#037 – Problem First, Kubernetes Second: Insights from Ahmed Bebars (New York Times | CNCF)

Itiel Shwartz: Hello everyone and welcome to another episodes of the Kubernetes for humans podcast my name is Itiel Shwartz and today with me in the show we have Ahmed Ahmed happy to have you

Ahmed Bebars: Happy to be here and thanks for the invite and really excited to see where is this conversation gonna go I’m excited about you know and everything around it

Itiel Shwartz: Yeah so as as I told you like a couple of minutes ago and the the show is quite clusive in the format and what I would have first of all is for you to introduce yourself what’s your background when did you start like I don’t know like doing some like a software engineering or operations and then we’ll like go through your story until the point where you learned about Kubernetes but you know let’s get started

Ahmed Bebars: Yeah let’s get started so I actually have a very interesting background but like I come from like an accounting background that’s my major and then I transitioned into that 15 years ago like I have a passion about technology so I started from there I always love to play with computers back in the day since my teenager age but then like after that I had my own startup I worked in different companies I always was passionate about cloud computing and AWS in the past few years I found myself more and more like swifting from like backend engineering to infrastructure more and then trying to like get into like what’s cloud mean for people like how things evolve and over the years things evolved from like just being cloud to Kubernetes to other things in the community right now I’m a principal engineer at the The New York Times but also like I work a lot is CNCF projects and I’m a CNCF Ambassador as well so that’s where I’m at trying to help Community understand what’s a platform I think my goal overall so to speak work-wise or Community wise is to help people being able to bring their products to life without having to learn all of the details because cloud native ecosystem and the tech ecosystem is coming faster and harder and larger and then you can compose it all in a small mindset so that’s that’s where I’m at

Itiel Shwartz: Okay so you know you said like a lot of interesting things in one in one in one paragraph so let’s get started maybe on a have to ask like from accountants to like you know like platform the The New York Times on the top of my head it does indeed sounds like a very weird like GenAIourney so how was like the first like change from how walk us through that

Ahmed Bebars: Yeah definitely so back back in the days it was like just like college just like I had to go through like a Commerce college or go through accounting but then like I started to find myself like just like college is not the thing that I wanted to go for it was more of like I’m just one of the college have a college degree and that’s what everyone want want to drop out of college back and doesn’t work this way so that’s where like I started to be like more connected to like computer stuff and then my Python started around like or what scripting mean what’s like web development mean all of the kind of elements to be honest and until this point don’t like front end I don’t work with it too much I just like it’s not anything of understanding but it’s more like I focus more on the back end and that’s where my transitions and then after college took a boot camp and that was like a nine months in software engineering and e-learning Technologies that’s where I learned a lot on how to do that and I decided to have my startup so I had a startup for a while for a couple years where like I was doing custom software development and that’s basically where I learned a lot like different languages different cloud ecosystems on-premises software and all the things like that after a while I just like start to work for companies climate didn’t wasn’t wasn’t good for a startup and like I wasn’t good at it because it was just my first thing after college so it wasn’t it wasn’t I learned a lot from the process definitely will try it another time but it wasn’t meant to be at this time and now I just like over the year the transition was like not something intentional to be honest it was more of like a step by step like in a couple years I work in backend engineering and then like I found myself like more swifting towards DevOps because I just like started to like the ecosystem I started to understand it more more and more so that’s where like start to be like a gradual change until a point where like like five years ago I was like I’m G to go like full DevOps full platform understand ecosystem work more on cloud don’t work more on backend engineering and that’s where the whole process happened

Itiel Shwartz: Okay no it sound it sounds like you know like quite a journey quite a journey and he also said that you open your own startup right why why how why

Ahmed Bebars: Why how like it was just like I was I was passionate about like what exactly I wanted to do and like to be honest I wasn’t sure yet if that if if like what exactly like kind of jobs up fits me I was like more into engineering but I also like wanted to see like how is the startup echosystem the problems that I was trying to solve like back in the day when I was when I when I were in Egypt like I had a lot of talent and they want to off like make good products out of it so like what we started doing is like building projects for different clients across we didn’t have a product it was more of like an a startup that like does some custom development for other companies and that was like the major thing but like you know young age out of college don’t have a specific Visions on where exactly this would go I didn’t have like that enough experience into like also business management all the kind of stuff I had enough experience in Tech but not in the other business side of things and that’s where like you I needed to have the balance and I was drifting more towards the tech rather than the business because I was passionate more about and that’s where things didn’t work out really well but it worked well for like a couple years but then it didn’t work well and then I learned my lessons but it was a fun journey to understand how the whole ecosystem work like how Tech work with business get some ideas all of the kind of stuff and then move over to like be completely focused on Tech but the business perspective that that brought to me is is helpful I think like you know

Itiel Shwartz: I think platform engineering is in a way you need to know how to balance like the technology aspects of things and most of the time the business requirement or like the business future I will say and how do you build the system in a way that makes sense for you know not just right here right now but a year from now right or a couple of years from now so maybe let’s talk a bit about Kubernetes when did you discover about it what did you fail and how do you feel currently on the Kubernetes and the ecosystem

Ahmed Bebars: I discovered about Kubernetes like about I love to say like seven or eight years ago maybe a little bit more just doing like you know all of that kind of deployment all of it was like instances and like you get your app running on like a machine and all we know is a virtual machine cPanel all of the old stuff and then like we were like oh there’s something that can manage container on a scale what do that mean and it was like oh holy like this is there’s a big thing that does this and I was like I don’t know anything about it let me just like dive more to understand like what exactly it does and to be honest it was like one of it’s a it’s a change because like it’s just like a drastic change for like how you do things it it’s I want to say it was a bad change or a good change right now I could say it’s a good change but back now I didn’t know like if if that was the things that but I have seen what I have seen in it is just like there was a problem and there’s a solutions that is trying to solve it it wasn’t not that popular back then so that how many years ago was it

Ahmed Bebars: About seven or eight years yeah yeah so yeah starting to gain the popularity but yeah yeah so it was about to start and like but containers were the things that like people are playing around with and done forever and then like I started to look into it and then like but I found a lot of useful use cases out of it the one thing that like really resonated did with me first it was an open source thing that like just like you know you just can play around it with different providers and then according to the journey all I have seen is like people not try to fight it they are trying to build on top of it which makes it easier for the community to adapt so like you find like all of the cloud provider now they have their flavors of doing it but still Kubernetes behind the scen just like it doesn’t matter and that’s where like I think like there’s value on having open-source Technologies in the market that’s widely adopted because also like me and you can talk Kubernetes right now and like we probably have different Stacks different things we do but it’s still like you know the same thing you might have like different couple things I have different couple thing and I like that you know have a shared language and I think that’s what open source and what Kubernetes built around the shared language is like we know what’s a deployment what’s a service all of the kind that construct are there and we don’t you know have to like explain what is that to one another maybe no no it makes total sense

Itiel Shwartz: Maybe share something that you know you know you’ve done quite a lot of of this work maybe share with us some like a harder story like some hard integration some hard M migration when you know when you tried it to like maybe like and like bring the new technology into the company how does it usually works like I know there is a lot of tension some tension right between Developers and operation and platform if you can share like something you know it can have like a happy ending but something that wasn’t that vanilla right like everyone saw that and it was crazy yeah

Ahmed Bebars: Nothing comes top to mind but like all I can tell you is like let me see if it flew into it like change is always hard because change requires effort like the one thing that usually makes more sense when you start thinking about abstraction because change behind an abstraction makes more sense of doing things so the one like I can’t tell a different story but like it might resonate with you so the one thing that I have seen not according to change were like a bad thing there but like one day in a in a in an organization that I used to work for we had we had a situation where like we want to move move they were in Kubernetes and then like we wanted to move from like one cloud provider to another and I can’t tell you like I’ll tell you the fast forward thing to it the DevOps team moved it in a single night like an entire set of a Kubernetes cluster just like you wake up in the morning and like hey we were in one cloud provider yesterday, and today we were in another it’s crazy it’s yeah it sounded crazy but it was like and that’s because of of Kubernetes that because like everything was the same they just like have to like move things around from one place to another behind the scene they build the right abstraction across tooling and then like you woke up in the morning like we good to go so don’t have to do anything and like I was that time as a backend engineer and didn’t have to do anything I just I was like sure my workflow is the same everything is the same so when you can do things like that I think it’s easier but change usually any new idea the one thing that I learned through any any new change that you have to bring you have to resonate the value proposition before you make anything like if I go to someone say hey you should move to Kubernetes the first thing that people will tell me you are insane like you shouldn’t do that and they didn’t even know because I went to the solution before the problem but then it guy go through a problem first and that’s what I usually say don’t beg the tool before a problem I love Kubernetes but it doesn’t fit everyone some use cases might be better with other things so that’s where like I feel it’s like you mentioned this earlier it’s all about how you handle it like how you really like propose the change if you come with like this is a change that we going to make but you can come what solution can help us solve the problems that we have both have the same results but one have is a hard way one have is a easy way

Itiel Shwartz: No you know even that I do love Kubernetes as a technology and as a ecosystem I agree with you like Kubernetes is not for everyone like it’s not the simplest tool I wonder like you know I guess given your background you saw quite a lot of projects is there anything in particular like any project running on top of cornetes that you think is doing a better jobs than the other ones I

Ahmed Bebars: I don’t think like something specific doing one thing over the other one but like there’s a lead in each area where like you find it like oh that has more contributions at more popularities that more features at more support and then I feel like from a CNCF landscape there are like some some projects have done that so one project that I have seen in the networking space is Cilium is doing like great job like lavaging eBPF all of the other things in the service and service mesh space Istio is doing good stuff like was the ambient coming and reducing sidecar problems all of the kind of Stark actually have a talk coming at KubeCon about sidecars and ambient and like yeah

Itiel Shwartz: Sh share a bit because personally I don’t like Istio like I think Istio is one of the more complex project that I saw in the cncf right like it’s not your beginner hey I just Le Kubernetes project so you know maybe share a bit what do you like you know it’s a I think people love to hate Istio but in the end of the day a lot of big companies are using it so you know yeah like

Ahmed Bebars: Here’s the thing I I’m telling you again it’s all about the problem Istio is could be the wrong choice if you don’t need it like if you are in a single cluster everything is smooth like you don’t have any you don’t have any problem with how to route your traffic you don’t have all of these situations that you want to test for is you might be the complete wrong choice for you but if you are in a multicluster environment you have like services and you want to build tolerance and then you want to have like your services in one region aware of the other your routing is flexible you can be able to say like if that happen in this case have a routing to that service but like when I move to a progressive delivery like let me run a canary release to version two and you can see like things like that works well so what I have seen Istio good ways is like mTLS stuff that’s great then like I used it for a bit but it’s not the biggest features that I’m looking for I mean the biggest features that I’m seeing in Istio to be honest is like how it shapes traffic in in different direction and how you can have like different ways of saying like how to handle traffic out of the box like circuit breakers is one of the things that is you can provide to you out of the box all of the failure scenarios like you don’t have the problem that comes what’s issue solving from my perspective is usually you can do all of that on the application code and I like you can do circuit break and I have seen people like Implement different scenarios but like what a product like Z brings to you is that like you just configure it you don’t build it so like I can say if my service if this traffic to this place go like above like x% just fill it over don’t do anything and I don’t have to write the logic to do that is is also helpful for like different situation where like you want to route your traffic differently in different situations like based on headers based on different things you can use it for for example like you can leverage how you virtualize a service into like a way of like I want to say like into like review environment or like Dev envir like something like that you can say like this service talk to this service but if this service not available talk to the other cluster in that service so having a mesh in a large scale cluster it’s helpful but like if you have two services and then like you tell me I need to install s we’re gonna have a problem because it’s just like I have a couple services and you probably shouldn’t have a Kubernetes cluster in the first place kill the Comm cluster then yes like go to fire base or something like that Oru and go do anything it’s just like it it’s it’s sells a problem KEDA is a is a one another project that I think like helps with the scaling mechanism I like I like where the project is going I think like it’s it’s super helpful and there are many others but you know in each category there’s something and there are like the you know core projects that we don’t talk about like containerd and other things that’s already embedded into the ecosystem that no one talks about them but like they are doing phenomenal job so

Itiel Shwartz: No so so you know like the point is interesting you know I will like another question on that because of the day I again like I feel people you like big companies do tend to hate is so you know when like did you like enter is into any company that you worked with like you was it you is it the team I feel that always when people are using Istio they have like two people only holding is your hand so it won’t get crazy so is it true or or am I like a like I’m like a exagerating here

Ahmed Bebars: You’re not exagerating and in some cases it’s true and I have seen it and the problem is it just like you it it’s it’s like let me tell you this Envoy on its own is complex imagine building the whole control plane on top of this if you don’t really grasp it really well it might be one of these like black hole areas we like I know my traffic is going I still know where it’s going it’s just going somewhere and then you need to have the expertise into it so it’s it’s one of the decisions that like I really like wanted to make sure that people understand like it’s a great project but like you just need to understand it comes with its own investment like when when you bring a value it’s not because one of the things that like we usually I try to tell anyone in the community is like open source is great but there is an investment that you have to make when you adapt open source it’s like similar to like buy versus build like we have to consider open source as a buy as well there’s an investment you have to make it’s not managed by a vendor I understand understand that but like still it’s it’s there’s an like what happened when you have a bug in next you what what are you going to do like you are going to fix it go ask contributors to do it like there there’s a process that and might require like more investment than you would require in a vendor in some cases but it’s good that like there’s a contribution to it like there’s people are solving the same problem and that’s what I love about open source it’s just like everyone is trying you know to do their best and contribute and you find some people are already devoted to open source I’m not at that level yet but like they are only doing open source projects and they maintain projects and they have done successful and phenomenal job at it

Itiel Shwartz: No the you know like I agree like you know the open source is what help us to move in the speed and velocity that we’re currently moving at you know looking at the ecosystem where do you think we’re going like what’s the interesting project projects that you think are going to be popular in couple of years same as you know for like Istio orada that you mentioned which I agree super popular so

Ahmed Bebars: I believe what we’re going towards is going to be interesting I think like Kubernetes is going to still be in the way but I think like we’re going to go more towards people are also like thinking of platform engineering that can abstract that so we want to have more options about how we scale compute without the need of like actually have to understand this this whole thing so like more abstraction is going to come I don’t want to say serverless will come hard on this but serverless has been there for a while and to be honest it helps it helps like I think like the one thing that like love hate problem about like between Kubernetes and serverless is Kubernetes is like you know when you have something encapsulated so there boundaries and into and out of it and everything is defined my problem with serverless it’s not a problem it’s just like you can go Rogue any minute because each component is so small that you can like so having having something that productionize this in a better way that make it like scoped in a way that like I can spin it fast but I don’t have to go into a trap of like being like doing everything can go wrong I think that might be but abstracting more different things like one problem that yeah I’m sure you’re aware of like Kubernetes ecosystem is great but it’s also getting complicated because like you add more and more to it like we have that Kubernetes Ai and then you have like a lot of CRDs that you’re building on top of that so gets you into like a massive state where like different things happen I think like more abstraction towards that like getting to a state where like it’s a simple app I just deploy a single YAML gets me all of I need I don’t have to worry about all of this so more and more manage like manage complete manage Kubernetes solutions that like you don’t have to think about anything other than just deploying your app I think that might be a thing that we have SE we going to see but and also like you know with all the GenAI happening around this it also helps like make sure that the trend is going into that direction because that encapsulation and that learning curve will make it easier that’s where I’m thinking we’re going to go but we’ll see

Itiel Shwartz: Any you know like one of the things that we are hearing quite a lot and to be honest like for full disclosure I am working on that quite a lot is AI and Kubernetes can we use GenAI in order to make I know developer life easier operation guys easier any thoughts on that area any interesting thing that you are seeing

Ahmed Bebars: You can use GenAI to make a lot of things easier I think like like I’m not debating Z I think like I just wanted to see where is the problem one thing that I really personally have seen that GenAI is good at is like I use things like that in a smart search in a smart completion like I don’t have to have like a fully like me telling like one one problem usually I have just like you write a command and they’re like I have to remember all of the parameters and everything and like what I really want to do well let’s imagine like I’m trying to get my the pods in the cluster I’m say like get me the pods and just like they they show in front of me or like there’s something wrong going there like what is that like could be or like build me a service that does that like it doesn’t it doesn’t really like it it doesn’t mean that I don’t know how to do it it could make me some time to do it and like maybe like a few minutes and then like if I ask a question and I can it rate on it I don’t know yet where that could be used in production or like how but I see it as a good start for how you like still have a human in the loop like kind of style like help me get there faster but then like I still have to validate what’s happening the one problem that I like I would I would just like argue everyone to make sure that they don’t get trapped into it let’s not make that take over on like the whole thing because like there’s there are things that like and I have these situations like you could ask ji something without guardrails it could come with a great idea that never existed and it happened it’s just like you can do that they’re like that never no one implemented that like I get the idea is fascinating but like this is where of the areas and I don’t know but guard rails around the ecosystem removing the small barriers and all the kind of stuff that makes it makes AI super helpful and let’s be honest we’ve been using AI for quite a while like with all of the Alexa Google stuff like for years like it’s it’s been there the difference is just like it’s coming with more to be narrative don’t have to really be trained it just like takes the information so I would I would use it as I have a data lake I would just like throw it at a specific GenAI and ask me to find the correlation give me the right resources things like that

Itiel Shwartz: No that that is interesting I think that yeah like it’s one of the problems that I think it’s very hard to hard-code like to to write like what is interesting while like just throwing it on a GenAI does provide like really superb resolve compared to you know coding it so it’s it’s a great example and what are the biggest challenges like you know when when you think about like the adoption of the tools that we just describe and so on and you know you know you mentioned serverless a bit like in that today serverless didn’t explode cores did explode and like what are like the biggest challenges that you see that awaits Us in know end of the day yeah

Ahmed Bebars: Biggest challenge in any adoption is the learning curve it’s just like you’re getting to it’s a change you get into something that you done now like how much effort get you to understand the whole ecosystem and if if you’re diving from like very very like way of like let’s talk about Kubernetes let’s talk like where where where should we start like like like is it is it like to start with a pod or with a node or with a deployment or a service or like there’s like there you can set for days decide I think adoption always comes in like there’s a change that one thing that like start like I always feel that it’s more appropriate it start with a learning curve start was a problem like what are we solving and this is like problem with Gen as well as like we’re going for Gen was like hey how how to implement GenAI like I would more rather like there are problems questions that you ask like how can GenAI help in that space so I think of we thinking about the problem we still like having a bit getting the solution before the problem but like that’s not a solution at this point it’s become like a batter like we have seen that helps in other areas let’s try to apply it in that industry and see if it’s helpful so that’s more like like there’s a problem what are the problems that this patterns can help us I think that’s that’s more a helpful approach in adoption and we’ll see but anything that like will explode in then like when you find the communities that have the same problem and there all there’s a solution that align that problem so and that’s what Kubernetes happen it’s just like there’s a problem solution aligning on it everyone like want to contribute to it and that then when it start but all Kubernetes came from like you know a company who built it for a specific reason and just gave it away so that’s that’s one one reason as well

Itiel Shwartz: No I think that that makes like again like total total sense any like closing work like we’re reaching like the end of the episode anything that you want to sh to say to you know anything else

Ahmed Bebars: Really just like oh again and again like big pick the problem before the tool that’s that would be super helpful and you know technology just like try to build things in a decoupled way where like things will make it easier that when you have to have interchangeable pieces that usually helps in applications and other things and Technologies are going to always evolve on the community side I’m super excited that I’m gonna be in KubeCon EU London in April and I’m gonna have a talk about sidecar and like server and sorry service meshes and ambient and with my co-speaker is Lin Sun she’s she’s one of the Istio maintainers so I’m super excited about it and also Al like we just announced Kubernetes Community Day in New York and this is going to be in GenAIune 4 this year so looking forward to see a lot of people coming to that event but at the end the open source industry is and the open-source ecosystem is great let’s continue to contribute to it and let’s make a difference

Itiel Shwartz: Okay strong Works to to finish the episode pleasure having you

Ahmed Bebars: Thank you for having me appreciate you bye bye

[Music] Kubernetes for Humans.

This is an AI generated transcript of the conversation