#222 April 8, 2024

KubeCon EU 2024

Hosts: Abdel Sghiouar, Kaslin Fields, Mofi Rahman

KubeCon EU 2024 was the largest KubeCon yet! Explore the trends and learnings from the event through interviews with attendees.


And additional Guest Host, Mofi Rahman.

Do you have something cool to share? Some questions? Let us know:

News of the week

Links from the post-interview chat

KASLIN FIELDS: Hello and welcome to the "Kubernetes Podcast" from Google. I'm your host, Kaslin Fields.

ABDEL SGHIOUAR: And I'm Abdel Sghiouar.


This week, we bring you a behind-the-scenes look at KubeCon Europe in Paris. We talk to attendees about their experience at the events.

KASLIN FIELDS: And by "we," we also mean our quickly becoming regular guest host, Mofi Rahman. So thank you very much, Mofi. But first, let's get to the news.


ABDEL SGHIOUAR: Kubernetes code source base adopted Go workspaces. Since its inception, Kubernetes relied on the GOPATH concept and some custom magic to allow maintainers to manage dependencies. But there were some pain points, especially when it comes to the customization that was in place. In a blog post, Tim Hockin detailed how moving to Go workspaces as a way to organize code and dependencies makes life easier.

KASLIN FIELDS: Fermyon, the serverless WebAssembly company, announced it applied to donate SpinKube to the CNCF as a sandbox project. SpinKube is an open-source framework for running Wasm workloads on Kubernetes. It combines an operator, the containerd-shim, and a runtime manager.

ABDEL SGHIOUAR: Istio announced the beta of ambient mesh starting version 1.22. The new Istio architecture brings Layer 4 sidecarless functionalities into the mesh using a dedicated node proxy. Sidecars will remain first-class citizen and will be used for Layer 7 functionalities. The new release includes secure overlay, metrics, Layer 4 authorization policies, and integration between the node and workload proxies.

KASLIN FIELDS: KubeCon EU 2024 was the largest KubeCon yet, with over 12,000 attendees in person in Paris. The CNCF's highlights blog is linked in the show notes for more information.

ABDEL SGHIOUAR: The CNCF has created the Kubestronauts program to recognize and partner with members of the cloud native community who have passed all five of the CNCF Kubernetes-related certifications. Kubestronauts will act as a cloud native education ambassador and receive benefits such as discounts to CNCF events and certifications.

KASLIN FIELDS: At KubeCon EU, the CNCF announced a new partnership with Udemy. Through this new partnership, developers will be able to leverage Udemy's technical course content, including a comprehensive library of cloud native content and customized learning paths, to prepare for many of the CNCF certification exams, including the Certified Kubernetes Administrator.

ABDEL SGHIOUAR: The CNCF, in collaboration with the United Nations, hosted its first-ever hackathon at KubeCon EU, CloudNativeHacks. The hackathon's focus was on advancing the delivery of the UN Sustainable Development Goals. Congratulations to Team Urban Unity for the first place, Team Forrester for the second place, and Team Potato for the third place. Each of the projects tackled at least two UN sustainability goals. The winning project created a map where urban planners can indicate upcoming projects, and locals can add their comments to those projects.

KASLIN FIELDS: The call for proposals for KubeCon North America 2024 is open. The event will take place in Salt Lake City, Utah, from November 12 to November 15. The CFP will close on June 9, 2024.

ABDEL SGHIOUAR: The CNCF Cloud Native Glossary is now available in Japanese. The glossary is an effort to explain cloud native concepts in approachable ways and to make them accessible in a variety of languages. The resource is now available in 14 languages.

KASLIN FIELDS: Kubernetes is turning 10. The CNCF is planning a big KuberTENes birthday bash for June 6, to be hosted in person in Silicon Valley and at various locations around the globe. If you're artistically inclined, a competition is open now to design the Kubernetes 10th anniversary logo. Submissions are open until Tuesday, April 30. And the winning design will be announced in June. And that's the news.


I'm Kaslin Fields of the "Kubernetes Podcast" from Google, and I'm here at KubeCon EU 2024 in Paris. And I'm speaking with--

BENJAMIN KOLTERMANN: Benjamin Koltermann from Germany. Hi, it's nice to be on the podcast. I do cloud and Kubernetes security at AVOLENS, a company I self-founded. And so I'm from Germany.

KASLIN FIELDS: What were you hoping to get out of the conference?

BENJAMIN KOLTERMANN: So I'm especially looking at confidential computing features. I were here to have a look at the Confidential Containers project and also at the confidential features NVIDIA just released. It's very interesting to see.

KASLIN FIELDS: What are the big trends you're seeing?

BENJAMIN KOLTERMANN: Because I was looking especially at those confidential features or confidential computing features, I see that these projects get a lot of interest. So the talks-- there were many people, and some also, the projects are developing rapidly. So it's nice to see that there is coming something new. Yeah.

KASLIN FIELDS: What is your favorite thing you've learned so far?

BENJAMIN KOLTERMANN: There's a lot of new AI things in the cloud native world. Basically, every third or fourth vendor somewhere has AI in its products. I mean, I didn't see it until KubeCon, so it's kind of interesting to see. And I love to see new AI use cases.

KASLIN FIELDS: What would you like to see at future KubeCons?

BENJAMIN KOLTERMANN: I hope in the future to see more security things because I love doing security. So I hope that there will be some smart folks figuring out some new cool or starting some new cool security projects.

KASLIN FIELDS: Awesome. Thank you very much.


KASLIN FIELDS: And I am speaking with?

GABRIELE BARTOLINI: Gabriele Bartolini.

KASLIN FIELDS: Would you like to say a little bit about yourself?

GABRIELE BARTOLINI: Aw, thanks. So I'm vice president of Cloud Native EDB. So you're probably wondering what it means. But my mission is to promote Kubernetes adoption with databases in general. So I'm a DoK Ambassador, a Postgres contributor. And this has been my mission in the last five years.

KASLIN FIELDS: What were you hoping to get out of the event?

GABRIELE BARTOLINI: So my interest this year-- you know, everyone is talking about AI. But my interest has always been in databases. And there's a lot of relationship between these two disciplines. So my job is now to scale databases so that they can be used for AI purposes. And of course, I'm talking about Postgres databases, you know?

And what I've learned-- that there's a lot of interest now. People were afraid of running databases in Kubernetes. But I see that now it's not the case anymore. And I think we're working very well with the tech storage in Kubernetes. And now I hope to get more storage vendors involved.

KASLIN FIELDS: What big trends are you seeing at KubeCon?

GABRIELE BARTOLINI: As I said, AI has been the major thing that people have been talking about. And databases-- you know, you need data to run AI. So I talked, for example, on DoK. On Tuesday, the room was full. It was a nice surprise. And yeah, I hope to see this interest grow even more in the future.

KASLIN FIELDS: AI is a form of stateful workload.


What is your favorite thing you've learned so far?

GABRIELE BARTOLINI: So I'm kind of an old guy. So I owe everything to open source. And what I'd like to say is, technologies like Kubernetes and Postgres-- first, Linux. Linux was my first love. Then Postgres came. And Kubernetes has been my passion for the last five years. So so far, this is what I'd like to say I've learned.

And then what I learn every time I come here to KubeCon is the focus on people, on diversity. It's so refreshing to see all this creativity brought by all these people getting together and have fun. Like today, for example, we had the jam organized by Kubeshop. It was amazing. It was so nice. I hope this is going to happen again, so it's good.

KASLIN FIELDS: What would you like to see at KubeCon in the future?

GABRIELE BARTOLINI: More jams. No, I'm just kidding, but--


GABRIELE BARTOLINI: --you know, I would like to see the data on Kubernetes community grow even more. Again, people are afraid to run databases in Kubernetes. I'm a database guy who learned Kubernetes. And in my opinion, there's no better way to run a database than in Kubernetes. So that's what I'd like to say. And I would like everyone to try, for example, our CloudNativePG operator, which is open source and which we want to reapply again for the sandbox to donate it to the CNCF and to everyone.

KASLIN FIELDS: Awesome. Thank you very much.

GABRIELE BARTOLINI: Thank you, Kaslin.

KASLIN FIELDS: [LAUGHS] And I am here with--

LACHLAN EVENSON: Hello, everybody. I'm Lachlan Evenson. I'm a Product Manager at Microsoft, work on Kubernetes. And I've been in the Kubernetes community for about 10 years. So it's great to be here, and I'm very excited about the 10-year anniversary of Kubernetes.

KASLIN FIELDS: What were you hoping to get out of the conference?

LACHLAN EVENSON: Connections. I really come to the conference to make connections. I find those connections to be really valuable and durable so that, when I'm working in the community, I can reach out to folks and they know who I am. They know the face behind the name. So I really look forward to making meaningful connections to people I've known for many years but also newcomers to the community, because I love building that next generation of folks that are coming up in the community.

KASLIN FIELDS: What big trends are you seeing?

LACHLAN EVENSON: I think there's three big trends. One, obviously, AI. I think that's a big trend. So the industry is changing. I see a rapid innovation. It kind of takes me back to the beginning of the container journey, that massive explosion of ideas around AI. And folks are figuring out how to apply AI and change their businesses.

It's super exciting, and I'm really excited to be a part of that. And I think in the next six months, I'll be really excited to see what folks do with AI. So I think there's that early nascent excitement. But we've got this real energy and this great community that can really leverage AI. So I think that's one.

A couple of others are FinOps. I think I've seen the rise of FinOps in the context of cost optimization, so really getting the most out of your clusters, running them at much higher utilization and scaling much quicker. So I think cost optimization and FinOps is another one.

And then finally is sustainability. So I think there are a lot more folks in the world that are conscious about their impact and utilizing hardware correctly. So I think that goes hand in hand with cost optimization. But it's great to see folks saying, hey, what's the impact of the workloads I'm running? And how can I utilize tools in the cloud native ecosystem like Kubernetes to actually improve my sustainability and my footprint? So I'd say those are the three things.

KASLIN FIELDS: What is your favorite thing you've learned so far?

LACHLAN EVENSON: The application of AI. So even though it's out there, and it feels very nebulous-- for example, I had a conversation the other day where somebody was using-- they had a cost optimization situation where they wanted to get more out of their cluster. They were taking tools like KEDA and Karpenter and doing this really cool autoscaling. And they were actually putting some AI in there to make some really interesting decisions about when they scale, how they scale, what their nodes are going to do, and preemptive scaling. So I think we're going to see a lot in that area.

KASLIN FIELDS: And what would you like to see at future KubeCons?

LACHLAN EVENSON: Yeah, I think I've been very reflective about the last 10 years. And I went back to how my Kubernetes journey started. And there was this moment when I first used Kubernetes where I had that kind of wow, like the spark-joy kind of moment. And I saw that first workload go to a running state, and I connected to it. And it just brought me so much. I connected with Kubernetes.

I think folks are coming, and they have no experience with Kubernetes. And they're looking to get that first experience. So I don't know if it's Mean Time To Kubernetes, MTTK, but just a really simple experience where somebody can spin up Kubernetes and witness the power of it, because I think that first moment where you connect with the platform really underpins the rest of your journey and what you feel you're empowered to do-- so I think creating small forums where folks can spin up Kubernetes and really see the power of it so that they can just get that wow factor that I had when I first used Kubernetes.

KASLIN FIELDS: Excellent. Thank you very much. And I am speaking with?

TABITHA SABLE: I'm Tabitha Sable. I am one of the co-chairs of Kubernetes SIG Security. And I am delighted to be chatting with you here in the glow of the Kubernetes SIG Security goose unofficial logo at our booth.

KASLIN FIELDS: What were you hoping to get out of the conference?

TABITHA SABLE: I was coming into this KubeCon with a panel discussion about community building through the creation of inclusive spaces in meetings and a maintainer track talk presenting the work that we've been doing in SIG Security, which has a lot of interesting technical content and also a lot of really important community-building content.

So I was hoping to be able to talk with other folks who are interested in increasing the amount of teamwork that folks have within-- really within the world, but also within Kubernetes. And I've had really good conversations as a result of that. So I've been really pleased with that.

KASLIN FIELDS: What big trends are you seeing?

TABITHA SABLE: One big trend that I have been really heartened by is the continual progress of security tooling and security features within Kubernetes. In something that is as large and mature as Kubernetes is, it can be very, very difficult to make changes. And folks are doing a good job of continuing to find the places where improvements can be made without blowing up everybody's compatibility.

So I'm really excited about the referential RBAC work that is happening. I have been really excited with some of the work that is happening in SIG Security, like with the continued expansion of the Hardening Guide and, in general, just seeing that folks are continuing to find areas for improvement.

KASLIN FIELDS: What is your favorite thing you've learned so far?

TABITHA SABLE: I think my favorite thing that I have learned so far, because I love silly technical tricks and I love tools that have some spooky implications to them as well-- and I'm sorry I don't remember the name of the tool, but someone's tool that they were releasing whereupon they hook the libc functions so that you can do network debugging from a laptop on a pod that is running in a cluster.

Or ultimately for development, you can run the code on your laptop, but it feels like it's in the cluster. Like, the network calls are emitted out through a pod. And I see where that will make developers' lives so much easier. And also, there are so many cool attack possibilities in that kind of model as well. So I'm excited to go back home and do some more reading about that.

KASLIN FIELDS: What would you like to see at KubeCon in the future?

TABITHA SABLE: Actually, one thing I would like to see is a continuation of the SIG booths at the Project Pavilion and the location of the Project Pavilion sort of near the vendor hall but not directly in the middle of it. It has been a really good experience to have folks be able to stop by here, talk to colleagues, talk to old friends, make new friends, as a bit of a respite from the more loud hubbub of the vendor hall. So I would love to see that continue.

KASLIN FIELDS: Awesome. Thank you very much. [LAUGHS]

TABITHA SABLE: Thank you so much for your time.


MOFI RAHMAN: We are here at KubeCon EU 2024, and suddenly, a wild goose appears.

IAN COLDWATER: Hi. My name is Ian Coldwater. I am Co-chair of Kubernetes SIG Security and about to start as Principal Security Architect at Docker.

MOFI RAHMAN: Ian, what were you hoping to get out of this conference?

IAN COLDWATER: Well, I am here at KubeCon to connect with the community, get things done with my SIG and my leads, and generally be excited about the energy of bringing people together and getting things done.

MOFI RAHMAN: And I'm going to ask you this knowing very well what the answer might be. What big trends are you seeing at this KubeCon?

IAN COLDWATER: I am seeing a lot of people appeal to executives with what they think is the appeal of AI and a lot of rank-and-file practitioners perhaps being a little more skeptical about that.

MOFI RAHMAN: Fantastic. Yeah, that's why I prefaced the question of knowing what the answer might be. What is your favorite thing that you have learned so far?

IAN COLDWATER: I have gotten to learn actually some really interesting technical things this time around. There have been a couple of talks that I've seen about people doing really cool things that maybe are not 100% intended, I think, by the original makers of that technology. And that's been really exciting. I really enjoyed the privilege escalation talk by Iain Smart and that talks about how to use DNS in unexpected ways. And that's been really cool. I really enjoyed that.

MOFI RAHMAN: And what would you like to see at KubeCon in the future, either change or different or more of?

IAN COLDWATER: I'm really excited about the community collaboration aspects of KubeCon, about everybody getting together, getting the work done, and just getting to kind of have a reunion after 10 whole years. And so I think I'm excited about seeing more of that happen. I hope that in the future, we all get lots of time and space to be able to get that work done. And I feel really good about that.

MOFI RAHMAN: And the final question I wanted to ask you is, at the 10-year mark of Kubernetes, as a community, as someone in the security space, what are you most proud of as a community that we have achieved to do in the last 10 years of Kubernetes?

IAN COLDWATER: We have come so far in the last 10 years. When Kubernetes first came out, there really wasn't any security to speak of in a lot of ways. You could execute commands as root with an unauthenticated curl call. And you just can't do that anymore. We've really made a lot of improvements through a lot of work for a lot of people over the years so that, now, the bar to being able to hack Kubernetes is much higher than it used to be.

And I love this both for the security of end users, and I love it, frankly, as a hacker. I love a challenge. I love leveling up. I love leveling all of us up. And I love that we get to make things harder for people who are attacking and also leveling up the education and knowledge for attackers, and also making things safer for end users without necessarily having to have that kind of education and knowledge. Like, it's easier for them to do things right now, and I feel great about that.

MOFI RAHMAN: Thank you so much, Ian, and until next KubeCon.

IAN COLDWATER: Thank you very much.

MOFI RAHMAN: And now we're talking to--

JAMES BLAIR: Hey, team. My name is James Blair. I work at Red Hat, but I'm also an etcd maintainer.

MOFI RAHMAN: What were you hoping to get out of this conference?

JAMES BLAIR: This conference-- well, there is a saying that goes something like, you come for the technology, but you stay for the people. And this is my fourth KubeCon now, actually. And for me, it's kind of been exactly that. Like, I came for the technology. I was excited about the technology. I still am. And the thing that draws me here every time to stay is the people. I get to see friends that I haven't seen for a long time, and I love it. Yeah.

MOFI RAHMAN: Yeah, that's very true. It's almost the same for me as well. What big trends are you seeing?

JAMES BLAIR: We cannot escape-- you know the word I'm going to say.


JAMES BLAIR: You know, AI-- everyone's talking about that. Obviously, the implications for our Kubernetes ecosystem around how we make accelerators available, like GPU, et cetera-- that has been consuming vast amounts of mindshare here at this particular KubeCon, for good reason.

Beyond that, it's actually a little interesting KubeCon. As a project maintainer, you end up spending so much time with people, you actually don't get to go to very many talks, which is actually a bit of a challenge. But interesting talks still continuing around WebAssembly and the role of WebAssembly within the ecosystem, particularly in relation to bringing runtimes for WebAssembly alongside existing runtimes for ordinary containers.

MOFI RAHMAN: Yeah, I also see a huge plugin architecture kind of developing around WebAssembly, which is also exciting to see. What is your favorite thing that you have learned from this conference?

JAMES BLAIR: From this conference? Well, one of the talks-- again, it's hard to get to very many as a project maintainer because you've got so many things going on. But I went to a pretty awesome talk looking at using CRIU, looking at checkpoint/restore to basically freeze or take a checkpoint of a process, maybe a process running in a container, and then maybe move that to another node and restore that running process.

So let's say you've got a process that's doing some stuff with files or network sockets, that sort of thing. Being able to kind of freeze that process in place, move it over to another node, and resume it-- pretty cool stuff. So that's been a highlight for me this KubeCon.

MOFI RAHMAN: Yeah, that also, I think, opens up a lot of use cases for even AI/ML and serving use cases where people want to-- because this resource is-- like, starting and restarting is pretty expensive. To be able to freeze and move-- yeah, that is pretty exciting.

JAMES BLAIR: So the particular use case for the talk was in relation to running Kubernetes on Spot. So if you've got Spot instances where you get maybe 30-seconds heads-up that you need to move your workload, being able to actually not just terminate and reschedule a workload, but to actually freeze it and restore it with all of its memory and everything-- it's very cool.

MOFI RAHMAN: Yeah, and I'll have to probably check that talk out. It seems really interesting. The last thing I'm going to ask you is, what would you like to see different or new in the future KubeCon iterations?

JAMES BLAIR: Oh, one thing I have really enjoyed this KubeCon-- and after four, it's the first time I've really seen it like this-- the Project Pavilion this year was pretty awesome. So for anyone that hasn't been to a KubeCon before, there is a CNCF Project Pavilion where you've got a whole bunch of stalls for each of the projects with maintainers working there staffing those.

And this year, we had a bunch of tables and a really nice area to actually collaborate and do some work right there within that Project Pavilion. We had tours running around the Pavilion. So I think for next KubeCon, I really hope we see a similar kind of focus and emphasis on the Project Pavilion, yeah.

MOFI RAHMAN: Yeah, excited to see that. And as the growth of KubeCon, space is always an issue. But one of the conversation I had with a couple of the folks from CNCF is that, every year, they're looking. And now, their choice pool is obviously limiting with the size of KubeCon. But they are taking great care and figuring out what spaces would be able to fit similar type of space that we got for this year. So that was pretty awesome.

JAMES BLAIR: Yeah, yeah.

MOFI RAHMAN: So thanks, James, for chatting with me in KubeCon, and until next time.

JAMES BLAIR: Hey, no worries. Thanks.

MOFI RAHMAN: And I am here with--

SREERAM VENKITESH: Sreeram Venkitesh. Hey, everyone. My name is Sreeram. I am from India. This is my first time attending KubeCon.

MOFI RAHMAN: So, Sreeram, what were you hoping to get out of the conference?

SREERAM VENKITESH: For me, one of the main takeaways from the conference was meeting everyone. I've been talking to a lot of folks through Slack and through Zoom calls. So meeting everyone in person and hanging out with them was one of the things that I was looking forward to.

MOFI RAHMAN: OK, so it sounds like you have been involved in the Kubernetes project in the open source capacity for a while now. How was that like? And how did that translate meeting all of these people for the first time in person?

SREERAM VENKITESH: Yeah, meeting people face to face and building that sort of personal connection with them sort of takes the edge away from just talking to them in Slack or Zoom calls. And it makes the work that you do really special because now you know the people. And the community is welcoming. So yeah, that's been one of the biggest rewards that I got from KubeCon.

MOFI RAHMAN: Fantastic. What big trends are you seeing?

SREERAM VENKITESH: Since this is the 10th year of Kubernetes, and there have been a lot of discussions about what cloud native is going to go to in its second decade and all the new projects that are coming up-- and I'm seeing a lot of trends related to AI and running AI and LLM workloads and things like that in the cloud native space. And I've also been seeing a lot of things about eBPF and the work that Isovalent is doing. Yeah, these are some areas that I would like to explore more.

MOFI RAHMAN: What would be the memory you cherish the most from this KubeCon?

SREERAM VENKITESH: Walking around the solution showcase and randomly bumping into people, like people like Tim Hockin and the folks from the "Kubernetes Podcast" by Google and getting to do such sessions with them. There are a lot of people at KubeCon, and it can get a bit overwhelming. But there are always folks around who you might know form different communities. And you can always bump into them, and then suddenly, your day just gets better.

MOFI RAHMAN: That's awesome. And the last question I'd like to ask you is, what would you like to see at KubeCon in the future?

SREERAM VENKITESH: First of all, I would like to see all the beautiful people I met at KubeCon this time again at future KubeCons. And yeah, the contributor summit and the SIG meet and greet has been really valuable for me because I get to have productive discussions about the work that is being done in Kubernetes and meet the different teams who are working on these things. And yeah, that is something that I'm looking forward to again.

MOFI RAHMAN: Awesome. Thanks for the chat, Sreeram, and hope to see you in the future KubeCon.

SREERAM VENKITESH: Thank you. Thanks for having me.


ABDEL SGHIOUAR: Hi. This is Abdel from the "Kubernetes Podcast" by Google. I am today at Cloud Native Rejekts, and we are with--

BENAZIR KHAN: Hi. I'm Benazir Khan. I'm a Community Program Manager at Microsoft. I'm so happy to be here. Thank you, Abdel.

ABDEL SGHIOUAR: Thank you. It's fantastic, I have to say. We're going to talk a little bit in details. Can you tell us what is Cloud Native Rejekts?

BENAZIR KHAN: So Cloud Native Rejekts is the B-side human-sized conference. I like to use all of these adjectives together because it's absolutely true. Ask any seasoned attendee or a first-time attendee. It's basically the small-scale conference of around 200 to 250 attendees that we put together right before KubeCon week.



ABDEL SGHIOUAR: So where did the idea came from? How did this all started?

BENAZIR KHAN: So I have to give complete credit to Chris Kuehl, whose brainchild was Rejekts. The idea was, years ago, when there were lots of KubeCons and everything happening, at some point, a lot of talks started getting rejected from KubeCon because not every great talk can make it to KubeCon.

And the idea started off with, How about we create this platform for rejected talks from KubeCon? which are still wonderful and still KubeCon worthy. But we create a separate platform for them. And they can actually still come around and give those KubeCon-worthy talks at a conference which is correctly titled and aptly titled Rejekts.

ABDEL SGHIOUAR: Yes. I like it. I like also the usage of the K instead of C, which is really nice.

BENAZIR KHAN: I'm so glad you caught on to that, yes.

ABDEL SGHIOUAR: Very clever branding, by the way.


ABDEL SGHIOUAR: So, I mean, I have to assume that, at this stage, with the size of KubeCon, there should probably be a Cloud Native Rejekts Rejekts because you get a lot of submissions that get rejected from KubeCon. So how do you deal with that as well, like if you are receiving more submissions than you are able to accept?

BENAZIR KHAN: Well, you know, the response to Rejekts has been so overwhelming over the years, and it's scaled so much in that sense that so many people now know Cloud Native Rejekts and they come and attend regularly-- that they have regularly started submitting to the CFP of Rejekts every time it's open and probably even prefer Rejekts to KubeCon many which ways because they know that in the two-track format for two days, at some point or the other, they will get to have closer discussions with focus groups and folks who are truly interested in their product or what they have to say or some of the projects that they're working on. So the upside is always like, it's all great, up, up, up, up, up.


BENAZIR KHAN: You know? Like, more and more popularity. I think one of the things that we're dealing with in terms of challenge is, definitely, there are so many submissions. And there are folks who get rejected from Rejekts.


So if your talk is rejected from Rejekts or has been in the past, then I think the only thing that we can say is, try again. [LAUGHS]


BENAZIR KHAN: You know, there's a good chance that when you're consistent and persistent with that, and you keep playing around with a lot of experimental topics, that something gets chosen.

ABDEL SGHIOUAR: Yes, I mean, you basically throw a bunch of pasta at the wall and see which one sticks.


ABDEL SGHIOUAR: And also, don't forget that there is two editions every year.


ABDEL SGHIOUAR: Is there going to be a third edition this year for India?

BENAZIR KHAN: So we have toyed with the idea. At this point, we're completely at capacity in terms of the team. And you know how much we take care of Rejekts on our own and everything. I do a lot of the heavy lifting, of course, for both EU and any editions, and I'm one person. [LAUGHS]

But we'd love to. I definitely want to say that it would be an amazing extension for us too, and we couldn't be more enthusiastic. But yes, we'll have to look into the finer details, if I can put it that way, because that's the absolute truth. So as much as we'd love to, we really have to see how it could possibly be done and come together.

ABDEL SGHIOUAR: Nice, yeah. I'm excited about this first year in India. So it's going to be interesting to see how the community-- I mean, there's quite a lot of people from India and Southeast Asia that participate in KubeCon, generally speaking, Cloud Native Rejekts. So yeah, we'll see how it goes. So we are here at this interesting spot called ESpot, pun not intended.


ABDEL SGHIOUAR: Can you describe to people who haven't made it this year what is this place we're at here?

BENAZIR KHAN: So ESpot is a fun gaming venue at the heart of Paris. And when I say heart of Paris, I'm not even misplacing it. This is bang in the center of Paris.

KASLIN FIELDS: Yes, I can confirm that.

BENAZIR KHAN: You're here. So it doesn't get more central than this, as we're right across the Louvre too. And it's at Rue de Rivoli. It's this fantastic venue that we found, which is actually like a gaming arena. And there are folks that are actually gaming and playing and stuff like that. And we have some spaces reserved for the conference.

So we're hoping that attendees get the best of all worlds where they actually arrive at our venue. They know that a lot of things to see in Paris are also close by in walking distance. They get to have lunch at some great bistros and cafes around the corner. And at the same time, at the venue, they also get to have the option of attending talks, networking, chatting with folks, partaking in the amazing barista service that we are offering too and on site on the venue and, yeah, just probably playing a game or two if they want to on their own time.

ABDEL SGHIOUAR: Nice, yeah. As we are sitting here-- of course, this is audio only, but there is a bunch of Playstation 5s around, and there are people gaming. There is a whole downstairs and upstairs with just PCs for LAN games, right? So it's a pretty-- I would say it's a pretty nerdy place.


ABDEL SGHIOUAR: Or pretty geeky or nerdy. Depends how do you want to describe it.

BENAZIR KHAN: We were hoping that there's something in it for everyone when they came. [LAUGHS]

ABDEL SGHIOUAR: I think a bunch of attendees will really appreciate being here. Nice.

BENAZIR KHAN: That was the idea.

ABDEL SGHIOUAR: Nice. So are you excited about KubeCon?


ABDEL SGHIOUAR: What are you looking forward for this year?

BENAZIR KHAN: At KubeCon, I'm definitely looking forward to, of course, being at the Microsoft booth. I will definitely be at the Azure booth on 20th. And I'm looking forward to have some conversations with folks there. And not just in terms of what the overview of things that happens in the Azure Arc-- that's something that I've always found those conversations fascinating from different perspectives, sometimes from end users, sometimes different customers and everything, and sometimes from people within and how they've seen it grow and mature over time and stuff like that.


BENAZIR KHAN: And also kind of talk to other folks around with maybe other bigger companies and smaller companies and see what they have in store and what are the projects that they're actively involved in and how they want to go ahead with it this year.

ABDEL SGHIOUAR: Awesome. Well, thank you very much for your time.

BENAZIR KHAN: Thank you so much. Thank you for having me. [LAUGHS]

ABDEL SGHIOUAR: Thank you. And I'm here with--

OLIVIA AL-JOUNDI: Olivia. I'm a senior product manager for Microsoft.

ABDEL SGHIOUAR: Awesome. This was your first KubeCon, you told me?

OLIVIA AL-JOUNDI: Yes, it was. I've been working at Microsoft for almost four years now. And finally, they were able to send me. [LAUGHS]

ABDEL SGHIOUAR: Awesome. And you choose one of the biggest to come to.

OLIVIA AL-JOUNDI: Yeah, exactly. I specifically chose Paris. [LAUGHS]

ABDEL SGHIOUAR: Nice. Oh, so you chose the city? OK.

OLIVIA AL-JOUNDI: Yeah, exactly.

ABDEL SGHIOUAR: OK, so how has it been?

OLIVIA AL-JOUNDI: It's been amazing, honestly. Like, there's so many elements to it. There's the booth sessions. There's walking around and meeting new partners and new customers. Meeting a lot of my team has been honestly great just because I usually work remote, so so many friendly faces that I've worked with for the past four years.

And then been able to see a lot of sessions. The keynotes were amazing. Lots of themes around like AI, GPU utilization, cost optimization. There's also themes around observability, but I think that's pretty common. And then, yeah, got some free coffee, some free smoothies. It's been great.

ABDEL SGHIOUAR: Nice. Yeah, so you touched on a bunch of themes there. I think AI have been a big theme this year. There was a lot of stuff around AI. The keynote was pretty much 50% AI.


ABDEL SGHIOUAR: Anything else that kind of piqued your interest that you didn't expect to hear about or any striking conversation you had?

OLIVIA AL-JOUNDI: I think the conversations that resonate the most with me are always with my customers or prospective customers and just kind of talking about how they're using Kubernetes, what open source tools they're using, and how we can make that into a more simplified add-on that's managed by Kubernetes. And then my product, which is Azure Linux-- just kind of trying to hear a little bit more about that experience and feedback there.

ABDEL SGHIOUAR: Nice, nice. So you told me you work on Azure Linux.


ABDEL SGHIOUAR: Right? How does that fit in this whole cloud native ecosystem?

OLIVIA AL-JOUNDI: Yeah so we launched Azure Linux at Microsoft about four years ago. It used to be called Mariner. And internally, we have Azure Linux as containers. But then we also have Azure Linux as the AKS node on AKS.


OLIVIA AL-JOUNDI: And that was released externally about a year ago. So that growth has been really great. We have thousands of enterprise customers since then. So just kind of walking around, seeing if people know about it, seeing what partner solutions are-- like, Palo Alto we have support for, Datadog, et cetera. Just kind of connecting with those people have been really fun.



ABDEL SGHIOUAR: Well, thank you very much.

OLIVIA AL-JOUNDI: Yeah, thank you so much. It was great chatting.


KASLIN FIELDS: Thank you so much to everyone that we interviewed at KubeCon. It was an awesome time. And here to talk with me about it, I actually have Mofi. Welcome, Mofi. [LAUGHS]

MOFI RAHMAN: Hi, Kaslin. Glad to be back again.

KASLIN FIELDS: [LAUGHS] Yeah. I said in the opening segment, which you haven't heard yet-- but I said in the opening segment that you're quickly becoming a regular guest host. [LAUGHS]

MOFI RAHMAN: Yeah, the podcast is always a good time. Again, I have been a regular listener for a long time. And then after joining Google, I've been even more involved with the team. So yeah, it's been really awesome to see the community grow as well as the podcast itself grow. And this KubeCon was the biggest of the lot after 10 years. So it's a clear sign that the community itself is actually in the upward trajectory of growth. So it's awesome.

KASLIN FIELDS: Yeah, and we really appreciate that you were up for helping us out with the interviews this time. It was such a big event, and there's so much going on. It was really hard to find the time and the space to do all of the interviews. So much appreciate you stepping in. So I'd like to kind of explore the event with you. What are some of the things that stood out to you about this KubeCon?

MOFI RAHMAN: I think folks that have listened to the whole interview and, at this point, have already hired from pretty much all the guests talking about AI being a big thing. But I think, for me, more than just AI being a big thing, I think the thing that stood out most is how all the other aspects of Kubernetes is accommodating the AI workloads.

Like, security, storage, scalability-- there is a lot of talks about those things that are kind of shaping the AI-- the AI workload is like one part of the bigger puzzle. And so many other things has to happen for that to work properly in the world of Kubernetes. And it's really awesome to see how quickly you mobilize the project of, I don't know, hundreds of thousands of users at the same time, right?

It's not super easy to move this many people towards the direction. So there is obviously that aspect of hype and other things related to AI. But at the same time, you also see people kind of mobilizing behind it, so which is kind of awesome.

KASLIN FIELDS: Yeah, and AI is your area of expertise, so [LAUGHS] I'm sure that you catch things that others might not.

MOFI RAHMAN: Yeah, so again, AI as the workload itself for the most part, I held a long-lasting belief of, it's just workload. But now that we have been involved in a little bit more and talking to folks that are coming from the world of machine learning and data science into Kubernetes, there are some nuances.

Kubernetes is now about 10 years old, about to be in a few weeks. And a lot of the community kind of came from the world of Docker, kind of came from that world of application development, and grew together with Kubernetes-- had this whole-- like, 10 years' worth of knowledge and other skills built up of how to take an application, put it into container, have all the best practices of security and other things.

And a lot of the machine learning discipline and that data science discipline kind of grew parallel with outside the boundary of containers. And in 2023, with the whole end of 2022, the movement of making machine learning as part of your, what I like to say, critical path of your business-- a lot of that work that we have done in the machine learning space can now have to be kind of jammed into the world of application delivery at the same time, right?

So that is a big shift. And the folks that are coming-- like, I was probably a lot stricter about being more like, OK, this is how you do container. This is you do Kubernetes, versus now, after speaking to people, after seeing the pain points of bringing your workload to Kubernetes, I have probably-- I don't know-- changed my mind quite a lot about what it means to get an AI application working on Kubernetes. It's not necessarily just throw it in a container and run it. There is a lot of consideration you have to make.

A lot of the accelerators require you to have root access, which over time, in 2017, 2018, we tried to convince people that root is the root of all evils. But I think all of a sudden now, if you're in the world of TPU and GPU, all of a sudden, you need root again because you're trying to talk to a device driver.

So we have to kind of go back and rethink that our strict principles of containers and Kubernetes is accommodating all the users. And people are coming from various angles of running application and machine learning workload, which we're now trying to accommodate in Kubernetes.

KASLIN FIELDS: Yeah, something that kind of blew my mind as I started trying to understand the landscape of AI on Kubernetes-- and we did a talk together at KubeCon about the processing unit landscape, how you use processing units in Kubernetes. And what I found mind blowing about that was that it's very hardware centric. Like, you've got these workloads that have such strong need for high performance out of these specialized types of hardware that it matters how the hardware itself does calculations.

When you're thinking about things at that level-- like, the whole way that Kubernetes has abstracted the hardware layer underneath-- you've got to think about that in a new way. It's not that it goes away. There's still a lot of the management of that underlying infrastructure that is really beneficial that Kubernetes gives you.

But it's definitely not going toward the serverless world of things, where you forget that there are servers underneath there. It's a world where you care a lot about the hardware and the servers underneath your infrastructure. And you want to use it to the best of your ability. And Kubernetes can kind of work in that space too. So it's been very interesting to learn.

MOFI RAHMAN: Yeah, I think for me, also, the serverless aspect of Kubernetes is not necessarily you forget about servers. For me, it's more about you not having to actively wrangle servers, right? From the get-go, the name "serverless" was probably one of those names that was more marketing and catchy than actually true. And anybody who dealt with serverless very quickly found out that there are servers, but you generally don't have to think about it. And the less you have to think about it, that means more time you can spend in doing things with the server that you don't think about, right?

So in Kubernetes, with cluster autoscaling and things like that, these workloads, when you're talking about fine-tuning or training workloads for machine learning models, these are expensive workload that happens on a period of time, and then they go away. So if you keep those hardware around, if you have a data center, then you have just a bunch of GPUs and TPUs lying around, then you don't really think about scaling, right?

In that sense, you don't really get the benefit of Kubernetes where things are scaling down for you to not spend any money on it. So in that sense, making Kubernetes very a scalable environment where it is scaling up to run the workload and then going away-- that is a serverless behavior that I am seeing more and more people kind of embrace and kind of coming towards Kubernetes.

The world of GPUs, like maintaining those drivers, going up and updating those things, installing the right libraries in the right places-- all of those things are challenging. And the bigger a data center you have, the bigger of a machine learning model you're trying to train, that additional burden of maintaining those servers-- that is going to slow you down.

And in this time of large language models that are coming up better and better every month, every week, spending one extra whole day to set up those 1,000 node or server cluster, all the GPU drivers-- that's a time spent that you probably don't want to do. And of course, you can get this done using just some scripting, some automation.

But the umbrella of Kubernetes that just gives you all these things as something reproducible as well as something that you can codify really nicely, as well as take the skills that we have built up over the last 10 years of application delivery, kind of bring that over, in the world of ML, that's really exciting. And I am excited to see where people take it.

KASLIN FIELDS: I think you're absolutely right. And this leads us very nicely into another trend, I think, that we saw at KubeCon, which is platform engineering. Everybody is trying to create some form of self-service platform. And it's all about the ways that different roles within an organization interact with the hardware, how they run the applications that they need to run. And serverless Kubernetes, all sorts of tools have a role to play in that, I think.

MOFI RAHMAN: Yeah, I think the whole conversation of platform engineering in my mind even started with the whole movement of DevOps. And in some ways, platform engineering seems counterintuitive to DevOps, where in DevOps, we are trying to bring dev and ops more together. In platform engineering, we're saying we should separate them.

But fundamentally, we're all saying the same exact thing. We want to decrease friction, and we want to make it easy for the folks to do the thing they want to get done faster, right? So in DevOps world, like the initial movement of DevOps, the friction was that dev and ops did not communicate. So there was a lot of, oh, I can do this as a dev problem. I can do this as the ops problem. So you bring them together.

But as your system gets more and more complex, at some point, you need to start thinking about, what can I do to not let everybody think about everything at the same time? How can we let people do the thing they want to do and do the thing they need to do more better, right?

So I think it can go too far to the point where no one knows how anything works other than the people maintaining the platform. But I think the core concept of, give people the right abstraction to get their work done quicker and better-- that's the core tenet of platform engineering in my mind. And that's what I think we should strive for.

It's not necessarily just, oh, yeah, the platform team will take care of everything, and I don't need to know anything. It's more about, I know exactly how this thing works, but I don't need to. I just have this abstraction that's given to me. That abstraction could be Kubernetes YAML, could be some new DSL, could be a ClickOps, that you click a button. And better yet, it could literally just be your GitOps. You have your code pushed to Git. And some magic happens, and your application gets to the place it needs to get to.

KASLIN FIELDS: With all of the policies and security considerations that your organization wants you to have just kind of built in, hopefully, minimal overhead. [LAUGHS]

MOFI RAHMAN: Yeah, this also kind of ties back to the original thing I was initially saying about a lot of the machine learning workload got thrusted into the critical path. Before, a lot of the data scientist and machine learning workload was one-off experiments. People did that, found an answer to a question. I want to know what exactly happened with this data.

Now as machine learning models, large language models becoming part of the core business, all of a sudden, you need to have reproducible way to train the model, fine tune the model, update the model, have the model registry, have some sort of a lineage of what happened with the thing, have regression and other things that people in the machine learning world.

Unless you are one of those bigger businesses or the companies and the teams that built up those practices, a lot of machine learning was kind of one-off experiments. From there, going to this reproducible pipelines of things that you are doing, having a platform to do this is going to be super valuable long term.

KASLIN FIELDS: Yeah, and I love this context around a couple of big trends that we saw at KubeCon-- AI/ML, of course, and platform engineering. What other trends do you think we should call out here from KubeCon?

I know one I saw was WebAssembly. Definitely heard a lot of WebAssembly talk going on, which is-- we did an episode from WasmCon last year, where we talked a little bit about WebAssembly and how it's developed as kind of a different form of-- what's the right way to say it-- isolation of applications on machines, a different approach to it than containers, that can also be compatible with Kubernetes as long as you have the right drivers in it.

So there's a lot of interest in that in the AI space because it can be really performance friendly. It can be a very lightweight form of isolation that allows you really close access to the hardware. So there's a lot of interest in that one. Any other trends you want to call out?

MOFI RAHMAN: Yeah, I mean, oftentimes, something that big as AI-- it kind of can drown out a lot of the cool things that are happening in the space. But a lot of the things that have been happening over the last few years continue to grow. The whole movement behind eBPF for getting access to kernel-level information has been growing strongly.

The CI/CD movement is still going strong. I saw a bunch of vendors providing support for CI/CD. There's a bunch of vendors growing around Argo CD, the open source project there.

KASLIN FIELDS: GitOps-- heard a lot of people saying the word GitOps around still. It's still a thing.

MOFI RAHMAN: For sure.

KASLIN FIELDS: Multi-cloud, multi-cluster considerations-- still a big thing in the community as well.

MOFI RAHMAN: Yeah, I think, in many ways, the whole movement behind AI and ML kind of pushes us to think about these larger clusters and how do you maintain all these clusters, maintain your workload. Like, everything needs to kind of evolve to accommodate workload that requires a lot of resources and nodes and clusters at the same time, right?

So the whole need of multi-cluster probably comes from running really large batch workload. A lot of folks will probably notice that their CI/CD platform needs to update to be able to help their ML engineers do all these experiments that they were previously doing in notebooks before. We're probably going to see a lot more need for better storage orchestration because, again, machine learning is nothing without data. Security is going to be a big concern because you don't want your data or your model to leak.

So it is kind of like the analogy of "a rising tide lifts all boat" kind of thing. So to keep up with this increased demand of machine learning workload, I think we're going to see a bunch of the other areas of Kubernetes also needing to evolve and be able to accommodate larger workloads than we were used to.

There were conversation about creating even new SIGs to be able to handle these kind of newer types of workload. Is this novel enough for us to have a conversation and a separate discussion rather than just considering ML as a subset of apps or storage? Can we think of the serving workload or this training workload as their own thing?

And how can we make the Kubernetes project grow and evolve to help users that want to run this workload on Kubernetes? So that was a interesting conversation that was happening on the side as well. Like, how do we make sure that we can make Kubernetes a good home for these kind of workloads?

KASLIN FIELDS: Yeah, Tim Hockin mentioned that in a conversation that I was having with him. It goes to show that the trends that you see at KubeCon, or any event, really, are all connected. They don't exist in a vacuum. Especially in the cloud native ecosystem, every trend affects the others. So I think this shows that really well.

And I want to wrap things up here. But one last thing that I want to call out is that the "Kubernetes Podcast" hosts, myself and Abdel, [LAUGHS] completely dominated in the CrashLoopBackOff game that happened during the KubeCrawl party at KubeCon. And, Mofi, you got to watch. How was that? [LAUGHS]

MOFI RAHMAN: Yeah, I had a front row seat. It was fantastic. And [? Jifi, ?] the host that ran the show-- we got to talking as well. Hopefully, we get to bring this CrashLoopBackOff game in the future KubeCon to even more folks. Like, last year, there was two games on the KubeCon NA. And this KubeCon EU, we had also two game between Kaslin and-- Christopher?

KASLIN FIELDS: Christoph Blecker.

MOFI RAHMAN: Christoph, yes. Christoph, and then we had Abdel and--

KASLIN FIELDS: Mitch Connors.

MOFI RAHMAN: Yeah, Mitch Connors. And so, in many ways, it was a battle between Kubernetes SIG Contrib Ex and the service mesh systems.

KASLIN FIELDS: That's true, yeah.

MOFI RAHMAN: Abdel and Mitch are really, really into the whole service mesh ecosystem. And both Kaslin and Christoph were very involved-- are involved in the SIG Contrib Ex.

KASLIN FIELDS: Yeah, I'm one of the new co-chairs of SIG Contrib Ex. And Christoph was one of the tech leads, who is now emeritus, who I took over from for SIG Contrib Ex. So that was really funny. [LAUGHS]


KASLIN FIELDS: But we should probably explain the game. We're just talking about it, not explaining what it is.

MOFI RAHMAN: Yes, the whole concept of the game as a viewer is there are some novel tasks that are-- there are three tasks that are given to both participant a week in advance. But you are not allowed to do any kind of prework. You can look into it. You can actually do some research if you want to. But if I recall correctly, none of the users actually had the time because it was one week before KubeCon to really do much research.

But on that day of, on the moment, you get to know one of the three tasks that you have to accomplish. The task Kaslin and Christoph had, I think, was set up a ray cluster and run some job there and--


MOFI RAHMAN: KubeRay, yes, which is like Kubernetes operator for Ray that you can run. And the task that Abdel and Mitch had was to set up some sort of cloud native way to stream your laptop webcam to your phone. So those are the two tasks.

And again, these are not something you do day to day, a lot of folks in Kubernetes. The KubeRay one, if you're in the machine learning space, if you're using Ray, you probably do it. But again, it's not something probably Kaslin or Christoph does day to day. So it is like more of, like--


MOFI RAHMAN: --go off onto the docks, learn about how to do it, and how quickly within half an hour you can get things running. So yeah.

KASLIN FIELDS: Explaining KubeRay was not part of the challenge, which was nice. [LAUGHS]

MOFI RAHMAN: Yes. So yeah, both Christoph and Kaslin actually got close, but Kaslin's one was actually scaling the job that was running. So in a decision made by [? Jifi, ?] it was Kaslin who was deemed the winner in that competition. And Abdel found a technicality where--


KASLIN FIELDS: Thanks to Bob Killen. [LAUGHS]

MOFI RAHMAN: Thanks to Bob.

KASLIN FIELDS: Volunteering.

MOFI RAHMAN: Yes, thanks to Bob slipping up, Abdel found a technicality where he set up a StreamYard stream to YouTube and opened the stream on his phone, which was very funny at the same time-- again, technically a win, which is best kind of win. So yeah, congratulations to both Abdel and Kaslin for having that win.

KASLIN FIELDS: Thank you. And I hope, listeners out there, we did you proud at KubeCon. We held the "Kubernetes Podcast" banner high. [LAUGHS] I wore the hat through the whole thing.

MOFI RAHMAN: And yeah, if you are coming to KubeCon in the future, there might be more chance for yourself to get involved with one of these CrashLoopBackOff. These are always a lot of fun, and you kind of learn new things. As an audience, I learned a bunch of new things. So yeah, I'm really excited to see where this goes. Seems like there is a lot of interest from the audience. This was in an open-- it was not even in a room. It was in an open space in the middle--

KASLIN FIELDS: In the showfloor, yeah.

MOFI RAHMAN: --of KubeCon. And the space--

KASLIN FIELDS: It was awesome.

MOFI RAHMAN: --was packed. It was a little bit-- yeah.

KASLIN FIELDS: So thank you very much, Mofi, for joining me to talk about KubeCon. And thank you, everyone, for listening. If you're at an event that we're at in the future, make sure that you come find us, because we do these mini interviews of folks who are at the events. We would love to interview you and have your voice on the podcast. And with that, we'll see you next time.


That brings us to the end of another episode. If you enjoyed this show, please help us spread the word and tell a friend. If you have any feedback for us, you can find us on Twitter at @KubernetesPod or reach us by email at <kubernetespodcast@google.com>.

You can also check out the website at KubernetesPodcast.com, where you'll find transcripts and show notes and links to subscribe. Please consider rating us in your podcast player so we can help more people find and enjoy the show. Thanks for listening, and we'll see you next time.