

7:44

you go.

7:45

>> Well done. Well done.

7:46

>> Check's in the mail.

7:48

>> Okay, gentlemen, this is fantastic. Is

7:49

there some more stuff you can show us?

7:51

>> Absolutely. I can show you not only

7:53

building the lab, but building a

7:54

scenario and getting some coaching tips.

7:57

>> You got to show us that.

7:58

>> A lot of your listeners, a lot of the

8:00

people who come to us and learn with

8:02

Cisco, they are on a certification or a

8:04

career journey. They want to not only

8:06

learn something, but they want to make

8:08

sure they know it well and they want to

8:10

practice it with intention. So, one of

8:12

the other things that's really great, if

8:14

you're looking for a study partner, if

8:16

you're looking for someone who can help

8:18

you

8:19

test you, quiz you, that type of thing,

8:22

we have the ability to create not only

8:25

labs like I did, but scenarios.

8:28

So, you may have heard we just released

8:30

a new blueprint for the CCNA. We've

8:33

up-leveled that to include more

8:35

troubleshooting skills. So, I uploaded

8:38

that blueprint to Cursor and I said,

8:41

"Look, I'm preparing for my CCNA, man,

8:44

but

8:45

>> [gasps]

8:45

>> this troubleshooting thing, it's got me

8:47

a little worried. I don't know if I'm at

8:49

that level to do the troubleshooting.

8:51

So, can you look at this blueprint,

8:53

figure out what is a good

8:54

troubleshooting task for me to do, and

8:57

spin up a scenario for me,

8:59

let me work through it, and then take a

9:02

look at what I did, take a look at my

9:04

console history, take a look at this the

9:07

config that I finally ended up with, and

9:09

evaluate me and give me hints." So, it

9:12

built the lab, it built this lab over

9:14

here.

9:15

Down here it says CCNA troubleshooting.

9:18

So, two routers, two switches, two

9:20

clients.

9:22

And then I went through earlier and I

9:24

did some things. I

9:26

went in and I I tried to fix some of the

9:28

problems. How did I know what problems I

9:30

needed to fix? Well, the LLM used the

9:33

CML guide to tell me everything I'm

9:36

supposed to do to make this scenario

9:38

work. It sent me a broken scenario what

9:41

to do. So, I did a few things, but I

9:43

didn't do everything.

9:44

>> have to interrupt you, right? You're the

9:45

guy who made the MCP, right?

9:47

>> I'm the guy who wrote the MCP server.

9:48

>> done, dude. Sorry to interrupt you.

9:50

>> Um so, but it's a team effort. I it it

9:54

needs to

9:54

>> Credit where credit is due, dude.

9:56

>> needs it needs the creativity. So, the

9:58

LLM said, "Okay, great. You did some

10:01

things." And then it said, "Here's don't

10:04

peek into what you tried." Okay, so I

10:06

scrolled down and I said, "Okay, can you

10:08

check where I'm at? I didn't do

10:09

everything. How I've done so far, give

10:11

me an eval and some coaching tips."

10:13

It went in and it said, "Well, what you

10:15

did well was you went into router two

10:17

and you ran some show commands ahead of

10:19

you. You didn't just start making

10:20

configuration changes. You looked and

10:23

set and saw what was wrong. And then

10:26

you fixed all of that, great. And then

10:28

what I what I like here is what you got

10:32

right down here in exercise three, you

10:33

ran show IP route and you saw this, but

10:36

you missed a critical part. There's a

10:38

missing route for your network." So,

10:40

[snorts] now I have not only an

10:43

evaluation of did I get it right or did

10:44

I get it wrong, but it gives me some

10:47

some both positive cuz LLMs love to be

10:49

effusive and positive, but also, "Hey,

10:51

here's where you could have done a

10:53

little bit better. Here are things that

10:54

you could have

10:55

done that made that would make you a

10:57

more effective network engineer or a

10:59

more effective troubleshooter." And

11:01

these are great use cases

11:04

for the MCP server. The last thing I

11:06

want to show you is let's say you're

11:07

dealing with a very complex topic. This

11:10

is BGP. It can take the packet capture

11:13

from CML and analyze it down to a ladder

11:17

diagram and explain to you how a

11:19

protocol works. So, it pulls out all of

11:21

that packet capture, shows me the entire

11:24

flow of BGP all the way down to here

11:27

where you can see the updates with the

11:29

NLRI, with the routes being learned. And

11:32

this helps me to really crystallize some

11:35

of these complex things. And this is all

11:37

available in all flavors of CML. The MCP

11:40

server itself ships with 2.10, but it's

11:43

open source and works all the way back

11:45

to 2.9. So, we're really hopeful that

11:48

people start to do this. I mentioned I

11:50

appreciate you give me the credit, but

11:52

it needs the creativity and we would

11:53

love to hear from people to see what

11:55

kind of use cases they come up with

11:57

this.

11:58

>> Beats, I'm impressed.

11:59

>> Thank you.

12:00

>> Fantastic, man. So, I know you're really

12:01

humble,

12:02

but that MCP thing was amazing. Dalton,

12:04

you've got some more, right? But there's

12:06

more.

12:06

>> But there's more, but wait.

12:08

>> [laughter]

12:08

>> I do. Yeah, so I'm going to show you our

12:09

new wireless fabric that we have in CML.

12:12

So, this has always been kind of one of

12:13

the problems with CML or really any

12:15

virtualization platform is to do a

12:17

wireless simulation you still have to

12:18

connect externally to a physical radio,

12:20

physical AP. So, with this new wireless

12:23

fabric you can actually run that inside

12:25

CML. So, what we're doing with this

12:26

first kind of approach that is running a

12:28

Ubuntu-based WPA supplicant for our APs

12:32

and for our clients. So, what that

12:33

allows us to do is to be able to connect

12:35

these clients. So, in this lab for

12:37

example, I've got two APs. They're both

12:40

sharing the same SSID. And I can see

12:42

with this client right here the signal

12:45

strength between that AP and this

12:46

client. And one of the really cool

12:48

things about this is whenever you drag

12:50

this client across the canvas, it

12:52

actually changes the signal strength

12:54

based on how far it is on the screen

12:56

from the AP. So, it really kind of

12:58

mimics what it would look like if you

12:59

were physically moving away from the AP.

13:01

So, you'll see signal degradation. You

13:03

can actually simulate roaming. So, for

13:05

example, if I go way past that AP1 over

13:08

here to AP2, it's going to recognize

13:10

that. And I've got a BG scan set up in

13:12

here. So, when it passes like a certain

13:14

threshold, it'll automatically search

13:15

for another SSID. So, that's what this

13:17

is going to do. Yeah, and as you can see

13:19

here, it just automatically snapped over

13:21

to our other AP. Yeah, it's really fun

13:22

to play with. How on earth do you get

13:24

Wi-Fi in a lab, though? Is this Is this

13:26

is amazing because it's always It's been

13:28

a problem forever.

13:29

>> Right. Yeah.

13:30

>> So how do you Is it just like magic?

13:32

>> It is. No.

13:33

>> [laughter]

13:35

>> I wouldn't really know how to describe

13:36

that, actually.

13:36

>> So, um the

13:38

I did not write this one. This is Justin

13:41

Quan

13:41

um he he created this. He had this

13:43

vision. He's had it for a while. He He

13:45

was a wireless engineer. I think he's a

13:47

wireless CCIE.

13:48

Um he figured out that there is a way to

13:53

multiplex what the wireless um the

13:56

wireless type of network does with uh

13:59

what CML does. And he's got a patent

14:01

pending on this, actually. Uh Cisco had

14:04

this thing internally because we need to

14:06

test our own gear called WSim, the

14:09

wireless simulator. He extracted bits

14:11

from that and figured out that there was

14:14

a emulated or virtual version of the

14:17

kernel module that does Wi-Fi for Linux

14:20

that he could adapt into CML. And he

14:22

created

14:24

network engineer software developers,

14:26

especially, they're they're kind of um

14:28

they're kind of geeky in how they think.

14:30

So, he created two components, the air

14:31

handler and the air duct. It's wireless,

14:34

so it's over the air. So, these are

14:35

back-end processes that enable that

14:37

wireless fabric to work together between

14:39

the client and the AP. But the cool

14:42

element that Dalton's showing you is

14:44

this

14:45

at the fore- at at the foreground of

14:47

this or in the in the in the um

14:49

uh you have the back end and the front

14:51

end. That's the word I'm looking for,

14:52

the front end. It's all regular Linux

14:54

wireless. So, WPA supplicant on the

14:56

client, host APD on the AP. And I don't

14:59

want to take Dalton's thunder, but we've

15:01

got something else coming.

15:02

>> Um that will make this even cooler for

15:05

wireless engineers.

15:06

>> this is amazing. So, Justin, if you see

15:08

this video, dude,

15:09

>> Very good.

15:10

>> really impressed, man. You have now made

15:12

the lives of network engineers so much

15:14

easier. Sorry, Dalton. Carry on.

15:16

>> Yeah.

15:17

>> there's more.

15:18

>> Yeah, and even to add on to that. So,

15:19

this is kind of our first stab at doing

15:21

wireless inside CML. So, it is based on

15:23

Ubuntu WPA supplicant, but in the future

15:26

here very soon we're going to have

15:27

actual Cisco virtual access points

15:29

shipping with CML. So, they'll have like

15:30

the full tool set, the GUI, all that.

15:32

So, you can stand up your your 9800 WLC,

15:35

you can have the Cisco VAP on there, you

15:37

can have some clients, and then really

15:38

have a pretty robust wireless network.

15:40

>> So, this is great. Just show me like how

15:42

do I build this? Is it just like I drag

15:44

like in like a router or switch in the

15:46

same kind of process I drag a wireless

15:48

device in?

15:48

>> Yeah, same kind of process. So, it'll be

15:50

in your nodes definition, your node

15:51

pallet. So, if you were to just drag

15:53

that on there, and then you can come in

15:54

here

15:56

and I did create a day zero config with

15:58

a BG scan. By default, it doesn't have

16:00

that on there. So, if you do want it to

16:01

have like the ability to do those kind

16:03

of scans past a certain threshold,

16:04

you'll have to import that. But, yeah, I

16:06

know you can come in here, you can

16:07

actually manually force roam if you want

16:09

to, you can see what's happening on the

16:11

link directly and which AP you're direct

16:12

you're connected to at the time. You can

16:14

even run like packet capture on those

16:16

links. I think that's one thing I didn't

16:17

talk about yet is running wireless

16:19

packet capture.

16:20

>> Are you serious?

16:21

>> Yeah. So, I'll run that and you can see

16:23

see what happens when I go from say AP1

16:25

to AP2.

16:27

So, there's AP1 right there the AABFF2.

16:31

And when I come over here, once it snaps

16:33

over, you should see it start connecting

16:36

to What is this other one? AA06. See, I

16:39

give it a second. The UI catches up with

16:41

it eventually, but it is in beta, so

16:44

that is expected. There it goes. So,

16:46

yeah, that is snapped over. So, now it's

16:47

on AA062F.

16:49

So, yeah, you can actually see that

16:50

happening in real time. And like I

16:52

mentioned, you can stream this directly

16:53

into Wireshark, too. So, if you wanted

16:55

to have Wireshark running on another

16:56

monitor, have your wireless simulation

16:58

here in CML, you can have that.

17:00

>> So, you're saying that I can run Wi-

17:02

Sorry, I can run run Wireshark and

17:03

capture all the Wi-Fi stuff.

17:05

>> Correct.

17:06

>> From the AP side or the client side?

17:08

>> So, the the AP packet capture um right

17:12

now filters on packets originating from

17:15

the source because because it's shared

17:17

medium. If you did everything it gets

17:18

very noisy but you do get the dot 11

17:21

frames which is super cool if you're

17:23

trying to learn wireless and understand

17:25

what those frames look like.

17:27

>> Wow, that is an that's a massive update.

17:30

It's a big one.

17:30

>> I mean MCP, Joe what you did is amazing,

17:33

what Justin did yes amazing. Dalton you

17:35

managing a bunch of rockstars dude.

17:37

>> I know I'm lucky to have them for sure.

17:39

>> So Dalton you talked about the the

17:40

streaming packet capture this is

17:42

actually a really cool feature because

17:44

if I'm trying to troubleshoot something

17:46

yes I can like I want to know what's

17:48

happening with my BGP here so I can

17:50

click on this link I can right click and

17:52

I can say packet capture and I can say

17:54

start it up here and it runs I can come

17:57

back over to here and do a clear IP BGP

18:00

star and I get stuff but you know

18:05

it's nice I can drill in but there are

18:07

other features of Wireshark that I would

18:09

like to use. So now in 2.10 there's a

18:11

new version of the breakout tool so if

18:14

you've never seen the breakout tool you

18:15

go to tools breakout tool

18:18

you download this you need to get the

18:19

new one that comes with 2.10 cuz there's

18:21

this new feature um I got this packet

18:24

capture running I'll come over here to

18:26

my breakout directory and I run breakout

18:30

pcap and it will list all of the active

18:33

packet captures I've got running on the

18:35

server I need to get this key so this is

18:38

the one active one I've got I'll rerun

18:40

it but I'll add the key this time it'll

18:43

automatically launch Wireshark for me

18:45

now I'm I'm I'm sucking in all those

18:47

live packets so let's go ahead and redo

18:50

that clear IP BGP again

18:52

and check out what Wireshark sees.

18:54

Now we're not on we're seeing all of the

18:56

packets here I can filter them I can

18:58

follow TCP stream I can look at latency

19:01

very easily between packets and I can

19:03

drill in remember I talked about that

19:05

those update messages that the LLM was

19:08

describing I can see all of that here as

19:11

well.

19:11

>> That is great.

19:12

>> And this is a great troubleshooting

19:13

tool. This is a great learning tool to

19:15

be a companion with your CML.

19:17

>> And then what I love about what you guys

19:18

have done, right?

19:20

Is it's there's a free version.

19:22

>> Mhm.

19:22

>> And this all all that we showed works

19:24

with the free version.

19:25

>> That's amazing.

19:25

>> Yeah.

19:26

>> When you studying for CCNA or any

19:28

networking cert, you need hands-on. And

19:30

I mean, this is fantastic.

19:32

>> Thank you.

19:32

>> Really, thanks so much. This is amazing.

19:34

>> Thanks for having us.

19:35

>> Yeah, absolutely.

19:35

>> Thank you.

19:36

>> We'd love to get to your follower, let

19:37

them know this is um it's there for you.

19:40

Download it now. Start playing today.

19:42

Let us know what you think.

19:43

>> Yeah, so guys, please put in the

19:44

comments what your thoughts are about

19:46

this and what else you want to see. I'm

19:48

sure everyone's got an opinion. Give

19:49

feedback. The product just gets better

19:51

every single year. Guys, thanks so much.

19:53

>> Thank you. We appreciate it.

Interactive Summary

The video highlights the major updates in Cisco Modeling Labs (CML) version 2.10. Key features include an AI-driven Model Context Protocol (MCP) server that allows users to generate and configure lab topologies using natural language, a new wireless simulation fabric that enables end-to-end wireless testing without physical hardware, and enhanced packet capture capabilities that allow for direct streaming into Wireshark. The presenters demonstrate how these tools can be used for both building lab scenarios and troubleshooting complex networking protocols.

Suggested questions

3 ready-made prompts