Serverless, Microservices, Software Architecture and Food 😋 with Yan Cui

Hi everyone. My name is Patrick Akio and if you just didn't serve it as Technologies, software architectures and Fook talk, this episode is for you. Joining me today is Jung Choi. He's a blogger consultant and an aw, serverless hero. There's some good perks that come with that title. I'll put all his socials in the description below. Check him out. And with that being said, enjoy the episode.

I mean, I've been to Portugal for the last month and I eat a lot of foods, okay? I can't handle avocado as much anymore. Like, it upsets my stomach and there's one thing I can't handle any, coriander cilantro is not a thing. Oh, running Portugal. They throw it on everything. So the first meal I got, they have this beautiful dish, it's called a cut up Lana. It's like a fish stew and make this one pot meals. Like a saucer, and they put on the table in the open.

It's like a whole presentation. And the waiters all thing with it and some cilantro area, I was like, well, shit, I should have said something, I guess. And then every single restaurant, if I would forget that say I can't handle cilantro, it would just be there and I hope you do that. That's my fault. That's why it's take anything. And I never knew that was the thing is, I guess it was similar

to me when I was in the thing, was it that we went to Europe, right? After what after math? I got married, we went to Venice for our honeymoon, beautiful. It was it was great. But then, every single restaurant goes to every single dish, you have all starts with Mustard relish cheese and tomatoes. So out of our street Asus three or four days straight of everything, just being cheese and tomatoes, we're just the time for something else to even just like getting a burger or

so, yeah, I'm gonna get that. I mean, I told you, I go friends, Italian and it's just, they're simple Foods. I could tell him food is amazing, but they're kind of basic kind of generic. As you say, you had some bread Montserrat tomatoes that the most beautiful, tomatoes, and, yeah, at some point, you want some variety? Give me something else. Yeah, I can't do it. I can't talk too much about Italian food and, and bad in the same clothes, variety. And yeah.

Okay. So something that's not so tomatoey, that's hilarious, man. That's actually what I ate a lot in Portugal as well because they do have amazing Tomatoes, but we would switch it up like we live together during covid. We moved in together and I came like before then I was living with my parents, so I didn't really cook Couldn't cook because my mom would cook for the whole family, then I live with her and she knows how to cook pasta and I was usually mostly a. So then we had to invent.

Okay, what else are we gonna eat? Because I would otherwise, I would have pasta for lunch and pasta for dinner, and it will be the same pass. That would be also with tomatoes and mozzarella or eggplant pasta or called Alfredo, which is the cold version of pasta.

And what's your Le? So, yeah, but then finally, I got some, my mom's recipes, my parents are both Turkish, and now we're starting to make Make more Turkish meals especially now because we've been living together for like three years now, eat a lot less pasta which to my surprise, she actually likes as well and enjoys. Okay, that's good. Yeah, our parents are going to come over and of this week. So it's gonna be a long probably forehead foods and cultures.

Like I didn't know it was this much of a thing, but two, Italian people food is very, very important where we eat. And I mean, in Portugal. For example, she researches all the restaurants that are there. And then we pick the Swans. And if we have to wait in line for half an hour, she's like, yes, there's a line that means there's good food, and I like know there's a line. I don't want to wait. So here we find a balance within there somewhere.

You should do end up waiting. And the foods really good though, let's get the iodine. If I appreciate that, you know,

food culture. It's also very big we on, from as well, originally from China. Yeah. And they say, yeah, like, we're at the city, where I came from the number one hobby that everyone have a, which is that talk about new restaurants as self and yeah, even which ones are good. It's almost like almost like a like a Prestige thing that knowing the good restaurants that the job just opened that no one's known and knows about yet. That's exactly. Yeah.

Feels like yeah. I mean in Amsterdam, me of the same thing. I don't know if you know there's a new Pizzeria it's called Michaela's pizza and that one originally was in the UK. That was the only Kerr, the only place they had my girlfriend used to study there and I used to travel there, come visit her, and we will get pizza there because it was one of the best pizzas we had right now. It opened up a joint in Amsterdam and it was a whole

thing. And apparently On nuke is there was a huge line on opening day way too many people way too little staff, so it was actually me quite disappointing, the pizzas and stuff like that. Some people will get the wrong order and just was a bit of a mess and now they're rolling a bit better. But yeah, my girlfriend knew her sister note, knew it was a whole thing. They would like their opening on. I don't know. March 13th. And I'm like, how the hell do you know, if any everything's on

Instagram? Yeah, that's funny. Yes, very funny. So I said this, there's a lot.

A burger restaurant. Well, but the joints in the Amsterdam is so, and I've basically tried almost every single one of the so-called best perk in Amsterdam and the one, the best one, I found was actually, it's like a tiny little joint just in the east and west of Amsterdam code that had to have people. They do this really nice. So medium-rare burger and it just blows away. All the other ones, the so-called realized. Yeah best book in the go there then.

Okay yeah check it out. Happy Bull and I think they only open about five o'clock onwards. Okay. So you can't just go there for lunch and fortunately, for the weekend, but yeah, I said, I do a lot of research as well, in terms of just know, we're which restaurant to go to. Yeah, one of the things I noticed as well in Amsterdam or maybe just Netherlands in general, eating out is quite expensive. It is a to London. Yeah. Like especially for Asian food.

You know, we go to some of the restaurants in the new center of Amsterdam and, you know, it Every dish is partner 17 Euros, 27 euros and the same food in soft. Chinatown in the in London would be 10 10 pounds or maybe even less. So it's actually found it a lot more expensive to eat out over here then compared to London, which is quite surprising to me. Yeah, I agree. I mean I've cleaned so long and quite often and I agree with

you. I'm stirring prices are high and they keep Rising. Like, I've just come back and I'm just going to be Dreadful. Because Portugal, everything was cheap. Like, I will get a whole chicken and so we don't cook them was 750. I'm like, what the hell? Now, we get a salad. Like, not even a big salad. Like, we're talking cucumber tomato, onion, and that was also 7 57.

Mm, like, what am I gonna get the solid, but this chicken and actually in language because I stayed in language of the month. They had a place and it had the top 50 best burgers at how do you say that? It was listed on the top 50, best burgers in the world. Now, I don't know what listing that exactly was, but I believe that like some marketing thing,

I'll go there and there. But that burger was delicious, actually have pineapple and just are all the standards of a burger, but it was really, really good. Pineapple. It has a interesting ground. Yeah, I also thought so, but man, that Burgers delicious. I went twice smaller restaurants. I went twice to it. I was like, I want, I want that burger again. Yeah, this is good. Food, thing I found was that that is very hard to find the

lamb over here. Even if I go to keep up over the, Donald keep up, is its beef. Yeah. And that's really strange as well because of coming from UK, you know, doner kebab. There's always lamb and get a lot of lives of lamb dishes in general. But over here are fun and really difficult to find the lamb tissues apart from say, like gettin Curry or something like that. You can get some lamb Curry but a lot of restaurants don't serve lamb now in any reason why that's the case, no clue, no

clue. Like I make a lot of lamb dishes myself. I go to the Butcher and get it but our restaurant, I don't see it. You're right. Yeah like up behind the no think they got lots of beef and local roads of a pork. But just, you know, very few, very few places you can actually can lamb. I guess, I can, I haven't tried butcher yet, maybe was the to the next time.

Yeah, I always go to the Turkish Butcher and there they have it because I make a good Steward and especially right, sometimes I love a good stew, but yeah, you're right. I haven't seen that much. The only place where we get lamb is from this thing. Is there like a Turkish butcher? Course Sahand Mark Supermarket. Okay, at this, one in the East Amsterdam, East News the park, that's the one. I know that's where we used to live. And we still get the motor push it.

There is a lot of really good you know, oxtail and things like that and Lamb Chops but outside of that is we can almost never find them anywhere. Now I would have to wait. Yeah, I don't know like I don't like any of the Turkish restaurants that are in Amsterdam and then like I've tried a few and I'm like, it's just not as good and maybe I'm biased because my mom's a great cook, but still, I'm like, I can

pick a better meal than this. So we don't really go too many turkeys restaurants but I know like, It should be a lot of lamb dishes in the Turkish Cuisine. There's a lot of eggplant for sure like everything everywhere you go you'll find that plan on on a Turkish Cuisine yeah we have lots of really good Turkish restaurants in. London just lots of ideas all over the place and there's always you can always get a really good you know, Donna keep up or something like that.

Some no or lamb skewers or things like that. Yeah, that's one of things I really miss for sure. I went to one in London and then I was my girlfriend booked this trip to Istanbul. For my birthday. We went to a Whoa, and I found the same place. So apparently the original one was in Istanbul eyes, and then you had the one in London as well. It was really good, it's really good.

I like it when food is cheap, when going out is cheap, and it is not cheap here in Amsterdam. No, that's that's quite sad actually. I guess food prices in general are going up. I heard like, I mean, I've been away for a month, but I've heard prices are even more increasing now. And I guess, if you're here more on a day-to-day you just don't notice it. You build just keeps increasing slightly, but I've looked at the

supermarket. And I'm like yeah, everything's like five percent more expensive again. And I also guessed that they also making portions smaller as well. Are there. Yeah, I've had quite a few cases. We have seen, you know, start the packaging spot the same but then it's the actual portion that you get. It's smaller. Yeah. So instead of a raising prices or maybe as far as raising prices, they're also making the portion smaller as well.

You also get the inflation and the string string cajón or something like that, what does that there? So take the string. The size is, but I think the price is same. Yeah. Exactly. And sir it's a different way of doing the same cost-cutting solve a cost-cutting, I guess. And the from what I heard London you can generally support even worse in terms of the inflation and the supermarket prices going up. Maybe you're more so than the Amsterdam. Yeah.

A ballistics Energy, prices are coming down with been to keep an eye, on my, have my energy bill, and the nice kind of back to what I was saying. Before the, the hopes of energy energy prices a crisis happened was started. Yeah, I don't know energy bills. Just something that escapes me, I looked at it and I was like, I think I should be paying more on a month-to-month basis. Like I don't have a great contract, but it was like your usage is less than a one-person

household. Right? And Mike. How's that possible? Like our apartment is quite old. So I thought if we heat up something, it should take a lot of like heating to do that. If I heat up the apartment where with two people, so we use a lot of electricity, we use all the gas and at the end of the year, they were like, no, no, you're going to get money back and I was like, I don't know how I got into this position, but my electricity bills were actually quite decent this year and it's

pretty cool. Yeah, so I'm happy with that kicking myself that when I signed up with a budget that I didn't say I didn't lock it in for three years. Oh yeah, everyone lucky for like one year and then so just as the prices go out, I was only 15. Prices for about two months and then the 0 or suddenly, you know, you go from paying, I don't know, 30, 40 euros a month to yeah 120. Yeah. I was like crazy.

I thought I was going to live in this apartment for a year so I was like, I don't want to lock in for anything anything more than a year. So I was like, I'll just I'll just do the bare minimum. I think I locked in for half a year and then like three months extensions and then I was like, oh my God, I made a mistake but apparently the bill was fine. All right guess we're lucky with that one.

Yeah I mean I I looked you up online. And, and there's a title that I always think, is interesting because your says, 80 is served as hero. Can you touch upon how you got that tight on the first place? So it depends runs these Heroes program similar to the Microsoft for MEP program. So is the recognition of people that are contributing to the community in some way.

I think initially the program was targeting mostly people that runs Community meetups event organizers, having said that, but then they They decided to expand the test of Heroes program to have a most of people that are not just all good, even organizers. But just people that are contributing to the to the community in some way. And they wanted to I think have the separate several categories for machine learning for developer tools for serve less

for containers. So the heroes program is so nominated by internal what people internal to AWS. so, I've been blogging for Hold on about maybe 12 13 years now and I've been doing it for a long time, the Fairly consistently and I guess I was one of the earlier, you know, served as adopters and there's no on the earlier people does was really sharing a lot of the stuff that was learning and to figuring out along the way.

And so when they broadened the heroes program, they serve set up the the containers and and service category. And I guess I was one of the people that was Nominated to be the first batch of people that they invited into the service Heroes program. And so that's how you get a

title. And this year they've started to see, require you to have consistent one, more activities so that if we don't have people who got the title and then stop doing things and then they were still, because they as a hero of in, they want to just make sure that they give a title to people who are still continuing a contract. To the active, and helping a community in some way, that's also parallel to the heroes program. There's also a Community Builders program, okay?

Which is a program that you cannot nominate yourself and then they have, I guess some committee internally that the sides are we are. Okay. You get the, you get to get to be inviting to the program and you don't have things like that, the heroes program. I guess the biggest perk you get as part of the heroes program is that you get A free free trip to reinvent so they pay for the flight, that's a hotel. Yeah, and the conference take it

as well. So, and also, during the keynote that they reserve the first, I think the first row of something for the, for the heroes. So, so, yeah, that that's probably the nicest perk that you get besides that, you also get regular, I guess like no sessions with oedipus's service teams so they give View like a road map of what they are thinking about and occasionally, you get access to private beta for new services or new features.

You can try them out. But usually when you, when they get to that point where they engage the heroes and the it's already quite late in the cycle. So there's not other things that they can actually change. They would just get some feedback. Yeah, and that's what I think that actually, you know, I was hoping that they would maybe change so that get people involved with the earlier, get the feedback beer. And I'm seeing some of that

already. In terms of, you know, is the acting team, for instance, getting feedback from people not just from the heroes program before customers that's all. Yeah, for like strings have been working on like the JavaScript resolver things that released last year and as well as the new merch API think that they feel released just a couple weeks ago, she has it. So hopefully that's going to be more of a common thing you see across the, it was a things but obviously everything's different.

Different. Yeah, it's no one company with like hundreds and hundreds of first Hops and so we're within the within the same company. So it's just a lot of inconsistencies in terms of how different teams that work.

But yeah, the heroes program is it's been it's been really good for me, just in terms of you're giving me some, I guess the authority as a consultant that know that really helps even though, I guess, the program doesn't really recognize the competence is not a competence program. It's not Is a community contribution. Yes. Like a visibility thing. Yeah, but I still good thing for visibility.

Yeah, for surely. I mean the, the perky touched upon going to reinvent, like Activia we get like a study budget and it's rather substantial budget and people save that budget until the end of the year. When we invented the book, The Flame, the hotel, and the whole conference, and they save up, and they spend all of that on that conference. So, a lot of people will be jealous if they heard. That's that's what you get. Yeah, that's awesome. Probably not the best use of a

study. But they're making use of it and making use of, it's good for this for me. Now, the most efficient use of high school because he is quite expensive and honestly, I mean, the last couple years, I know last time, the last couple times I've been to reinvent, I haven't really been to any sessions because all of them are recorded. Another breakout sessions anyway. Yeah, you go to the there, was it a chalk talks. Those are more interactive and the workshops are pretty good as well.

And spend most of my time in bed. Is just no kachunk people. Because that's the only time when everybody around the world hadn't just get it together. Yeah. And the law people that are only ever talked to online. Finally, the house to be my one chance every year to see them in person. And yeah, and this kind of know have that Network and after socializing I which is what I tend to do.

Most my time with the reinvent, what a breakout sessions are pretty much available the next day on YouTube. So anything that's good that you can always just watch it on YouTube and your own. Time and also at 1.5 speaker is faster more efficient through exactly what it sounds like an awesome experience that guy is so, but if I hear people and talk about the stories that they have on there and just the experiences and meeting people, it does feel like I'm missing

out so maybe I should. Maybe I should go. Yeah. It's just that now you've ever been before? I definitely definitely try to go at least once especially now that the person events are kind of back no back. I think the last week Was like a mix of in-person and virtual. So from what I've heard from a lot of people that when is stupid weird stupid, they're awkward. But I think this year is going to be fully in person. So so yeah maybe save up yourself. Yeah, good stuff.

I mean, just to preface this. I'm a fan of service Technologies. I haven't been using them as long as you have, but from what I've seen is just I love getting fast up and running getting out of proof of concept as fast as possible. A low overhead with regards to operations and think about that and just really focusing on delivering value again. As fast as possible. Is that also an especially early on what you saw where the

service technology was going. Yeah, yeah so I guess the, I said, my first exposure to some Lambda with what 2015, I think that was just before they introduced the Apex Gateway integration. So at that point we're still feels like a toy. We know. Does the S3 integration there? The Canisius integration, which is okay for some stuff. But then you can't really build the entire application entire system on top on using just Lambda and things like that. But then once introduced API

Gateway that became okay. Now, you can actually build a rest API using API Gabriel Lambda and don't have to manage any infrastructure and I haven't been on not been using any basis since 2010. So I went through the whole journey of using ec2 and the containers And then Lambda and the remember back in the day I was just, no I was building very different things but then you kind of do the same thing over

and over. Every time you got a new project which is okay, if you know you decide your web framework, whatever you need right application, this bunch of a boilerplate stuff. You got to do it before. You can write your first on the business logic. Yeah, after that. Okay, there's a lot of the polling place. Have you got to do around infrastructure, you got to configure the networking which requires a premature hold University creatures to understand how different things you got to do.

Doris Auto scaling or the packaging machine images. And also, don't forget, you have to do this every couple weeks, every couple months to make this an issue that maintainer. You've got the latest security updates and all of that stuff. Which again, you know, just things that you don't have to do once you give that responsibility to a debeers.

And I remember spending spending spending about a 70% of my time, just doing a lot of that, you know, stuff, so that I can spend 30% of my time doing the stuff I should need to do. Which I mean you still quite fast compared to what I was doing before because I started my career in banking and you know we were waiting for I don't know about 69 months to get a new server into our farm and, you know, you you have this a crazy meetings with 20 people talking about, oh yeah.

I need to have this thing installed on the new server and it have that thing installed on the new server. Yeah, and once installed everyone memorizes, the login credentials, and an IP address because you going to be logging onto the server, pretty, pretty frequently. So that's like a whole year long project. Again, you serve into your farm and then going to a DPS is a

five-minute thing. Again, I get a new server, so that's already revolutionary, but then, you know, there's still a lot of boilerplate stuff. You got to do is especially as a DPS, adding more and more control. You can have around the networking and also scaling all of that stuff which is great. But at the same time and if I just want to write some web app, I don't really need all of that. And then unless I'm dealing with a super high, a scale super high, throughput basic box than

the thing that works for. 90% of cases is just fine for 90% of our cases. Yeah, and so no one slam, the claims around and became more capable in terms of being able to build API rest, apis with it and build it. All these other things with it and it just became a no-brainer for me to use it. And I haven't looked back since just because I know how much work it takes to actually do easy to is a containers and yeah now don't want to spend all of

that effort. Don don don things does not going to add any value to my customers and also doing a

project for a client. I think about 23 years ago you know the client had an idea for a social network. The first server template is that they hired a guy to do the build it in PHP and the they do something pretty quickly together and then they launched it and launch, it just kind of fell apart. They had two servers, I just couldn't handle the number of people. That's trying to login. Trying to sign up. So it was kind of a kind of embarrassing, how how bad they

went. And so they got some more funding and they had a second attempt at the second chance and they really wanted to just make sure that they get it done, right. It's get done quickly as well because it's a there behind. Yeah, it's a social network for University student. So I have to be tied to

semesters know. So, you know, so, you know, may or September when the next semester starts, everybody's going to be. That's when One's going to stop start to create an account so that they can book a sporting sessions with the University's the gym or something like that. Okay. And so now that's a pretty short deadline to get it done and the Indigent done.

And also, make sure that, you know, whatever they end up with is going to be able to handle some kind of scale, which is this is not like heavy around load, but it just in talking about now, tens or hundreds people are signing up and okay, him? The use the app the same time, Which is fine. And so, yeah, we did with app sink and the Lambda functions and things like that and that we got to having done with in before weeks.

And I was the back when I was the one man that can team was built a whole thing, including standing up the entire Oedipus organization, multiple accounts. Following a lot of the best practices around the security and an account management and then build the application. But we also build a CMS as well. Again, all that stuff, it just so easy and so quickly.

So quick to build So that we are able to get it the get it done with before a deadline and also you know launched a lost people signed up film, we've got 15,000 people signed up in the first week or two. Everything's went smoothly. There's no errors and no report for many customers. So the client was really happy I can imagine. Yeah, it's the sort of thing that we have taken like whole team much longer to do if they have to worry about all these other infrastructure stuff, worry about.

All of these things that besides what the customer is actually asking you to build, Old because you do those things to be able to you know scale to demand not have any problems when the fitness you got more than a handful people standing at the same time and things like that and we've served as you got a

really good. So Baseline for a resilience and for scalability and for security and of course you know if you have full control of your environment that, you know, very well what you're doing and dedicated to doing specific things that really well, you can do better than what you get, real amateur but for 90% of cases, you don't need better.

You just need good enough and the world and the keeps you out of the box is good enough for even better than most people can do on their own because they have to worry about so many other things besides the network security.

Yeah. Actually going to worry about actually building the application that the Quran is not your users want besides just, not focusing on security, and anybody can dedicate the entire teams of people just aren't looking after your network environment and you can do much better job than most people can. Yeah, then it's not your problem anymore.

It's not your problem anymore. And When there's a downtime me show, you still have to worry about the resilience and things like that and you get much better software based on resilience just because you get multi AZ out of the box when using things like Lambda and dynamodb and API, Gateway, where you have to provision servers in multiple regions yourself, you could work out how much capacity these have in each region.

So that if one region goes down, sir, wonder when a z goes down, you still can cover all the traffic you've got, and you've got to set up a load balancing and all these things like that. Which not besides the fact that you more work to do but you're also paying for a lot of redundancy that you're not going to be using. Yeah. And so for a lot about a lot of the applications, we see out there unless they're doing really high.

Throughput Lambda is pretty going to be cheaper, and safer and faster and better than to run your own infrastructure. Yeah, of course, you got examples like the Prime video where they got the sounds like, a very specialized workload, but have to cater for quite High throughput, which they didn't foresee at a time. So it's intense out there, okay? Because of the traffic that they weren't ended up having to serve to serve it made more sense to move the workload to containers.

And that's not that's one of the things that were once you get to certain sources report and scale, using service component is going to be more expensive because you're paying for that premium per request. So if you have a really high number of requests consistently, then the that premiums going to be more expensive than say if you were to hire someone to come in and On managing infrastructure and that's the cost that a lot of people don't see when you think about.

Alright, I can run containers for another 100 bucks a month versus running. This training on a peg where Lambda would cost me, 150. Let's just run those containers. Oh but wait, we need someone to let you do this thing can I hire guy? Yeah. Sure that'd be 5,000 bucks a month. So your actual savings $50 versus you. Actually finish a person you hire, which is Usually much more expensive but no because it depends on how much internal

skills you have people that you. Can you say, okay, you're doing containers for other teams. Already just no help us as well which could be the case for large Enterprises, which has got dedicated teams that it's going to be supported larger infrastructure footprint. They have for lots of different assets, but the for lots more companies. He just doesn't make sense to hire someone to help look after

your infrastructure. When you don't have that throughput to No to to make up the difference in terms of the operational cost on a DPS versus the hiring someone to do the same job. Yeah.

And I mean I've been a bit on Twitter as of late a little bit more sense of doing this since doing this podcast. But when the vine video example, for example, happened, people love, shitting, all over there. Like either it's not good on a cloud or they shouldn't use Lambda 2 should have used it from the start like or microservices wasn't a good option anymore. They're like look this is a prime example. They go Haywire, they go all over the place and I'm like, okay.

Well, when you start, you probably don't know what your end situation is going to really help like you iterate, you improve and this is a big Improvement for them, right? And this is Prime video, we're talking about billions of people using the service watching videos, right? That's high throughput. Anyway, that's actually the interesting thing, it wasn't Prime video. It was the one team within Prime video.

Does that build a monitoring? And and what do people like a tool that goes through the frames and the think they look for some problems, you encoding or something like that? Is one team. Not just not the whole Prime video. It's know. You also get to know Amazon is very big company. Prime video is also not small team. Yeah, so one team may be different may have decided to do. Something doesn't mean that the entire company has made way for a specific technology.

Yeah. And also even read a blog post is that you're very very reasonable what it talks about. You know you talked about how they didn't add it, is this going to be this already existing to? They're just building a new version of it when used by that many people. So they build it didn't think it's going to be high throughput. But in terms of the to was such a quiet good. So more and more team started to adopt the tool that the field is

internal tool steel. But I guess it's the, the scale comes from the number of videos you have to analyze the number of streams have to processing in parallel. So and they could be a lot. It could be a lot of, right? Because there's a lot of videos that you did that so very important Prime video and so and so you know they they they they hit a skinning problem that they didn't foresee which you know, you can't blame anyone for not being the predict the future.

And so when the the context changed a refines the the The review, the architectural decisions. And also after having run this system for, well they can specify they can spot the inefficiencies which may be easier to build a shelter at the time but now becomes less efficient. Once you have to operate a certain scale. So what they're doing you know, is textbook evolutionary architecture where you adapt and evolve as your contractor needs and your contacts change and is

the absolute. It's absolutely fine. But people just not read the text with a specific assumption in mind. The thing that they wanted to kind of look out for and they did the need, the nitpicking small bits of of context out and taking it out of context. And just all, this is the mic. I'm a sense. Moving away from microservices, the can't make sense of microservices find their own truth. Yeah. Yeah. So they say, yeah, it's very frustrating.

And anyone who's such a rate that the post for what it is, with, with objective mindset that we see, okay? This isn't it. This is normal. Yeah, this is what I expect people to do to do. I think based on what they know at a time and then and then review and the evolved architecture.

When the when the change and one of small things that people didn't so pick up on as well, is the fact that there's that one paragraph that talks about how when they migrated from things that functions to just basically have like a orchestration, run it inside one container or something like that. They didn't have to change much code because, you know, most of the code is the same. Yeah, I'm so that talks about,

okay, people that taught you to talk. About how vendor lock-in you know once you use Lambda you can move out. It's just cold. You know they were able to take the same code, maybe do some small read write and put into containers and they go, right? So it's it's it's also one of the things that talks about the people don't didn't see the fact that okay, kind of, just can't debunk, a lot of the so vendor lock-in are such a buzz term when walking such lazy term as well.

Most of the time, just two people trying to so Russian eyes. Why they don't want to do this? Don't want to learn this thing was the time. I think it's just because we're pumping too. Identities that people have built with around the tooling that use. Yeah, in terms of containers, I mean, I've learned, I spent all his time learning containers, you could be an atheist and now you tell me to learn something else. Of course, I'm going to be well, we've invested in this case of the I know.

Yeah. And and there's no Samantha pushing a different sort of Paradigm, different Technologies, kind of invalid in my skill set, my effort that I've put into learning this other thing. And okay, what's the, what's the easiest argument? I can use a stir around the, which is, when the Locking, which is the brief. Strating, because I've talked to a lot of people that time to throw the argument at me, but then they can never quite substantiate exactly what are talking about me. Yeah.

But it's actually mean, and I mean, if he were the same thing, the same sort of lock in exist with pretty much any position utility take along the way. You know, if you use this a rest of our now, I want to use the Expression as well, guess what? You clarify the application. If I want to use say, MySQL now wanted use postgres. Well, guess what? A lot of your codes. Got a good because we've written a lot of your work is Kirby has been redone.

The same thing with the moving technology from say containers to Lambda. And that's a lot of different ways. You can actually make that portability a lot easier. There's the things, like, Hassan architectures and modularity that you can do at your code level, or that is just not quite easy to. And we saw in the Prime video example that, hey, this is like a real world example, some some, some of whom Of their workload from using cell functions and Lambda 2.

Moving containers and didn't have to rewrite entire applications. Just the orchestration layer. Just some parts of the application. Yes you'd expect because of the decisions they made. Yeah. And I've had in the past that

projects, where, you know, we weren't quite sure. What one, our previous jobs, full-time jobs was working for a company called the zone is the is the the Netflix of a sport streaming. So, you know, they do, they stream live sporting events and pick the head like it. Two million concurrent concurrent viewers when I was there so bad. After I left this for you know three million or whatever by now but you know it really high

throughput very spiky traffic. So certain parts of the of the subsystem we just say you know you can't use Lambda because we're hitting skidding limits that Lambda has around the sudden burst of going from 0 to say how what was the picture we can go? No, we've immediately instantly and after that they can only go up by five hundred concurrent execution. Superb. Minute. So we are waiting that upper bounds of weird sub L skin ability of Lambda so start to

which certain limit. Yeah, but for all of our apis other desktop not on the critical path of user coming in login, and start watching, say, maybe 90 playing my city for the FA Cup. Final things like that, we can know we use Lambda and functions a lot for love the payment processing and other things as well. Yeah, and I forgot where I was going with this photographer but for years You know, it's, there's no technology that is, that's the perfect fit for every single scenario and your

contacts is not always the same. So it's a contest change. You need to know you can and you should evolve and review your decisions and in terms of the, the riot are and everyone. I was always going to go with that story. I've said some of our apis that we want to show as the time I'm right. Hit that if he's going to be end up if it's going to end up running in containers or Lambda

functions. So we would never write it so that using Ideas from a second architectures, where the Core Business logic is wrapped inside, its own modules. Yeah. And then we have wrappers for say if know that we can use inside the express tree sap or inside Lambda function that calls into our domain logic. So that way we were able to really easily. So just take the API and move into a container and running the expression years or putting to Lambda functions.

So that probability is actually really easy to do. Just a very simple code pattern that you can employ. Where was the other thing I was going to say with the with the, with the Zone example was that? Yeah. So, you know for different apis. You know you're going to have shit. Look at your different requirements in terms of both what you need to do the functional requirements but also your non-functional requirements. Like what other soft scalability

you're going to need? If it's something that's for example, for the Social Network that was users are used to worked on, they don't have super snow, they don't have zero to two million people. Requirement these times. Yeah, that kind of piqued throughput. But they do happen is lot of a really spiky traffic pattern just because a lot of the traffic is driven by say, influences doing something. So these are the social network

is built. When forward one previous jobs are in back in 2016. That was warm, I guess, my first, so Venture into service and we remove the work with the team to move a social network to run, pretty much entirely on servlets having Was running on containers before, we are lots of problems around scaling and other things. So we started to react attacked and so, fixed things that one

one piece at a time. And we thought the service was really good fit for us. Yeah, and so we use the word and I'll be using this triangular pattern to move things. Okay, one vertical at a time into those interests different services. And as we do, that rewrite, we and we architecture, we will also read enough to use the rewrite, the code into Lambda functions instead, and One of things we saw was that the traffic pattern was raised

spiky. Sometimes you get no 20x the spikes just because an influence, the post is something. And so for those kind of spiky traffic, that's an unpredictable Lambda is Lambda is really good because you got a lot of scalability out of the box, but with this is own example. The problem is that we hitting upper bound of how far Lambda is allowed allows you to scale in

one jump. Yeah. So, the certain cases where, you know the scalability, there's like a range in which There's pretty good compared to containers in the zones example, because a lot of that traffic spikes are predictable is because there's a match on at 7:00 and everybody at logs in at the 659. Yeah, so we go from like almost maybe a few hundred requests per second to know, tens of thousand requests per second with in like a twenty second, twenty second window.

So, that's kind of spike is predictable when it's going to happen, but this is the spike is so high that it is so big that we We were hitting limits on the what you can do with Lambda sirs first capacity. Yeah, so forth. That's why, you know, we made the decisions that for anything that's going to be on that critical part of and it's going to experience that kind of traffic Spike.

We're not going to put stuff in Lambda function, going to use containers instead, we have a bill control around the scaling just based on timing based on our schedule, for when the matches are going to be on. But for others, no things that worked on. Like The Social Network example, where the traffic spikes are unpredictable, but then The not so big that they were going to hit any sort of skinny limits around Lambda. So Lambda is really good fit for

that. Yeah, that makes a lot of sense.

Did you work with area, Weinberger by, by any chance at the Zone? I didn't want with him directly but I was a, but I know of him and we had some conversations afterwards as well. I think he also did a course on, I think next JS or something like that or you need to me. I think the quite well on the Summit is courses. Yeah, I had some conversation with him but I never really worried. Um direct okay, cool. Yeah, he's come on the the podcast as well.

Okay I did a job interview this was like a year and a half ago and he was sitting across from me actually okay on the other side of the table and I remember he worked at the zone as well. Some some interesting challenges with regards to traffic and stuff there. Yeah. So the zone of us are really good.

They're range think a technical Challenge and I think this this it was so there's also I don't know if he spoke with many many other at his own people at there are some really interesting stories around how it is related to a officer was put together. The haven't heard anything about that. Okay, yeah, so he was put together very quickly. I would hand out to people but then the high enough managers.

So there was some initial problems in terms of just this, like maybe like a bit of a lack of Direction. I guess that because you have too many people, but then there's not enough managers and not enough for support structure around them. Don't happen often those it?

Well, but yeah, it was. So, yeah, that there's a really interesting sort of a learning experience for me, in terms of just The scaling and organization very, very quickly and the putting a lot of things that into place things like platform teams and the hiring practices, and yeah, interview practices and things like that. Awesome. I'm thinking back because again, I'm a fan of service Technologies. I'm a fan of this movement that allows people to go to the cloud and deliver value faster.

And I don't care what service technology it is that much. Because as you said, you have to be evolutionary with diversity or technology, right? So you use what's most apt Or whichever use case you have and not everyone is a Prime video, so speak traffic like Peak traffic, you have to figure out what your own Peaks are going to be, and if they hit higher than you expect, then you adapt accordingly, right? So, the normal process, the same is with your skill set.

If you've learned Docker and kubernetes, you still have those skills, you can still try other things. They'll be afraid to adopt new technologies in that way. But then still in this era I'm again, going to reference an

article. I saw there was this article and I think it was a few months ago where there Was this guy and they said we're going to save like Millions, probably in the five or six millions by moving off the cloud this time and there was a whole post about. Okay, this specific part of our

infrastructure and disturbing. I think it was of images, specifically don't know if it was images but in any case he laid out this whole thing about the costs and the benefits of running his own racks and knowing exactly what to do, but for me, exactly the part that wasn't there was the operational cost of the people that was going to maintain this part. And I don't know if it's going to add up to the Millions.

But obviously people jumped on that the same with the Prime video example, where they said, okay maybe we shouldn't put everything on the cloud. Maybe in this case, wasn't such a good idea. Do you remember the article? I'm talking about. Don't stop that. There were a couple. There was one from. Think th the guy that founded the boot camp, and he was talking about the hater cam. I think I that's the one, that's

the one, that's the one, right? Yeah, there's another one from another company that's think they do SEO. I've got a name tag. A now, but they were talking about how they were going to stay. Would have saved. They will have spent 400 million or something like that on a TBS if they had one there's that the stack on the database instead. Yeah, that was crazy. Yeah, it's also completely wrong as well because I wasn't there

for that for the estimate. I think he just looked at just this, some background back of the napkin calculations and what car we are. This this this is how much we're paying now and based on the cost of the, the CPU door. And I'm going to get on a TBS surface of liquid equivalent service on the hourly rate is going to how much you end up spending. Yeah.

So very black and white. Yeah and also when you consider the fact that I think we're also striking at the the amount of demand money that Netflix spent on ec2 and use use, use it was less than the amount that they they reckoned. They was spent on a database for easy to and when you consider that, this is SEO company. Not not. Netflix, some pretty sure that they wouldn't have spent any idiot nearly as much as they did. They said. But, but, yeah.

And the, the a chick's article remember that, as well, he said,

he's, I don't know, he's the old one, he sir, he's a busy, no quiet, very smart guy, and share a lot of things in his career, but then the he say he's totally against his throat against the microservices. And so the in the cloud I don't really know where that's coming from. The same time, I mean, I guess some of the same arguments you see talks about how in some socket know. Microsoft microservice is not right for everyone, but nothing is the rifle. Exactly.

And I'm personally, I still think that, you know, monolith is is nice. Good choice for a lot of use cases. Even when doing Lambda function meetings with served as you know I still quite prefer no booty monoliths especially when I when I just put in the first version of something, I'm the only person working on it, it doesn't mean that they have to have. Of to have just one Lambda function or something like that because Lambda function is not a

microservice know. It's just a component that's part of your system. I'm only fair. You know, I've got one code base, one repo, one deployment for the once of system, and that system happened to be my graph qap. I'm running an app sing with bunch of different Lambda functions and down, TV tables.

That's the very prolific to me. So when I want to change something, that's one, place to go, and one Department that deploys, everything that's to me, that's not a micro services and When I've got a company with no 500 people working on the same system, I want to have microservices because the allows me to scale the team so that individual developers only touch one small part of the system. And the only need to understand

how that particular system. Works independently independently and importantly, also, if they mess up, or something happens to the system, is not going to bring down the entire application, the entire system, right? So microservices is, is really, really powerful tool for scaling the Ation understand that they know for hate account. Maybe they don't have hundreds of hundreds of Engineers that the necessitates that micro services.

And also I appreciate that a lot of the one of the most common misconceptions about microservices is that people tell me this all the time is that all we are is there to scale. Your application is absolutely not necessary application. If you don't know how to scale, I'm on the left. You're not going to know how to scale. A micro service architecture.

Yeah, and one of the third, the most scalable model if I know is I used to work for Company and they would take the gold bought by a supercell and the from that I need and I don't know. I've spoken with a lot of Superstar engineers and the softener got a glimpse of their stack which is a very very simple is one big jaw, Java application jar, they deployed to easy to read in a horizontal scaled and hold on his skin and the shore on his shot at the mySQL database.

Very simple. I to Street here application, but I did they handle millions of millions of controlling users just by scaling the model a very simple way. Horizontally. Ali. And if you don't know, those are skating patterns and apply to Individual Services because we're looking the video services within microservices is the same scaling challenge. Yes, getting model. If a sponsor and a one service

one system. And so now is absolutely not about scaling your make your application sparse, killing your organization so that you can have the hundreds hundreds of people working on the same project. The same time. If you look at no construction projects, we have 3,000 people working on the same site but that if they always just hammering the same there, You're not gonna be able to have that many people working on the same

project at the same time, right? So our service is the same way, is that it allows you to segregate your bigger building site into lots of different small area. So that different people can just work on, you know, one part, one floor, maybe a one part of the of the job like mixing the the, the material. So I don't know.

Nailing the, the boards or something like that so that you can have people working on individual tasks within the sport smaller software domain, So they can become easier to unborn new people as well. So yeah, microservices is good. For some cases may be more around your organization. Attend, the technical challenge? Are you trying to solve? Yeah, I came into a project because I mean, work at a

consultancy company as well. And there they started off with the micro Service Act architecture. And I was like, okay, interesting because I wouldn't start off with it like that, right? And then very soon, we notice that two of the services were very into dependent on Each other, and I was, the killer makes sense, because we just decided those were going to be separate, microservices. But when we needed to change one, we also need to change another, and sometimes, even a

third one. So then, we're like, okay, these are not really an interdependent. So, we just made this kind of distributed system, and it didn't make sense anymore. So, we were like, okay, wipe that off the table, we're going to start with the monolith and then we're going to see, okay, what parts need to actually be separate from different teams. And we haven't even gotten to that point. Yet is the same scale where Justice that we're even faster because we don't have to go to

different. Repositories change different things, it's all in the same spot and now when there is going to be a second separate team because that's going to happen pretty soon actually, then we're going to be like, okay, these things were going to split off and that makes sense for that team. That's how we're going to grow organically, not by deciding. Okay, this is what it's going to be in state and then working like that because that just, it

really diminished our speed. I noticed and it was a real hard argument. Because the person sitting across from me, was like, okay, it's really hard to go from a monolith to microservices and I'm like, no. The way we have a Now is really hard to operate. It and it's way easier to transition when it actually needs to happen. But that argument.

Yeah, we had to, we had to learn how to trust each other, and how to actually evolve the system and wipe some things off the table and redo, it just for the fact that we were losing a lot of speed. I felt like yeah, after seeing the quite a few times. In fact, that one of the things that if he now you've listened to a lot of The microservices Advocate, when I first thing to say is that don't start with Wanna Live. Yeah. Until you know the system, you know where the microservices?

Yeah. I start start with us. Start with monolith. Yeah. Until you you know, where the discussed the separations are, and then then then you can break our things into microservices. But yeah, the micro service is active Advocates, won't know. They don't tell you to start with microservices. Now they tell you to start with monolith, learn about system and way it makes sense to split. Then you split it.

Ya like it says, no, it's not. I mean, I just go instead straight away unless unless you're moving from a say, like is an existing system. Where you already been running this thing for a while and that's just that's the case with the Zone. Where, you know, if they the first version of it was built by a third-party third-party company. Was one big one, the left. So, by the time I joined a company and started hiring hundreds, hundreds of ha of Engineers.

We already know about a system and whether the software where the separation should be. Yeah. So we had a pretty good idea in terms of how to split the team teams at the time, were reasonably argue into an identity, in terms of how the stability and the Zone also uses a microphone as well, which is a microservices applied to the front end, which is quite interesting.

But yeah. So a lot of that is because, you know, we ended up with we're going to have a company with hundreds of Engineers. We're going to want them interview with to be able to own specific part of the system and be able to iterate independently. And also if one team have a problem with their services, don't bring down the host, the whole system, which is actually something that you have to invest in the so build with that

in mind. Yeah. Because we Is really easy to build those into servicer, the appendices that's the pipe coupling that one service goes down, you have this recorded a domino effect, cascaded Cascade failures. So you have to build those, your system with that in mind and that's one of things that I

really love about. Event-driven architecture is it's really a very natural way to to remove that tight coupling between different Services is by depending on event from each other as opposed to relying on synchronous API calls to different server. Opportunities and services.

Yeah I think the for me the hard part and I'm thinking back to this argument is experiences what drives people, right? So if you've seen something work well in the past you're going to lean on top of that experience and be like, this is what worked well in the past. And if you saw something that didn't work so well, in the past, it's hard to be like objectively, look at, okay. Why was it that? It didn't work that well, and why in this case, can it work

well, can it work better? What can we do differently? You can still make it work. I feel like that really drives a lot of arguments and in this Ace maybe even with the example of interlocking, it really drives a lot of arguments and then it fails or some people fail to look at it objectively and be like, okay, maybe in this case we can do differently, maybe we can do the same, but have it be a success instead of a failure in the past? Yeah, I think that's one thing

that's missing from a lot of the online content is the people give you the the hell. I know, something happened. So we change but like I was right? Look at it. Yeah. But I didn't give you the context of why? Yeah. And, and that's the one that has the task does a bit. I should look out for that. Know, why did that work for you and not for somebody else? Because something always works for somebody enough for somebody else, you know, medications, the same thing.

You know, I think we talked about intermittent fasting. I can do one meal a day for a 10 ounce. Maybe you can, or maybe maybe I can eat. Avocado, you can't. Yeah. So it's nothing is always going to work for some people and not others. So that context is is how we go. How you go beyond the surface or just Pattern surface that you know, that naive fantasize that this thing doesn't work.

Yeah to say okay understand when these Frameworks and that's something that I think is missing from a lot of the content out there and that's what I want to things that I've had a few colleagues and Friends cannot come to me and asked me to serve, you know, review their blog post. If that's one of the things I was always so kind of feedback to be ghetto. People who are starting to to block is going to the explain why explain the context, I like

that. A lot of people they go straight into The solution to say, here's why I did hear some know. Here's the, my beautiful elegant solution and one of the first feedback I often, get give, you sir, sell the problem first, and you people care about the problem, you know, you're trying to trying to solve. They're not going to really care about your solution or they're not gonna understand why your solution works and for you but not for them when they tried to copy it. Yeah.

And you always sort of brings me back to that that that that scene from Wolf of Wall Street where you know his name, The Punisher actor. Who's he He's got a, he was that they were all sitting at a table and and DiCaprio's character asked him to know. Sell me a pen. He doesn't isn't going to describe how to pan is awesome. You know, tells you oh yeah, what forgot, what were your first is that? It was a whole story where the end he actually needed the pan and then this was the pain.

Yeah, he created a need for a pen. I think something right down like a check or something like that. I'll give you a check but independent writer, something like that. Yeah, so he created need for the pain and then he starts out of the pen and that's something you need to do for, you know, your content as well. Create a need for running for for your solution before you try to sell it to solution to your

readers. I mean, it's something that's missing it at from a lot of content out there but waiting to see them that they know what they're really good at. Okay, explains, here's my see, here's my situation. He's my context. And this is the problem that we had and the solution. Here's our solution and it works for a context because we are large company.

Mike Services works for us because we need to be able to allow people to come in and just work on one part of system without having to know everybody and everything else that's happening in the system and be able to have an application that is going to be resilient to small failures, too small part of the seat of the application and the actual, some nice examples, as well from what happened. Wait, don't do that properly. One of the two companies I used

to work for and they have this very day went through, kind of the microservices. So Revolution against large and large company not Enterprise, but not okay, amateurs, stop with lots of users and the last Engineers, they kind of did their microservices thing on their service boundaries but then they were still using a shared database. And so there was one time when the Sphinx, someone from the customer support team did something in, like, internal to,

which ran a really heavy query. No, took down a database. Yeah. And what because they didn't have a read replica for for the, for the custom support to so. And then that because everybody, a lot of different Services was using the same shared database. So you've got these tributed model if. Yeah. So it took down the entire system for the whole weekend. The funny thing is, you know, these a food delivery service. So you know the cost of moving for customers is really low.

So all the customers went to another food delivery service and funny thing is the other services also wasn't doing very good. In terms of architecture. So it took down the outlet that other services as well. So I guess the farm business point of view. They didn't really suffer too much from that outage but it was enough to for them to rethink a lot about the architecture and the actually hire a really good person. I know into the company just to fix that.

A lot of that problems that they didn't the that it had displaying for. But yeah. So yes again, that's kind of shows, you know, one of the problems that you have with know when people I don't really understand what microservices and just doing this of surface level, you know, we just had a service but they're not really following through and understanding that. Okay? We have microservices Services, should be autonomous, they should be independent.

Yeah they shouldn't be sharing the same database. Otherwise you can create this implicit coupling and the single point of failure which is one of the yeah. Really. I sister but maybe not nicer but very interesting story that I've had to have formed from that. Yeah, I can imagine I mean it's the Same with people looking for Solutions, right? Or explain the solution more so than explaining the context is also. Okay, I can put a micro servers up but it's really needed.

Like why do I need this micro service in the first place, right? And do I understand why microservices or how they work in the interdependently from each other? What a, well, well laid out architecture would be if someone runs a query on the database is going to Cascade and it's going to like, flatten my whole system or do we have a replica for that?

For example, you have to understand why and when to use WhatsApp But that's not the fastest way because the fastest way it's just the use the solution you think is right? Yeah. But I can have severe consequences. Yeah. So that's the so that's that's something that's hard to integrate integrate with were all of my yourself a Content just know making sure that I always explain the context and try to do as much work as good a job as I can to explain the why

the solution works. And we know when the solution works because stuff, man, this was a lot of fun. Yeah, I'm gonna round it off here. Um thank you so much for coming on. I'm going to put all yawns socials and links to his blogs in the description below. Check them out. They'll know you came from a show and with that being said, thank you for listening. We'll see you on the next one. Thank,