- [Announcer] Previously... - But like, if I'm starting today is there any kind of path
that you would recommend someone take based on what you've seen like do I become a CCNA? Do I go and do DevNet? Is that...what are the
sort of the technologies we did this before but I'd
like to get an updated list. Are there any top five top 10 technologies that you think are really important and are there any...is
there any path to get there? - Well, you see, it depends on whether you're looking
short term or long term. Short term today, I would
go with automation cloud. These are the things that will be probably in greatest demand in the
next three to five years. Long-term learn the fundamentals. IP routing hasn't changed in 40 years. Ethernet hasn't changed in 40 years. Everything we did was up the speeds. 400 gig Ethernet is still ethernet. - [David] Yeah, yeah. - It's still running IPv4
and IPv6 on top of that. And it's still transporting
more IPv4 than IPv6. TCP is slowly changing with
quake and things like that but OSPF is there, BGP
will not go anywhere. ARP will be there forever because you know if you have layer two and layer three you have to map them somehow. So regardless of what pays the bills learn the fundamentals cause eventually you will need them. - [David] Yeah. - And eventually every
environment will get to a point where the network will be down and who will it troubleshoot that? The Python guys? The cloud guys? Or someone who took time
and learn the fundamentals? (high upbeat music) - Okay, so what are the skills that you reckon are important? And when you say fundamentals,
how would I learn that? - CCNA is not bad.
Yep. - But look at CCNA not as cert. Well, yeah, we're cert It's, you know, just to
prevent being thrown out of the huge pile of CVS. Yeah, just go for it and get the cert and you'll learn something, but more importantly, figure
out what are the things that you should get out of CCNA. How does Ethernet at work? How does routing work? What is IP subnetting? Do they do port channels?
- Yes. - Port channels, Spanning
Tree, OSPF that stuff. And then, you know what? Built a lab with Cumulus VX boxes and do that same thing on Linux, just for the giggles. And then you will learn how, It's different CLI, is different approach. It's monolith versus multiple processes. But in the end you will get OSPF running between a Cisco box and a Linux host. And then you will
understand how things work a little bit better than before. - So when we spoke about,
it's a year ago already you mentioned like some skills
like you need to learn Linux. You've kind of mentioned that again. So for the short term. So, I mean everything's
short term these days, but let's say in the
next two to three years, what...or five years even, what do you think are the top
skills that I should learn If I'm trying to get into this industry. You've mentioned CCNA kind of like get you the basics of networking, but do you have any other
like, sort of high-level skills that you think are really important? Like if you had to choose
five things or 10 things what would you pick Linux perhaps? And then? - Well, we mentioned learning, right? - Yeah
Yeah. Deep work. I read the book Deep Work
by Cal Newport, I think? - Yeah, it's good book. - It's an awesome book. And it sent me thinking, so, your brain is your
most valuable muscle. And if people are willing
to spend time in the gym and train the other muscles you should train your brain as well. Which means that you should
be able to take a problem and focus on that problem
and work on that problem for like three to four
hours uninterrupted. And if you can't do that,
you have a serious problem. The other thing is just do honest work in the evening look yourself in the mirror and ask yourself did I do
everything I could do today to get the job done? And if the answer is, nah,
then you have a problem. - I'm gonna push you Ivan,
because you've mentioned like a it's more philosophical type stuff which I think is really, really important. But are there any practical
skills that you think I should, like you mentioned cloud, should I like, and I don't... if you're
not happy with the question, that's fine. But like you mentioned CCNA should I go into do AWS,
like associate cert? What do you think are the
big trends or the certs or the knowledge that I should try and get in the next few years? - Well, definitely still
don't ask me about the certs because honestly I don't care about them. - Yeah, that's fine. So give us like, sort of the technologies. - Yeah, I told you. You need some basic
networking fundamentals. - Yep. - So how do networks work? Then clouds works slightly different than the traditional networks. So learn how cloud networking works. Don't only focus on networking because once you get to the
cloud, it's all intermixed. - [David] Yep. It's VMs, it's storage, it's containers. It's Lambdas, it's networking,
it's Load Balancing. It's security, Packet filters,
application, firewalls. And you know what? It's all free. All three big cloud providers
have a free tier, use it. The only thing it's
costing you is your time. And then the next thing
obviously is figure out how to automate stuff. So for example, you're studying for CCNA, for God's sake don't use
GUI to set up your lab. (David laughs) Use Ansible, use CLI, use
API, use anything but GUI. - [David] Yep. - You want to set up your OSPF lab. Don't even think about
logging into five routers and typing IP addresses. What have you learned? Typing skills. Hooray, we're really need that. Automate that, learn Ansible, learn how to build configs, push configs to the boxes. You'll master CCNA,
you'll master networking. You'll figure out how things work. Plus you'll learn something else. Once you start doing that
don't have scribd.1.1.a.3.z on your folder. Use Git branches. Figure out how you can
experiment and test things out. And destroy failed experiments and accept successful experiments with... for example Git branches. Everything you do, there
is always something extra you can do to learn something else. Not just the one thing
that you are working out. And you know what it's all free. Git is free. Ansible is free. Linux is free. Baseline cloud theory is free. Cumulus VX is free. Juniper vSRX 3.0 is free. Arista EOS is free. Well the Vagrant box.
- Yeah. - Cisco is not free. Who cares? Oh yeah, Nexus 9000v is. iOS is not free. Forget iOS, let's move on. Sorry Cisco. Gets your house in order. - That's funny that you mentioned that cause one of the questions
I wanted to ask you Ivan is Cisco, Juniper, Arista
what should I learn? I mean, you've mentioned
Cumulus quite a few times so you-
- All of them. - All of them. - It doesn't matter. It really depends on, do
you want to be a CLI jockey or do you want to be an engineer? - Yeah. - Do you have a brass mechanical engineer and an iron mechanical engineer and a steel mechanical engineer? I don't think so. - [David] Good point. - So why do we have a Cisco-CLI jockey and a Juniper CLI jockey
and Arista CLI jockey they all run OSPF is just
configured differently. Well, no I Arista is using
industry standards CLI. (both laugh) - So in other words, you're
saying learn the technology like OSPF don't worry
so much about the CLI. Is that what you're saying? - Yeah. No, honestly, what I did was
probably two months ago now I said, well, now I want to build a tool to set up my automation labs. - [David] I saw that, yeah. - Yeah, so I build that Python tool that takes the left topology and creates device data models. And now I can do initial config
and the next step I'll do for example, OSPF or BGP or what have you. But as part of that experience, I built boxes for as
many different platforms as I could get my hands on. NX-OS is downloadable,
VSRX 3.0 is downloadable VMX I think isn't but it's
a mess to set up anyway because they want to have two VMs one for data playing,
one for control plane so just go with VSRX. - [David] Yeah. Arista is downloadable. Nexus 9000v is downloadable. Cumulus is downloadable. The problem I had was that
I wanted to do this on KVM cause learn a new stuff. - [David] Yeah. - Vagrant on the MacOS or Windows. It works out of the box, you
can just pull down the boxes. It all works. It's all crap, but it works
and you learn nothing new. So I built all those boxes myself. And now I have a running
a lib variety environment on a Linux box with
automatic provisioning. And I just build the left
topology and bam it goes up and now I can really test anything I want between like four different vendors. That's totally different
than trying to learn the CRI commands for a single vendor. And it's not harder
because they are similar. Well, ignoring Juniper,
Juniper is special. But even Cumulus, once
you get it up and running it's a Layer 2 stuff is
special, interfaces are special. But once you get to OSPF and BGP is free range routing which is Quagga, which is really Cisco syntax. - Ivan, what do you think
are the next big trends? What do you think is a good
trend or trends to look at if you're just starting in this industry? So are there any waves that I
can ride if I've just started? - For a young person going into IT, I would say focus on the cloud. - [David] Okay. - Because everyone will
have to do something with the clouds eventually. And traditional environments
will have no idea whatsoever how to do that. So, it's like when we went
from Cobalt to JavaScript 30 years ago or what era it was. Or was it Java, probably
more Java and JavaScript. There were tons of people
familiar with the old technologies and no one familiar with the new ones and everyone wanted to be
hip and use the new stuff. And probably cloud is the new stuff at least in some geographies, obviously not for Silicon Valley,
obviously not for startups but for traditional IT. Do you think they can spell cloud? - [David] So something up
there in the sky, isn't it? - Yeah, exactly. It's rains right?
- Yeah. So Ivan you've mentioned get ready. So it's interesting. What's your opinion of Cisco DevNet and that those kinds of certifications - They can't hurt. So you see, it's always the question
of where you're coming from and why you need certain things. So if you are an existing
networking engineer and you think that you
have to become a programmer which is a wrong idea but we can go there some other time. Then yes something like DevNet
is the right thing to do because you already have
networking expertise, you know nothing about Python and REST API and Git and all that stuff,
you have to learn those things. If you're a developer, then you already know all those things. And if you want to be a
network automation developer then maybe you should focus more on understanding how networks really work. Because you have the developers
that can develop anything and are not good at anything. And then you have people
who have actually worked in some industry for ages. So for example you might have a mediocre
developer who has been working in chemical industry
for the last 20 years. He will beat any whiskey because he knows all the dirty details. All the requirements, all the regulations, all the audits you have to go through. Whereas, someone who might
be a brilliant programmer has no idea about the industry and the same thing if you
want to be a good developer in network automation space then maybe it's more than Python. Maybe you should know
how networks really work. - Do you think there's going to be a trend where network engineers, the
traditional network engineers are replaced with guys who do automation? Is it... are we kind of all
forced to do automation now? - Well, you see the way
we've been doing things in the past is crazy. - [David] Yeah. - I mean, just configuring the same crap on thousands remote
office routers manually and using Excel to replace the IP prefixes in your configuration- - Is the future. - What stone age are we living in? - [David] Exactly! - So, yeah, we need to clean
up the crap that we're in. And one way of cleaning it up
is through network automation. - You've been in this
game for a long time. What would you suggest someone
do if they starting out? So in other words, if
you could go back in time or let's say you were 18 years
old and you started today what would you tell your younger self? - Oh, my younger self, okay. - Or, anyone in the industry
what would you advise? - Well that's probably a bit different, but let's start with my younger self. - Okay. - The most important lesson, technology doesn't matter as
much as you think it does. So honestly, when you're
young and brave and stupid you think that you can
solve all the problems with throwing technology at
them, it doesn't work that way. Then you play MacGyver and you solve impossible
problems by configuring stuff in the way it was never
designed to be configured. It works in a lab. It breaks in production. You're called at 2:00
AM. on a Sunday morning, you have to rush off to the
customer to fix your crap. You learn. Then, sometimes it might
work for a year or two and then you install a
new software release. And of course it crashes because no one ever used the
same box in the same combo with the same config. And you felt so smart. And now you feel utterly stupid
and mad at the bloody vendor for sending you the bug forgetting that you are
the only idiot in the world using that particular code path. So minor things like that. Oh, there's one more. Old people aren't stupid contrary to what all
the youngsters believe. And I'm not like defending myself. I was in that position. If they do something and you
think it's totally stupid, it might be. I mean, after all, they
might have stopped thinking after a while and just do
things the way they do things because they've always
done things that way. But there usually is a reason behind that the reason might be irrelevant or it might be wrong or whatever. But you have to find out the reason first before you're trying to fix stuff. There's this English saying Ken I always forget what it is about not moving a
fence that you encounter if you have no idea
why the fence is there. - Yes, I mean basically you were like that when you were younger. I think most of us were- - Oh yeah, off course - You knew better than other people,- - Everyone else, of course. - Especially if they were older- - I was the smartest
person in the universe. - Of course. So, I mean, the running joke today is like boomers get out the way. So, you basically telling yourself if you were younger, that's a mistake. Is that right? - Well, think first. That's the most important message. There's a reason people are doing things figure out what the reason is. Don't just rush in and
take over and do your stuff thinking you know best. Sometimes you do know better. Sometimes they do stuff
for stupid reasons. Sometimes they really are boomers, but sometimes there's something behind it that you just don't get. - So even if... my counter to you would be like technology's
changing the world. I mean, when we were kids,
Facebook didn't exist Instagram, all these massive websites. So surely technology is
really, really important. Or what are you saying
to your younger self that you should learn
apart from technology - Listening.
That's a good one. - People skills, communicating, listening. Did I say listening?
- You did. (both laugh) - Listening. The most important thing
is listening to what people are telling you and trying to understand what they're really saying,
not what you're hearing. Cause sometimes, a lot of
times we have a problem because they're using their terminology and you think it means something else and in the end you do something or you implement something
that is total crap. Like my favorite one, you know it's stretching
VLANS across two data centers. - [David] Layer 2 your favorite - Layer 2 into the cloud. - Yep, that sounds like a great idea. - Yeah, it's awesome. I mean, earth is flat, right? What could go wrong? - I've heard this a lot. People talk about people skills and doing things differently. But the question I wanted
to really ask you is give us an example of
where you really messed up where if you had your wisdom and the experience that you have today you wouldn't have perhaps
gone and done something. So it's always nice to get a story of where you made a mistake. - Like...things like debug,
IPBGP on the core router that connected a country
to the wider internet. - [David] That sounds like a good example. So tell us a bit about that. - Well, we were setting up a BGP feed with the upstream provider, and we already had a number
of production customers on data box that was the router. That was the hub of the commercial internet
in Slovenia and BGP. We were getting some prefixes and I got the route
maps wrong or something. So of course, what do you do? Debug IPBGP. Let's see all the updates. Well, yeah, a hundred
thousand updates later on a console port of a Cisco router. The real problem is for the
youngsters who don't know that in the good old days, I don't know whether
that still works today. You were able to break into
iOS on the console port until you would send the brake signal and you would be in a debugger. - Yeah. - Which means that console must be the highest priority
interrupt on the system. Otherwise, something could lock it out. Now imagine that you
are generating thousands of highest priority interrupts per second just because someone is
printing debugging messages to the console. Obviously they router just goes like, ah I don't want to live anymore. - You know it's great to
hear stories like that because you very respected. And I mean, these days, people look at you and think, this guy can't make a mistake. So it's great to hear that. You are actually human
- No, Designing a network with
20 parallel ERG processes. - You did that. - Yeah, It worked. - Is that an example of
you, like you said earlier you taking your technical knowledge and trying to fix a bad design. Like, I mean I think you've mentioned
this in other videos where you talk about like OSPF craziness and just trying to sort out
mad stuff because you can. - Yeah, when you're young and crazy, you think that you can solve
every problem out there with technology. - Yeah. - And yeah, my ERPG design
was one of those things. Effectively, the customer needed MPLS VPN and we told them how to do stuff and install the core routers and then install the routes
filters on the core router so that the edge routers will
only get certain prefixes. They had a shared address
space for everyone and MPLS VPN wasn't even
on the drawing board. So we couldn't use that. And the customer said, no, no, no, no. For political reasons,
it has to be framed relay in the middle from company X and routers at the edges
bought from our company. The pie chart diagram, every supplier has to get certain
percentage of the business so no one will a complain. Layer nine or 10th, right?
- Yep. - In the OSI model. And so we had to do something and they wanted... think they wanted to implement the whole
thing with Frame Relay DLCI so that every tenant would have its own partial mesh of Frame Relay Circuits. But there was this still
this problem of the web. There were two problems. One was that they had the central site and everyone had to connect
to that central site. And then you get route
leaking and isolation is gone. And even more interestingly they had multiple tenants
sitting in the same building and they were not keen on
buying a router per tenant. I don't know why. I mean, after all, they
wanted to have Frame Relay and you know, anyway, so we
solved that thing by running, yeah. One EIGRP process per customer. - [David] Men! - Yeah, and I even asked
the friends at Cisco TAC what they think about that. And they go like, yeah, it should work, but do remember that you are the only one in the world doing that. So you might encounter a few extra bugs. - Oh wow. And I'm assuming that's
what happened, yeah. - No, actually that thing
worked, it was too successful. - Too successful. So you can like solve every problem now. - Well, yeah, they thought so. So what happened was that
the customer took that ran with it and then they figured out that they needed to
provide internet access to those locations and central
services and something else. And Oh, and they had the regional hubs. So in the end it was
this orthogonal matrix of EIGRP processes per regions and EIGRP processes per tenant. It's two way redistribution between them. - But at least you had a
job security, didn't you? - Well, someone, another
consulting company took over that design and they were milking
the customer for years - I could imagine. - But yeah, they had job security. (high upbeat music)
