Kubernetes is the platform to build platforms, like Cloud Run or Fargate

In this episode of Cloud Commute, host Chris Engelbert welcomes Abdel Sghiouar, a Kubernetes expert from Google, to discuss various aspects of cloud technology, with a focus on Kubernetes and cloud-native solutions. Abdel, who has been with Google for a decade and specializes in Kubernetes networking, security, and storage, shares insights into the evolution and current state of container technologies.

The conversation begins with Abdel introducing himself and his role at Google, where he works on the Google Kubernetes Engine (GKE). Abdel explains the importance of Kubernetes in modern cloud computing, highlighting its flexibility and scalability, which make it suitable for AI and machine learning workloads that require substantial compute power.

Chris and Abdel delve into the history and impact of containerization, discussing how it evolved from virtual machines to containers and now to Kubernetes. They touch upon the benefits and challenges of adopting Kubernetes and how cloud providers have simplified container services to make them more accessible.

Abdel also talks about his involvement in the Kubernetes community, including his Kubernetes-focused podcast and his newsletter, which covers updates in the Kubernetes ecosystem. He emphasizes the importance of making Kubernetes more user-friendly and less complex, advocating for solutions that abstract away the intricacies of Kubernetes to provide a seamless developer experience.

The Kubernetes Podcast by Google episode mentioned in the show can be found here: https://kubernetespodcast.com/episode/208-containerd/

For questions, you can reach Abdel at:

http://bit.ly/3LBOfOw

You can learn more about Google at:

https://www.google.com

The Cloud Commute Podcast is presented by simplyblock (https://www.simplyblock.io)

01:00:00
Where things start

01:00:01
to be interesting,

01:00:02
or where people

01:00:03
or where people start looking into

01:00:04
using Kubernetes,

01:00:06
specifically and of course we're

01:00:08
talking here about Google Cloud,

01:00:09
so GKE, is when you start having,

01:00:12
or start requiring things that

01:00:15
the simple containerless service

01:00:16
platform doesn't give you.

01:00:17
doesn't give you.

01:00:18
And since we're

01:00:18
Since we're in the age of AI,

01:00:19
we can talk about GPUs, right?

01:00:21
So if you

01:00:21
So if you need a GPU,

01:00:23
like the promise of a serverless

01:00:25
platform is very fast scaling.

01:00:28
Very fast scaling and GPUs doesn't

01:00:30
really go hand in hand.

01:00:32
Just bringing up a

01:00:33
Linux node and installing

01:00:35
NVIDIA drivers to

01:00:36
have a GPU ready,

01:00:37
that takes 15 minutes.

01:00:38
I don't think

01:00:38
I don't think anybody

01:00:39
will be able to sell

01:00:40
a serverless platform that scales

01:00:41
in 15 minutes, right?

01:00:47
You're listening to Simply Block's

01:00:48
Cloud Commute podcast, your weekly

01:00:50
20-minute podcast

01:00:51
about cloud technologies,

01:00:52
Kubernetes, security,

01:00:53
sustainability, and more.

01:00:55
Hello everyone. Welcome back to

01:00:57
the next episode of Cloud Commute.

01:00:59
The podcast about Cloud,

01:01:01
Kubernetes,

01:01:03
Cloud Security, Cloud Storage.

01:01:05
Cloud, Cloud, Cloud,

01:01:05
Cloud, Cloud, whatever.

01:01:07
Today with me, I

01:01:08
have a good friend.

01:01:10
We know each other for quite a few

01:01:12
years, I don't know,

01:01:14
it's going to be a

01:01:15
couple of years back,

01:01:17
Abdel, and I don't try to

01:01:19
pronounce your last name.

01:01:20
I'm really sorry. Maybe you can

01:01:22
introduce yourself real quick,

01:01:24
like who you are, what you do, and

01:01:25
where you come from.

01:01:27
Sure. Abdel is fine.

01:01:29
Yes, Abdel Sghiouar

01:01:30
if you want,

01:01:31
that's how typically people

01:01:32
pronounce my last name.

01:01:34
I'm based out of

01:01:36
Stockholm, work for Google.

01:01:37
I've been at the

01:01:37
company for 10 years.

01:01:39
I do Kubernetes stuff.

01:01:41
I do a Kubernetes podcast,

01:01:42
and I talk about

01:01:45
Kubernetes and containers.

01:01:47
I arguably do more talks about why

01:01:49
you probably do not need

01:01:51
any of these technologies than

01:01:52
why you need them actually.

01:01:55
Originally from

01:01:55
Morocco, that's the back story.

01:01:59
Have you made it to

01:02:00
Devoxx Morocco yet?

01:02:02
I was supposed to go this year,

01:02:04
but I caught

01:02:06
COVID the week before.

01:02:09
Oh, yes, I remember.

01:02:11
Yeah, that was real unfortunate.

01:02:14
I was so looking forward

01:02:15
to that. I was like, no.

01:02:18
I keep promoting the conference

01:02:20
everywhere I go,

01:02:21
and then I forget who

01:02:22
made it or who didn't.

01:02:23
Okay, well, maybe

01:02:24
2024 is what happened.

01:02:25
Yes, next time.

01:02:27
Next time.

01:02:28
Like last time, I

01:02:29
promised last time,

01:02:31
and I still failed.

01:02:33
All right, you said

01:02:34
you work for Google.

01:02:35
I think everyone knows Google,

01:02:36
but your role is

01:02:38
slightly different.

01:02:39
You're not

01:02:40
working on a specific...

01:02:43
Well, you work on a specific

01:02:44
product, but as you said,

01:02:45
you're trying to tell

01:02:46
people why not to use it.

01:02:50
Which I don't think my manager

01:02:51
would be very

01:02:51
happy if he knows that

01:02:52
that's what I do.

01:02:54
All right, so tell us about it so

01:02:56
I can send it to him.

01:02:58
So, yes, I work on GKE.

01:03:01
GKE is Google Kubernetes Engine,

01:03:02
which is the managed version of

01:03:03
Kubernetes that we

01:03:04
do on Google Cloud.

01:03:05
That's technically

01:03:06
the product I work on.

01:03:07
I think that we were

01:03:08
having this conversation.

01:03:09
I avoid being too salesy, so I

01:03:11
don't really talk about GKE itself

01:03:13
unless there are

01:03:13
specific circumstances.

01:03:15
And of course,

01:03:16
Kubernetes itself, as a platform,

01:03:17
it's more specific to an audience.

01:03:20
You cannot just go to random

01:03:21
front-end conferences

01:03:22
and start talking

01:03:23
about Kubernetes.

01:03:23
No one would understand

01:03:24
whatever you're saying.

01:03:27
But, yeah, GKE is my main thing.

01:03:29
I do a lot of networking and

01:03:31
security within GKE.

01:03:34
I actually started this year

01:03:35
looking a little bit into storage

01:03:36
and provisioning

01:03:38
because, surprise,

01:03:39
surprise, AI is here.

01:03:42
And people need it.

01:03:44
And when you are doing wild

01:03:47
machine learning workloads,

01:03:48
really, on anything, you need

01:03:49
storage because these

01:03:50
models are super huge.

01:03:51
And storage

01:03:52
performance is important.

01:03:53
So, yeah, so that's what I do.

01:03:56
I think predictability is also

01:03:58
important, right?

01:03:59
It's important

01:03:59
that the data model,

01:04:03
sometimes it's not like you want

01:04:04
it sometimes fast

01:04:05
and you want it sometimes slow.

01:04:08
You want it very fast.

01:04:09
You want it to have

01:04:09
same speed consistently.

01:04:11
Yeah, that's true.

01:04:12
It's actually interesting.

01:04:13
I mean, beside the product itself,

01:04:15
it's interesting.

01:04:15
A lot of these AI companies that

01:04:17
came to the market in the last,

01:04:18
I'd say, year, year and a half,

01:04:21
they basically

01:04:22
just use Kubernetes.

01:04:23
I mean, ChatGPT is trained on

01:04:24
Kubernetes, right?

01:04:26
It sounds like, oh, yeah, I just

01:04:27
need nodes with

01:04:28
GPU, Kubernetes, boom.

01:04:31
And to be honest, it

01:04:32
totally makes sense, right?

01:04:34
Because you need to have

01:04:35
a lot of compute power.

01:04:36
You have to scale up

01:04:37
and down really fast,

01:04:39
depending on how much users you

01:04:40
currently have, right?

01:04:42
So I think anything like that with

01:04:44
Kubernetes makes perfect sense.

01:04:46
Exactly.

01:04:48
So you mentioned you

01:04:50
also have a podcast.

01:04:51
Tell me about it.

01:04:52
It's not like I have listeners that

01:04:55
you don't have, but who knows?

01:04:58
Yeah, it's the

01:04:59
Kubernetes podcast by Google.

01:05:00
It's a show that's been running

01:05:02
for almost six years.

01:05:04
I think we're close to six years.

01:05:06
We've been doing it for two years,

01:05:07
me and my co-host.

01:05:09
So me and Kaslin

01:05:09
Fields from Seattle.

01:05:12
It's a twice a month show.

01:05:15
We basically invite people from

01:05:17
different parts of the

01:05:19
cloud-native world,

01:05:20
I would say.

01:05:20
So we talk to maintainers.

01:05:22
We talk to community members.

01:05:24
We are looking into bringing some

01:05:27
other interesting

01:05:28
guests that are--

01:05:33
one of my personal interests is

01:05:36
basically the intersection

01:05:37
of technology and science.

01:05:40
And I don't know if your audience

01:05:42
would know this,

01:05:42
but at KubeCon in Barcelona in

01:05:46
2019, one of the

01:05:47
keynotes that was actually done

01:05:49
was done by people from

01:05:50
CERN who came on stage

01:05:52
and showed how they could

01:05:53
replicate the Higgs boson

01:05:54
experiments on Kubernetes.

01:05:57
So that was before I was a host.

01:05:59
But we are exploring the idea now

01:06:02
of finding organizations

01:06:05
that their job is

01:06:06
not really technology.

01:06:07
They're doing science, but they're

01:06:08
using Kubernetes to

01:06:09
enable the science.

01:06:10
And we're talking to them.

01:06:12
So there's quite a lot of

01:06:12
interesting things happening.

01:06:14
That's how we are going to be

01:06:15
releasing very soon.

01:06:18
But yeah, so it's a

01:06:20
one-hour conversation.

01:06:21
We try to do some news.

01:06:22
We try to figure out who bought

01:06:24
who, who acquired who,

01:06:26
who removed which version,

01:06:28
who changed licenses in the last

01:06:30
two weeks since the last episode.

01:06:32
And then we just

01:06:33
interviewed the guests.

01:06:34
So yeah, that's

01:06:36
the show essentially.

01:06:37
Right.

01:06:38
And correct me if I'm wrong, but I

01:06:39
think you're also

01:06:40
writing the "What Happened in

01:06:41
Kubernetes" newsletter this week.

01:06:44
This week in GKE.

01:06:45
Yes.

01:06:46
Or just, oh, this week in GKE.

01:06:48
OK, no.

01:06:48
We're not talking about that.

01:06:51
Which is not only about GKE.

01:06:53
So I have a newsletter on LinkedIn

01:06:54
called "This Week in GKE",

01:06:56
which covers a couple of things

01:06:57
that happened in GKE in this week,

01:06:59
but also covers other cloud stuff.

01:07:02
All right.

01:07:02
Fair enough.

01:07:03
So you're allowed

01:07:04
to talk about that.

01:07:05
Yeah, yeah.

01:07:07
I mean, I cover--

01:07:08
I basically cover--

01:07:08
it's actually interesting.

01:07:09
It started as a

01:07:10
conversation on Twitter.

01:07:12
Somebody put a tweet that say this

01:07:14
other cloud provider

01:07:15
without mentioning a name

01:07:16
introduced a feature into their

01:07:18
managed Kubernetes.

01:07:19
And I replied saying, we had this

01:07:21
on GKE for the last two years.

01:07:24
It's a feature

01:07:25
called image streaming.

01:07:26
It's basically a feature to speed

01:07:27
up the image pool.

01:07:29
When you are trying to start a

01:07:31
pod, the image has to

01:07:33
be pulled very quickly.

01:07:34
And then the person replied to me

01:07:36
saying, well, you're

01:07:37
not doing a very good job

01:07:38
talking at it.

01:07:38
I was like, well,

01:07:39
challenge accepted.

01:07:41
Fair enough.

01:07:44
So I see you're not talking about

01:07:46
other cloud providers.

01:07:47
You're not naming

01:07:48
people or naming things.

01:07:50
It's like the person

01:07:53
you're not talking about.

01:07:55
What is it?

01:07:56
Not Voldemort.

01:07:57
Is it Voldemort?

01:07:58
I'm really bad at

01:07:59
Harry Potter stuff.

01:08:00
I'm really bad.

01:08:01
Maybe I am as bad as you are.

01:08:03
But yes.

01:08:04
OK.

01:08:04
Fair enough.

01:08:05
That's other cloud provider we're

01:08:06
not talking about.

01:08:07
All right.

01:08:08
The thing we're not talking about.

01:08:10
So OK.

01:08:11
Let's dive a little bit deeper.

01:08:13
I mean, it's a cloud podcast.

01:08:16
So when you build an application,

01:08:18
how would you start?

01:08:19
Where is the best start?

01:08:22
Obviously, it's GKE.

01:08:23
But--

01:08:25
Well, I mean,

01:08:25
actually, arguably, no.

01:08:27
So this is a very quick question.

01:08:29
I think that--

01:08:32
I mean,

01:08:32
specifically, if you're using--

01:08:33
I think that if you're

01:08:34
using any cloud provider,

01:08:36
going straight into Kubernetes is

01:08:38
probably going to be frustrating

01:08:39
because you will have to learn a

01:08:40
bunch of things.

01:08:42
And what have been happening

01:08:43
through the last couple of years

01:08:44
is that a lot of these cloud

01:08:45
providers are just offering you

01:08:47
a CAS, a container as a service

01:08:48
tool, Fargate on AWS.

01:08:52
I think it's called ACS, Azure

01:08:53
Container Services on Azure,

01:08:55
and Cloud Run on GCP.

01:08:56
So you can just

01:08:57
basically write code,

01:08:59
put it inside the container, and

01:09:00
then ship it to us.

01:09:01
And then we will give you an URL

01:09:02
and a certificate.

01:09:04
And we will scale it

01:09:05
up and down for you.

01:09:07
If you are just writing an app

01:09:08
that you needed to answer a web

01:09:10
request and scale up

01:09:12
and down on demand,

01:09:13
you don't need

01:09:14
Kubernetes for this.

01:09:16
Where things start

01:09:17
to be interesting

01:09:18
or where people start looking into

01:09:19
using Kubernetes,

01:09:22
specifically-- and of course,

01:09:23
we're talking here about Google

01:09:24
Cloud, so GKE--

01:09:25
is when you start having or start

01:09:28
requiring things

01:09:30
that this simple container service

01:09:31
platform doesn't give you.

01:09:33
And since we're in the age of AI,

01:09:34
we can talk about GPUs.

01:09:36
So if you need a GPU--

01:09:39
the promise of a serverless

01:09:40
platform is very fast scaling.

01:09:43
Very fast scaling and GPUs doesn't

01:09:45
really go hand in hand.

01:09:47
Just bringing up a Linux node and

01:09:49
installing Nvidia drivers

01:09:51
to have a GPU ready,

01:09:52
that takes 15 minutes.

01:09:53
I don't think anybody will be able

01:09:55
to sell a serverless

01:09:56
platform that

01:09:56
scales in 15 minutes.

01:09:58
It will be complicated, I guess.

01:10:01
It will be complicated, yes.

01:10:02
That's where people go

01:10:03
then into Kubernetes,

01:10:04
when you need those really

01:10:06
specific kind of--

01:10:07
and a more fine-tuning knobs that

01:10:09
you can turn on and off

01:10:11
and experiment to try things out.

01:10:14
This is what I like

01:10:15
to call the happy path.

01:10:16
The happy path is you start with

01:10:17
something simple.

01:10:18
And as you do in this case, get

01:10:19
more complicated.

01:10:20
You move to

01:10:20
something more complex.

01:10:22
Of course, that's not

01:10:22
how it always works.

01:10:23
And people usually just go head

01:10:26
first, dive into GKE.

01:10:29
I'm as big as Google.

01:10:32
I need-- Kubernetes right now.

01:10:34
Sure.

01:10:36
Knock yourself down, please.

01:10:38
Actually, managed Kubernetes makes

01:10:40
more money than container

01:10:41
service, technically.

01:10:42
So whatever.

01:10:43
So the container service, just for

01:10:46
people that may not know,

01:10:47
I think underneath is basically

01:10:49
also something like Kubernetes.

01:10:50
It's just like

01:10:51
you'll never see that.

01:10:53
It's operated by the cloud

01:10:54
provider, whatever it is.

01:10:56
And you basically just give them

01:10:57
an image, run it for me.

01:11:00
Exactly.

01:11:02
If people want to go dive into how

01:11:05
these things are built,

01:11:06
there is a project called Knative

01:11:08
that was also released by the

01:11:10
Kubernetes community

01:11:11
a few years ago.

01:11:12
And that's

01:11:12
typically what people use

01:11:14
to be able to give you this

01:11:15
serverless experience

01:11:16
in a container format.

01:11:18
So it's Kubernetes with

01:11:19
Knative, but everything

01:11:20
is managed by the cloud provider.

01:11:21
And as you said, we expose just

01:11:23
the interface that

01:11:24
allows you to say

01:11:25
container in, container out.

01:11:28
Fair.

01:11:30
So from a couple of years ago--

01:11:34
well, a couple of years

01:11:35
ago, about a decade ago,

01:11:36
I think now, people started really

01:11:38
going into first VMs

01:11:42
and doing all of that.

01:11:43
Then we thought, oh, VMs.

01:11:46
So it all started with physical

01:11:49
servers are s**t.

01:11:51
It's bad.

01:11:52
And it's so heavy.

01:11:54
So let's use virtual machines.

01:11:55
They're more lightweight.

01:11:56
And then we figure out, oh,

01:11:57
virtual machines

01:11:58
are still very heavy.

01:11:59
So let's do containers.

01:12:02
And a couple of

01:12:03
years ago, companies

01:12:04
started really buying into this

01:12:06
whole container thing.

01:12:08
I think it basically was when

01:12:10
Kubernetes got big.

01:12:12
I don't remember the original.

01:12:14
Google technology,

01:12:16
Kubernetes was basically

01:12:17
built after Bork was started.

01:12:19
Bork, yes.

01:12:19
Right.

01:12:20
Yes.

01:12:20
Bork is the piece.

01:12:21
It was probably Borked.

01:12:23
And that's--

01:12:24
Yes.

01:12:25
Yes.

01:12:27
And we saw this big uptake in

01:12:32
migrations to the cloud,

01:12:33
and specifically

01:12:34
container platforms.

01:12:35
Do you think it has slowed down?

01:12:38
Do you think it's

01:12:38
still on the rise?

01:12:40
Is it pretty much steady?

01:12:43
How do you see that right now?

01:12:45
It's a very good question.

01:12:46
I think that there

01:12:48
are multiple ways

01:12:48
you can answer that question.

01:12:49
I think that people

01:12:52
moving to containers

01:12:55
is something that is probably

01:12:56
still happening.

01:12:57
I think that's

01:12:58
what's happening probably

01:12:59
is that we don't

01:12:59
talk about it as much.

01:13:01
Because-- so the way I

01:13:02
like to describe this

01:13:02
is because Kubernetes as a

01:13:04
technology is 10 years old,

01:13:05
it's going to be 10 years old in

01:13:06
June, by the way.

01:13:07
So June, I think,

01:13:08
5th or something.

01:13:13
June 5, 2014 was

01:13:15
the first pull request

01:13:16
that was pushed, that pushed the

01:13:17
first version of Kubernetes.

01:13:20
It's going to be a commodity.

01:13:21
It's becoming a commodity.

01:13:23
It's becoming something that

01:13:24
people don't even

01:13:25
have to think about.

01:13:27
And even cloud providers are

01:13:28
making it such a way

01:13:30
that they give you the experience

01:13:31
of Kubernetes, which

01:13:33
is essentially the API.

01:13:35
And the API of

01:13:36
Kubernetes itself is pretty cool.

01:13:37
It's a really nice way of

01:13:38
expressing intent.

01:13:40
As a developer, you just say, this

01:13:41
is how my application

01:13:41
looks like.

01:13:42
Run it for me.

01:13:43
I don't care.

01:13:45
And so on the other

01:13:47
hand, also this is--

01:13:48
also interesting is a lot of

01:13:51
programming language

01:13:54
frameworks started building into

01:13:57
the framework ways of going

01:13:59
from code to

01:14:00
containers without Docker files.

01:14:02
And you are a Java

01:14:02
developer, so you

01:14:03
know what I'm talking about.

01:14:05
Like a Jib, you can just import

01:14:07
the Jib plug-in in Maven.

01:14:11
And then you just run your Maven,

01:14:13
and then blah, you

01:14:14
have a container.

01:14:15
So you don't have to

01:14:15
think about Docker files.

01:14:16
You don't really have to worry

01:14:17
about them too much.

01:14:18
You don't even have to learn them.

01:14:19
And so I think that the

01:14:21
conversation is more now

01:14:22
about cost optimization,

01:14:25
optimizing, bin packing,

01:14:26
rather than the simple, oh, I want

01:14:28
to move from a VM to a container.

01:14:32
So the conversation

01:14:33
shifts in somewhere else

01:14:34
because the technology itself is

01:14:35
becoming more mainstream,

01:14:37
I guess.

01:14:39
So a couple of years ago, people

01:14:42
asked me about Docker,

01:14:43
just because you mentioned Docker.

01:14:45
And they asked me what

01:14:47
I think about Docker.

01:14:48
And I said, well, it's the best

01:14:49
tool we have right now.

01:14:50
I hope it's not going to stick.

01:14:53
That was probably a

01:14:54
very mean way of saying,

01:14:56
I think the technology, the idea

01:14:58
of containerization

01:15:00
and the container images is good.

01:15:02
I don't think

01:15:02
Docker is the best way.

01:15:03
And you said a lot of tools

01:15:05
actually started building

01:15:07
their own kinds of interface.

01:15:08
They all still use Docker or any

01:15:11
of the other imaging tools

01:15:14
underneath, but they're trying to

01:15:17
hide all of that from you.

01:15:19
Yeah.

01:15:19
And you don't even need Docker in

01:15:21
this case, right?

01:15:23
I think just very

01:15:24
quickly, I'm going

01:15:24
to do a shameless plug here.

01:15:26
We have an episode on the podcast

01:15:27
where we interviewed somebody who

01:15:29
is one of the core maintainers

01:15:31
of ContainerD.

01:15:32
And the episode was

01:15:33
not about ContainerD.

01:15:34
It was really about the

01:15:34
history of containers.

01:15:35
And I think it's very important to

01:15:37
go listen to the episode

01:15:37
because we talked about the

01:15:39
evolution from Docker

01:15:41
initially to the Open Container

01:15:43
Initiative, the OCI,

01:15:45
which is actually a

01:15:46
standardization part,

01:15:48
to ContainerD, to all the

01:15:50
container runtimes

01:15:51
that exist on the market today.

01:15:52
And I think through that history,

01:15:53
you will be able to understand

01:15:55
what you're exactly talking about.

01:15:56
We don't need Docker anymore.

01:15:58
You can build containers without

01:15:59
even touching Docker.

01:16:03
Because I don't really

01:16:05
like Docker personally.

01:16:06
That's just--

01:16:08
So everybody has--

01:16:09
I think we're not the

01:16:10
only ones, to be honest.

01:16:12
So what I wanted to

01:16:14
direct my question to

01:16:16
is that you also said you have the

01:16:18
things like Fargate

01:16:21
or all of those

01:16:22
serverless technologies

01:16:23
that underneath use Kubernetes.

01:16:25
But it's hidden from the user.

01:16:27
Is that the same thing?

01:16:30
I mean, yes, in the sense that

01:16:32
yes, because Kubernetes is

01:16:33
a sign, it's becoming

01:16:34
a commodity that people

01:16:35
shouldn't actually--

01:16:37
I mean, to be honest

01:16:38
with you, and probably this

01:16:39
is going to be--

01:16:40
I mean, probably people would be

01:16:42
shocked to hear me say this.

01:16:44
I think Kubernetes should probably

01:16:46
have never gotten into--

01:16:48
I mean, the fact that it became a

01:16:50
super popular tool

01:16:51
is a good thing, because it

01:16:52
attracted a lot of interest

01:16:53
and a lot of investments.

01:16:55
I do not think it's something that

01:16:57
people should learn.

01:16:58
It's a platform.

01:16:59
It's something you

01:16:59
can build on top of.

01:17:02
It's like-- I mean, you need to

01:17:04
run an application,

01:17:06
go run it on top of a

01:17:07
container as a service.

01:17:08
Do not learn Kubernetes.

01:17:09
You don't have to, right?

01:17:12
Like, I can see how it's our put

01:17:13
it once in a tweet, which

01:17:14
is basically, Kubernetes is a

01:17:15
platform to build platforms.

01:17:17
That's what it is.

01:17:19
That makes sense.

01:17:20
You don't even know--

01:17:21
you need to

01:17:22
understand how it works.

01:17:24
And I think that

01:17:25
makes perfect sense,

01:17:26
because from--

01:17:28
how do you say that?

01:17:30
From a user's perspective, the

01:17:33
whole idea of Kubernetes

01:17:34
was to abstract away

01:17:36
whatever you run on.

01:17:38
But now, there's so

01:17:39
many APIs in Kubernetes

01:17:40
that abstract away

01:17:42
Kubernetes that it's basically

01:17:44
possible to do whatever.

01:17:45
And I think Microsoft

01:17:46
did a really interesting

01:17:48
implementation on top

01:17:50
of Hyper-V, which uses--

01:17:53
micro VMs, whatever

01:17:54
they call those things.

01:17:55
Oh, yes, yes, yes, yes, yes, yes.

01:17:56
The shared kernel kind of stuff,

01:17:58
which is kind of the same idea

01:17:59
as a container.

01:18:01
Yeah, I think it's

01:18:01
based on Kata Containers.

01:18:02
I know what you're talking about.

01:18:04
But it's

01:18:04
interesting, because they still

01:18:06
put the Kubernetes APIs on top, or

01:18:08
the Kubernetes implementation

01:18:09
on top, making it look exactly

01:18:11
like a Linux container, which

01:18:13
is really cool.

01:18:16
Anyway, what do you think is--

01:18:20
what will be the next

01:18:21
big thing for the cloud?

01:18:22
What is the trend

01:18:24
that is upcoming?

01:18:24
You already mentioned AI, so that

01:18:26
is out of the box.

01:18:28
That's a very big one, right?

01:18:31
It's actually

01:18:32
funny, because I am--

01:18:33
so I'm in Berlin this week.

01:18:34
I was chatting with some--

01:18:36
we are here for

01:18:36
conference, and we were chatting

01:18:38
with some of the colleagues.

01:18:39
And the joke that we were-- the

01:18:42
joke I was making was,

01:18:43
next time I go on stage

01:18:44
into a big conference,

01:18:46
if there are people talking about

01:18:47
AI at the same conference,

01:18:48
I will go on stage and go, OK,

01:18:51
people talked about AI.

01:18:52
Now let's talk about the things

01:18:53
that actually matter.

01:18:55
Let's talk about the thing that

01:18:56
people are using and making money

01:18:57
from.

01:19:00
Let's stop

01:19:00
wishful thinking, right?

01:19:03
I think-- so OK,

01:19:05
Kubernetes for AI is big.

01:19:07
That's going to be around.

01:19:08
AI is not going to disappear.

01:19:10
It's going to be big.

01:19:11
I think we're in the phase where

01:19:12
we're discovering

01:19:13
what people can do with it.

01:19:15
So I think it's a super exciting

01:19:16
time to be alive, I think,

01:19:17
in my opinion.

01:19:18
There's like a shift in our field

01:19:21
that lots of people

01:19:22
don't get to experience.

01:19:23
I think the last time such a shift

01:19:25
happened in our field

01:19:26
was people moving from

01:19:27
mainframes to pizza box

01:19:29
servers.

01:19:30
So-- and we're living through--

01:19:33
You had to mention the mainframe.

01:19:35
Of course.

01:19:36
I'm going to live

01:19:36
through the same thing.

01:19:37
So anyway, I think that that's

01:19:39
what's going to happen.

01:19:40
So the security

01:19:41
remains a big problem

01:19:43
across the board for everything.

01:19:45
Access, security, management,

01:19:46
identity, software security.

01:19:48
You're a developer.

01:19:49
You know what I'm talking about.

01:19:50
People pulling random dependencies

01:19:52
from the internet

01:19:53
without knowing where

01:19:53
they're coming from.

01:19:55
People pulling

01:19:55
containers from Docker Hub

01:19:57
without knowing who built them or

01:19:58
how they were built.

01:20:00
Zero ways of establishing trust,

01:20:02
like all that stuff.

01:20:02
So that's going to remain--

01:20:07
not a problem, I

01:20:07
would say, but it's

01:20:08
going to remain a theme that we're

01:20:10
going to hear about

01:20:10
over and over again.

01:20:11
And we have to

01:20:12
solve this eventually.

01:20:15
I think the other thing would be

01:20:17
just basically cost saving,

01:20:21
because we live in an interesting

01:20:22
world where everybody

01:20:23
cares about cost saving.

01:20:24
So cost optimization, bin packing,

01:20:26
making sure you get

01:20:26
the most out of your buck

01:20:27
that you're paying to

01:20:28
your cloud provider.

01:20:29
And I think that the

01:20:32
cloud native ecosystem

01:20:34
enabled a lot of people to go do

01:20:36
some super niche solutions.

01:20:38
And I think we're

01:20:39
going to get to a stage now

01:20:40
where all these companies doing

01:20:42
super niche solutions

01:20:43
will be filtered out in a way that

01:20:45
only those that have

01:20:47
really, really interesting things

01:20:48
that solve real problems,

01:20:50
not made up problems,

01:20:51
will remain on the markets.

01:20:53
So I think that that's--

01:20:54
That makes sense.

01:20:55
Yeah, that makes sense.

01:20:56
Only the companies take that

01:20:58
really have a solution that

01:21:01
solves something and

01:21:02
not just being around.

01:21:06
Which this is how it

01:21:07
comes in 2023, at least.

01:21:09
Yeah, and I think

01:21:10
that ethics will also

01:21:12
play a big part in the whole AI idea.

01:21:16
Well, in the next decade of AI, I

01:21:19
think we need to be

01:21:19
really careful what we do.

01:21:21
The technology makes

01:21:23
big steps, but we also

01:21:25
see all of the downsides already.

01:21:28
The main industry that is always

01:21:30
on the forefront

01:21:31
of every new

01:21:31
technology already is there,

01:21:35
and they're misusing it for a lot

01:21:37
of really stupid things.

01:21:39
But it is what it is.

01:21:42
Because we're already

01:21:42
running out of time.

01:21:43
20 Minutes is so super short.

01:21:45
Your favorite stack right now?

01:21:49
Yeah, you asked me that question.

01:21:51
Still unchanged.

01:21:52
Go, as a backend

01:21:54
technology, I don't

01:21:55
do a lot of front-end, so I don't

01:21:57
have a favorite stack

01:21:58
in that space.

01:22:00
Mac for development, Visual Studio

01:22:02
Code, VSS for coding.

01:22:04
I started doubling into IntelliJ,

01:22:06
but recently I kind of like it,

01:22:07
actually.

01:22:09
Because I'm not a Java developer,

01:22:10
so I don't really had--

01:22:11
I never had a need for it, but I'm

01:22:13
just experimenting.

01:22:15
And so Go for the backend.

01:22:18
I think it's just backend.

01:22:19
I only do backend, so Go.

01:22:21
Yes, probably I should have--

01:22:22
That makes sense.

01:22:24
Go and me is a

01:22:25
love-hate relationship.

01:22:26
I wouldn't say Go is

01:22:27
a perfect language,

01:22:29
but it's super

01:22:30
efficient for microservices.

01:22:33
When you write microservices, all

01:22:35
of the integration

01:22:36
of Let's Encrypt or the ACME

01:22:38
protocol, all that kind of stuff,

01:22:39
it's literally you just dump it

01:22:41
down and it works.

01:22:43
And that was a first for me,

01:22:47
coming from the Java world.

01:22:49
Making it something which is--

01:22:51
I wouldn't say--

01:22:53
A lot of people claim

01:22:54
that Java is very verbose.

01:22:58
I don't think that.

01:22:59
I think Java was always meant to

01:23:01
be more readable than writeable,

01:23:03
which is, from my

01:23:04
perspective, a good thing.

01:23:05
And I sometimes think Go did some

01:23:08
things, at least, wrong.

01:23:11
But it's much more

01:23:13
complicated, because Java

01:23:14
is coming from a very

01:23:15
different direction.

01:23:17
If you want to write

01:23:18
something really small,

01:23:19
and you hinted at the frameworks

01:23:21
like Quarkus and stuff,

01:23:23
but Go all just has that.

01:23:25
They build it with the idea of the

01:23:27
standard library

01:23:28
should be pretty much everything

01:23:30
you need for a microservice.

01:23:32
And it basically is.

01:23:34
Exactly.

01:23:35
All right.

01:23:37
We're recording that

01:23:38
the week before KubeCon.

01:23:40
KubeCon Europe, will

01:23:41
I see you next week?

01:23:43
Yes, I'm going to be at KubeCon.

01:23:44
I'm going to speak at Cloud Data

01:23:46
Rejects on the weekend,

01:23:46
and I'll be at

01:23:47
KubeCon the whole week.

01:23:48
All right.

01:23:48
So if that comes out next Friday,

01:23:50
I think, and you hear that,

01:23:52
and you're still at

01:23:53
KubeCon, come see us.

01:23:56
Yes.

01:23:57
I'm going to be at the

01:23:58
booth a lot of times,

01:23:58
but you will be able to see me.

01:24:00
I am a tall, brown

01:24:01
person with curly hair.

01:24:03
So I don't know how many people

01:24:04
like me will be there,

01:24:05
but and I speak very loud.

01:24:07
So you'll be able to

01:24:08
both hear me and see me.

01:24:10
That's fair.

01:24:11
All right.

01:24:11
Thank you, Abdel.

01:24:12
Thank you for being here.

01:24:14
I really appreciated the session.

01:24:16
It was a good chat.

01:24:17
I love that.

01:24:19
Thank you very much.

01:24:19
Thanks for having me, Chris.

01:24:21
All right.

01:24:22
Next week, next episode, and I

01:24:25
hope you're listening in again.

01:24:26
Thank you very much.

01:24:28
The Cloud Commute Podcast is

01:24:30
sponsored by simplyblock, your

01:24:31
own elastic block storage

01:24:32
engine for the cloud.

01:24:34
Get higher IOPS and low

01:24:35
predictable latency while bringing

01:24:36
down your total cost of ownership.

01:24:39
www.simplyblock.io

Kubernetes is the platform to build platforms, like Cloud Run or Fargate - Abdelfettah Sghiouar from Google

Related Episodes

Testing Methodology for the Cloud | Meryem Zid

Data Handling in Manufacturing | Jeremy Theocharis

JetBrains: Developer Tools for the Cloud Age | Hadi Hariri

Cloud Frontier

Cloud Commute