AWS re:Invent 2021 - Building a serverless banking as a service platform on AWS

Aug 16, 2023

AWS re:Invent 2021 - Building a serverless banking as a service platform on AWS

Solarisbank, Europe’s leading Banking-as-a-Service platform, enables other companies to offer their own financial services. Via APIs, partners integrate Solarisbank’s modular banking services directly into their own product offering. By migrating all of its core banking systems, digital products, and databases to AWS, it became the first bank in Germany to full migrate to the cloud. Join Dennis Winter, VP TechOps, as he shares how their all-in on AWS approach allows them to speed up the delivery process and drive more innovation, while simultaneously allowing the company to follow their internal approval processes to ensure compliance, security, and integrity across the overall architecture.

Learn more about re:Invent 2021 at https://bit.ly/3IvOLtK

Subscribe:
More AWS videos http://bit.ly/2O3zS75
More AWS events videos http://bit.ly/316g9t4

ABOUT AWS
Amazon Web Services (AWS) hosts events, both online and in-person, bringing the cloud computing community together to connect, collaborate, and learn from AWS experts.

AWS is the world’s most comprehensive and broadly adopted cloud platform, offering over 200 fully featured services from data centers globally. Millions of customers—including the fastest-growing startups, largest enterprises, and leading government agencies—are using AWS to lower costs, become more agile, and innovate faster.

#AWS #AmazonWebServices #CloudComputing

Content

3.04 -> That's how things start.

4.14 -> Hello, again, my name is Dennis.

6.72 -> I am Vice President Technical Operations at Solarisbank,

11.84 -> a leading banking-as-a-service provider in Europe.

15.76 -> And in the next roughly 40 minutes, 45 minutes,

19.47 -> I want to tell you a little bit about how we started out

23.097 -> and how we eventually ended up in the cloud.

28.17 -> So first of all, before I dive into our story,

33.8 -> I want to tell you a little bit, a couple of facts

35.78 -> about me, so that you know who's actually

37.39 -> telling you this story.

39.41 -> So I'm 42 years old, I have two kids, 15 and seven,

44.2 -> and I consider myself as an engineer by trade.

47.39 -> So I basically grew up in my grandfather's workshop.

50.22 -> He was an electrician and for those of you who know that

53.11 -> in the late '80s, early '90s, computer came without a GUI.

58.419 -> That was also my first interaction with source code.

61.55 -> So there was a program that was called "Gorillas,"

64.58 -> it was a game, and I just,

66.41 -> being nine, 10 years old, just fiddled around

69.68 -> with the source code, because you had

71.14 -> to execute it explicitly pressing a five.

74.39 -> And this is basically what I tried out.

76.167 -> And I won't tell you that I didn't know anything

77.86 -> about backups back then and had actually to wait

81.61 -> for the next computer that my grandfather would buy

85.74 -> that Gorillas.pass would actually work again.

90.5 -> So at Solarisbank, I joined in 2016

94.6 -> and I'm in my fourth role in the meantime.

97.94 -> So looking at the development steps that we had

101.94 -> at the company, I hope that some of the things

104.38 -> that I'm telling you today will maybe help the one

107.3 -> or the other person who are

109.47 -> in similar challenges that we have been.

111.86 -> And for this, let's have a short,

115.44 -> let's have a short look at the agenda.

118.13 -> So I broke it a little bit apart in three pieces.

121.7 -> The first part of my talk will be how we started out,

125.48 -> how we bootstrapped the company,

128.75 -> the immediate challenges that we saw back then

134.182 -> be it because we were a fast growing company

137.3 -> or also because we were just exposed to the regulations

141.1 -> from the German regulators as a fully licensed bank.

147.15 -> Afterwards, I will tell you a little bit

149.55 -> about how AWS helped us in resolving some of these problems.

154.77 -> And in the last part, I will tell you a little bit

157.16 -> about where we stand today

159.49 -> and what our plans are for the future.

164.95 -> In order to bring you up to speed for the stuff

168.57 -> that I'm telling you, I want to tell you a little bit

170.21 -> also about what Solarisbank is actually doing.

173.66 -> So as at Solarisbank, we consider ourselves

177.47 -> as a technology company with a banking license,

181.81 -> which means we are not the regular bank.

185.14 -> Like, you cannot just come to Solaris

186.89 -> and open your individual account, come into our offices.

189.82 -> You need to do this with our partners.

192.73 -> I will tell you a little bit more about this in detail

194.88 -> in the next slides.

196.34 -> So we have a business-to-business model,

198.913 -> a B2B2 access, we call it.

200.92 -> So Solarisbank is providing technology

205.04 -> that can be embedded from other companies

207.84 -> into their respective products.

209.61 -> So they enrich their products with financial services.

215.54 -> In the meantime, there are 650 employees.

219.05 -> We are actually more than 700, I looked it up.

220.92 -> So when I created this presentation,

223.47 -> that's a couple of weeks back, we are growing rapidly.

226.29 -> We have more than 60 nationalities at the company.

229.09 -> It's a very diverse company.

231.85 -> And just this summer, we closed our D round,

235.99 -> or our last funding round, which send us directly

239.66 -> into the Unicorn club.

243.34 -> And we acquired one of our competitors which gave us

246.9 -> also access out of continental Europe into the UK.

253.13 -> As one special thing that we are really proud about is that

256.75 -> we are operating out of five different countries

259.18 -> in Europe in the meantime.

260.08 -> So we used to be a German company,

262.55 -> in the meantime, we are an international company.

268.56 -> Now a little bit more about the actual idea

272.65 -> behind Solarisbank.

274.17 -> The word contextual banking has been floating around

276.95 -> at the company for two, three years now.

280.2 -> And what it means is that we truly believe that

283.77 -> from like, it's already happening and in the future

286.46 -> it will accelerate, that companies will integrate

290.96 -> granular financial services into parts

294.34 -> of their individual service offerings.

296.9 -> Which means back then, a couple of years back,

299.59 -> when you wanted to buy a used car online, you had to also,

303.05 -> additionally, in parallel, apply for a loan

305.85 -> if you didn't have the cash in your bank account.

309.17 -> What we enabled companies to do is to integrate the decision

314.17 -> from a bank, whether you are getting this loan or not,

316.75 -> into the checkout process.

318.5 -> So the whole identification of a person

321.07 -> and the whole decision, the scoring of a person,

324 -> happens in the background in a manner

326.45 -> as it wasn't possible a couple of years back.

329.57 -> And in the meantime, we have 50 products,

333.47 -> 50 different products, depending on how you combine

336.64 -> our APIs, that can be integrated through RESTful APIs,

341.79 -> into the solutions that other companies create.

346.67 -> So if you look into the overall trend that we are seeing

350.29 -> already for a couple of years now,

352.39 -> Apple has been doing different things.

354.65 -> In the meantime, you get an Applecart

356.63 -> and you can, with Apple page,

359.14 -> just buy at the point of sale.

361.48 -> Same for Google, they are actually,

363.08 -> like, they used to be a search engine

365.07 -> and they do many other things.

367.41 -> And why do they do this?

369.2 -> Because people actually expect nowadays

372.2 -> that things like this are possible.

374.09 -> It's a diversification of the product portfolio,

376.67 -> but it's also, you're allowing people to conveniently live

380.34 -> their lives in a manner as they want to.

386.603 -> Solarisbank's role in this is, if you look

389.16 -> at the sandwich, we are offering these five,

392.89 -> six product groups as a white label solution

398.57 -> in the background without interacting directly

401.59 -> with the end-customer.

403.12 -> All this came up because a couple of years back

406.13 -> our parent company, the company that founded

410.52 -> the Solarisbank wanted to build FinTech companies

415.15 -> just had this experience that this was impossible

417.47 -> with existing banks and we wanted to be different.

424.18 -> Okay, so telling you now a little bit

427.18 -> about the early days and the challenges that we had,

430.46 -> and how we overcame them.

433.72 -> So when I joined Solaris, we were 25, 30 people

437.93 -> and everyone had to be a superstar, a rockstar.

441.21 -> And how do you do this?

442.44 -> You motivate people.

444.04 -> And that was one of the core enablers for the success

446.99 -> of Solarisbank, that the founders, the early founders

450.42 -> really gave everyone the impression that

452.91 -> you are exactly at the point where you're supposed to be

456.53 -> to really help us be successful.

458.27 -> No one else would be able to do this.

460.47 -> We were all N-preneurs, not only the founders,

463.84 -> and that motivated everyone immensely.

468.13 -> Back then, it was still unclear

470.24 -> what our actual product would be.

472.04 -> It was just clear that, okay, we were granted

475.3 -> the banking license, with the banking license,

478.86 -> we applied for it.

481.34 -> But how, in particular, the products will look like

484.29 -> that we built first was something that we had

486.91 -> to test out in lead manner with the market.

489.67 -> So building up some foundation and at the same time,

494.91 -> checking just the market need was one

496.95 -> of the first things that we did.

498.45 -> And the market need was just immense.

502.28 -> Like, we saw that a lot of companies were out there

504.47 -> that were craving for another company

506.95 -> that would enable them to provide financial services.

513.61 -> Of course, we built POCs and we built monoliths.

517.81 -> And some of these monoliths also ended up

520.22 -> to be version 1, 2, 3 that we're running in production.

524.87 -> We all know this, usually you just want

526.49 -> to try something out and then build it in a better manner.

529.91 -> Also, in our case, we had to somehow live

533.06 -> a very lean approach in there and see

535.54 -> that we stabilized things as we go.

539.22 -> We already tried to ensure that,

541.48 -> from an architectural perspective,

543.12 -> and you need to imagine that,

545.41 -> the tech team grew basically by 100% each month.

551.41 -> We onboarded an immense amount of people

554.54 -> and onboarding not only means giving them access

557.55 -> to systems or to source code, and to repositories,

563.52 -> you need to tell them what to do.

565.68 -> And that was, of course, because it was the early days,

568.55 -> very challenging; we didn't do that well.

571.04 -> So why we tried to create architectures,

574.44 -> software architectures that already had some breaking lines

577.52 -> in there where we could say later on, we will break out,

580.03 -> for example, an authentication out of an existing service,

584.54 -> that part of the code had already changed

586.85 -> when we were just about to like,

588.8 -> somehow get it out because other people

591.08 -> had already implemented changes in there.

593.54 -> So catching up with the real world was a challenge.

597.86 -> What we did back then is that we said, we need to do,

601.26 -> we need to create a platform where we will be able

604.69 -> to horizontally scale our coding

609.79 -> or our software development, our product development.

612.79 -> And the notion in 2016 in Germany was

616.75 -> if you go to a hyperscaler from the United States,

619.07 -> the chances that you will get into trouble

621.33 -> with the regulators are pretty high.

624.18 -> So also there, let's come up with an approach

627.81 -> that gives us all possibilities later on, but start on-prem.

632.36 -> So we built an on-prem system,

633.96 -> an on-prem system completely based on Linux

637.81 -> on a lot of HashiCorp tools.

639.28 -> Thanks guys, I saw them yesterday somewhere here.

642.99 -> So they helped us a lot or their services helped us a lot

646.49 -> and everything was dockerized.

648.14 -> It was clear that we needed to come up with some kind

651.04 -> of a standardization, a concept that would give us

654.73 -> the possibility to, at some point in time,

657.02 -> exchange the sub-jacent system and do something

660.42 -> more efficient, better.

661.98 -> But for the time being back then, the solution was

664.73 -> on-prem Linux, and a lot of HashiCorp tools.

669.82 -> This is the system that we came up with eventually,

674.29 -> because we had to ensure integrity.

677.38 -> That is the most important part,

679.93 -> or one of the most important parts

681.53 -> when you're in a regulated environment.

684.42 -> The question from auditors, from regulators,

686.94 -> from officials, will always be how do you ensure

690.2 -> that you know what gets rolled out in production?

693.81 -> So this is the concept.

695.67 -> We called it Platform-enforce principles

698.23 -> that we follow until today.

700.9 -> Because what we made was a system that would

703.26 -> allow us to fully control the changes

707.81 -> that are introduced into the system.

709.42 -> If you look at the very left, there is the circle

711.75 -> with the Dev, that's the developer or our developers.

714.67 -> And the only interaction points that they had

716.7 -> with the overall system was GitHub Enterprise.

721.6 -> Concourse, for those of you

722.97 -> who don't know it, it's a CI/CD tool.

724.85 -> It's easy to configure it or it's actually supposed

727.98 -> to be configured through text files.

731.65 -> So also there, we have the possibility

734.405 -> to apply some principles, to enforce

737.5 -> four-eye principle, for example.

739.41 -> And that was a core enabler for us to show that

742.52 -> we are not only controlling the source code itself,

745.5 -> but also the environment.

748.672 -> I look into this, there is a similar picture

750.67 -> a little bit later, where we dive a little bit deeper

752.54 -> in what happens there.

755.04 -> So how we called this was compliance in code.

757.46 -> So compliance is usually something that is perceived

759.96 -> as it's important that you do it, but it's additional work.

764.47 -> And if you're a software engineer, if you're a product guy,

768.144 -> you just want to get your next feature out.

770.25 -> And compliance is something that, ideally,

773.21 -> it comes automatically.

774.61 -> But the sad truth is, it doesn't really happen.

778.73 -> So what we tried to establish is a mindset

783.3 -> of this is a chance for us.

785.88 -> Compliance is not something that should be considered

788.19 -> as a burden, but it is something that tells you

792.33 -> how to do things right.

793.67 -> Because I think most of you also have the experience

796.18 -> in software companies, the one thing that was

799.97 -> usually not in the best shape was the documentation.

803.92 -> And then that became a problem at some point,

807.36 -> and basically, compliance is not only documentation,

810.93 -> but it defines that you need to make sure

813.13 -> that your documentation is up to date.

814.46 -> And in the end you will also benefit from it.

817.31 -> So question is framing, what is compliance

821.02 -> and what is security?

822.56 -> And if you take it as an opportunity to really think

825.87 -> about things and make things cool and better,

829.46 -> and efficient, it's actually a creative process.

833.62 -> It's a creative requirement for people to do it.

838.95 -> And that's also what I tell people who are

842.01 -> in interviews with Solarisbank and who ask me,

844.407 -> "Wait a second, so you have a banking license?

846.667 -> "Am I actually, like, can I work the way

848.557 -> "how I used to work before?"

850.67 -> It's like, yeah, of course.

851.92 -> I mean, we do things in a certain manner,

853.9 -> in a defined manner, but we also need

856.38 -> to be more creative than others in order to ensure this,

860.23 -> that we can have this culture of, you know, allowing people

863.85 -> to fix problems in production, for example,

867.18 -> and not having this one team that is certified

870.55 -> to jump in if things go wrong.

872.68 -> That's not scaling.

874.61 -> So just to give you an example, and I'm circling back now

879.35 -> a little bit to the image that I showed you earlier,

881.92 -> these are some of the topics that our auditors

884.67 -> are checking us on.

887.68 -> So, and that's the list, and I'm not lying,

891.35 -> that came up when I was sitting at my kitchen table

893.6 -> in like, maybe 20 minutes and I could go on.

898.12 -> And just to give you an example of how most of these things

903.02 -> can be already covered by just one or two witty ideas

907.39 -> on how to do things, I want to talk

909.64 -> about the change process, the way how we introduce changes

912.81 -> into production in a completely compliant manner.

918.52 -> Back here, the engineer got a face in the meantime,

921.94 -> and let me just show you a bit how the systems

925.26 -> are operating when changes are introduced.

928.9 -> The engineer or end engineer is interacting

931.81 -> with GitHub Enterprise.

932.79 -> We have Terraform.

935.13 -> We use Terraform to configure GitHub Enterprise

938.76 -> in a standardized manner.

940.17 -> We protect the main branches.

942.33 -> Nothing gets into the main branches

943.94 -> without a proper pull request

945.8 -> and an approved pull request before it gets merged.

949.18 -> That's the work for the engineers.

951.56 -> Additionally, they create releases.

954.08 -> They are the ones who decide this specific merge

957.35 -> into the main branch is supposed to be released now

961.43 -> into production and into our sandbox environment.

964.9 -> That's their responsibilities so far.

967.31 -> If changes happen, our CI/CD tool, Concourse,

970.49 -> is doing things and these things are

973.02 -> not only building and testing,

974.63 -> that's the obvious part of what CICD tools are doing.

977.81 -> The cool thing is, if you look up CircleCI online,

981.45 -> it says something like your automate everything tool

985.01 -> or something, and it it's really like that.

987.4 -> You can use Concourse to execute scripts

991.99 -> or to define scripts that you execute in different manners.

995.87 -> So what we do during our rollout process is that

999.54 -> not only we are testing assets,

1001.07 -> not only we are tagging assets,

1002.65 -> everything is dockerized.

1003.74 -> Remember?

1004.78 -> We are not only tagging assets that are matching

1007.72 -> the Gitshot of a specific change,

1011.62 -> we are also creating tickets in Jira,

1014.41 -> and we are using these tickets

1015.74 -> for additional triggers within this process.

1020.39 -> The actual requirement in Germany from the regulators

1024.38 -> for banking-related functionality is

1027.75 -> you need to ensure that you have a technical review

1030.66 -> and a functional review.

1033.26 -> Which makes sense because usually you have someone

1035.68 -> who understands the domain, someone who understands

1037.9 -> the financial systems, the transaction networks, and so on.

1042.07 -> That's the functional review.

1043.95 -> And you have a tech person who understands the code

1047.86 -> that is supposed to be introduced.

1050.08 -> Now in traditional banks, this leads to release trains

1054.86 -> or bigger releases because you have to go

1056.68 -> through this effort to get these approvals in

1059.61 -> and people to understand what's actually happening.

1062.35 -> We are rolling out more than 300 times per week

1066.714 -> and we'd really want to keep it like that.

1068.83 -> We want to allow the product development teams

1072.55 -> to really be quick and fast in releasing new versions,

1075.95 -> to fix things, but also to, of course,

1079.01 -> deploy new functionalities that the market is waiting for.

1082.56 -> And how this process in detail works is

1086.54 -> the PR is approved and merged from all these merged changes

1092.25 -> into the main branch.

1094.29 -> The engineers can create releases.

1097.42 -> If such a release tag is created in GitHub Enterprise,

1100.56 -> it automatically creates a Jira ticket

1102.9 -> on a Jira project where only the product owners

1106.53 -> have write access to.

1108.82 -> They got a new ticket.

1109.95 -> This ticket contains the information on the release.

1113.86 -> This ticket is in a waiting state.

1116.96 -> And they can just drag and drop this ticket

1119.87 -> to approved or rejected.

1122.88 -> If they put it to approved, changing the ticket

1125.31 -> also triggers again, our CI/CD tool, which automatically

1130.1 -> rolls out exactly this specific version

1132.33 -> of the Docker container

1133.98 -> into the respective AWS account into production.

1140.88 -> And by doing this, we can prove the full

1144.31 -> or we have the full audit trail.

1145.69 -> We can fully prove who had, at what point in time,

1149.77 -> a look at the change and we have directly the names

1153.46 -> connected to the respective approval.

1155.96 -> That's something that's proved as bulletproof

1159.43 -> when we were talking with the auditors and regulators.

1162.68 -> Explain it to them, we were able to show them also, like,

1166.23 -> data, like, based on data, that all the changes were

1169.24 -> really went through these steps.

1170.98 -> And that's something that we will refine it even

1173.1 -> in the future, but at the moment, we can just concentrate

1176.83 -> on other things because this is solid.

1181.31 -> So now AWS Quite had an impact on Solarisbank

1188.769 -> and on the way, how we do things.

1191.95 -> Back in 2019, we were 12 products engineering teams.

1198.56 -> We are more than 20 in the meantime.

1201.57 -> And back then, we had one platform team.

1203.95 -> You remember, we were still operating on-prem

1207.9 -> in distributed system based on HashiCorp?

1213.96 -> And the problem that we had back then was that

1217.53 -> the platform team was basically responsible

1220.06 -> to provide everything that had to do with the persistence.

1223.53 -> You cannot just allow people to create databases

1226.17 -> in a manner where the databases need

1228.1 -> to be set up themselves as well, to be highly available.

1232.11 -> Also, just in terms of giving access to the crown jewels

1235.43 -> of everything that we have is something that

1237.25 -> we couldn't really allow teams to do themselves.

1242.774 -> That would have been not compliant.

1244.92 -> So we had outgrown our concept spectrum.

1248.63 -> 12 teams were just too many for the platform team

1252.18 -> to really be like a service provider for,

1254.64 -> so we had to somehow come up with a new solution.

1259.89 -> So status quo, what was there?

1263.67 -> We were dockerized back then.

1266.4 -> That was good so far.

1268.44 -> We had a lot of standards established and templates.

1272.12 -> So all the teams used base images

1275.4 -> that already came with certain functionality to,

1278.09 -> for example, register in console mode.

1281.1 -> And so that helped us speed things up but it was,

1285.53 -> at that point in time, not enough anymore.

1289.1 -> All the services were stateless.

1290.65 -> That was something that we also pushed the teams to do,

1293.06 -> just to be aware of.

1294.88 -> Having stateful services, it's just something

1296.66 -> that keeps you from really doing the good things

1299.07 -> at a later point in time.

1300.72 -> We had databases that followed a standard,

1304.17 -> so MariaDB, mySQL, that's something that is just known,

1306.9 -> and it was clear that whatever we would be using

1309.12 -> as a new solution, we would be good

1311.95 -> with MariaDB and wouldn't have to change

1314.32 -> also the services, themselves.

1317.34 -> And we have just a traditional in-service communication,

1320.35 -> RESTful APIs, usually synchronous,

1323.94 -> some of them asynchronous.

1326.52 -> And we knew we had to go to the cloud.

1329.25 -> We looked at the cloud already,

1330.51 -> like, drooling over the possibilities

1334.55 -> that we would have as soon as we get rid

1336.29 -> of our on-prem solution.

1338 -> So something had to happen.

1341.65 -> The choice was pretty clear because we already had invested

1346.1 -> some time to make, let's say a small comparison

1350.38 -> between the providers that were out there.

1353.28 -> And AWS was clearly our favorite

1356.29 -> because of its reliability.

1357.61 -> Its been there, AWS has been there for a very long time,

1360.41 -> in the meantime, and it has a good reputation

1363.7 -> on the European market and the German market.

1366.52 -> The quality, of course, of the products,

1368.41 -> because it has been there already for a while,

1371.28 -> if you just think about the possibilities that you have

1373.51 -> with RDS, if you know how painful it is

1376.21 -> to set up a CMOS synchronous mySQL setup,

1381.44 -> that's just something that you don't really wanna have.

1385.98 -> It's obvious the geographical possibilities is,

1389.58 -> of course, given with AWS.

1392.56 -> And a very important point for us was talent,

1396.35 -> which means if you're looking at the market

1398.83 -> for new engineers, the amount of people

1401.76 -> that already had encountered AWS at some point in time,

1406.22 -> it's just really high.

1407.47 -> And people really like working with AWS.

1409.54 -> So it was a clear choice.

1413.76 -> The migration.

1416.24 -> In 2019, we got a new CTO who came from Capital One.

1421.67 -> And he already knew, Capital One being like,

1424.81 -> one of the biggest AWS partners or clients,

1431.16 -> he brought in a push for us to really also invest this time

1435.77 -> and this effort to go into AWS because there was just

1439.36 -> so many things and so many requirements

1442.28 -> that came in just from regular business,

1444.31 -> that it was always hard to negotiate, basically,

1446.95 -> that now we need to start doing the big work.

1451.06 -> So it was clear that we had to do a plan,

1455.29 -> that we had to come up with some idea on how we can

1458.35 -> centrally track the migration of the whole company,

1460.93 -> back then, 300 services, including databases

1465.52 -> that had to be migrated over to AWS

1468.45 -> and a pretty packed agenda for all the teams.

1475.41 -> So what we came up with was, and I just want

1478 -> to point out maybe one thing out of here,

1483.01 -> as I mentioned, it was like a longer process

1485.44 -> to really come up with it.

1487.7 -> We just decided that there was a strategy, for example,

1489.97 -> going hybrid because lift and shift in a business

1494.95 -> that is very fast, it has a very high pace,

1498.04 -> it's just something that you cannot really afford.

1500.59 -> You cannot just pause, move everything over,

1503.35 -> see that it's working.

1505.65 -> It was clear we need to go into a hybrid mode.

1509.12 -> And we created a translation metrics

1510.78 -> and I won't go through all of it,

1512.3 -> but the translation metrics, and I will show you this

1514.2 -> on the next slide, was one of the core enablers.

1518.04 -> We thought we need to tell the teams

1520.19 -> how the solution will look like as soon as we are moving

1523.07 -> out of on-prem because they were used to think

1526.15 -> in a certain concept, in a certain manner.

1529.46 -> And the first thing that usually humans react

1532.42 -> or the way how humans react if you tell them we just,

1535.35 -> like, everything different from tomorrow on,

1537.46 -> they get anxious.

1538.7 -> So making sure that they really understand what's happening,

1542.14 -> making sure that they get like, a tangible idea

1544.57 -> of how it will look like as soon as we are

1546.11 -> in AWS was one of the core deliverables

1550.27 -> during these sessions.

1552.27 -> And if we look into the translation matrix,

1554.6 -> this is roughly how it looked like.

1555.99 -> It was not hard to do, it was just,

1558.15 -> it had this one decision in there where people were asking

1560.73 -> for Kubernetes back then, where we said,

1563.137 -> "Nope, our container orchestration will be ECS."

1566.38 -> Because it will be just easy for us to, instead of deploying

1569.56 -> to an existing system that digests target containers,

1573.81 -> you're just going directly also to ECS

1575.68 -> because it does the same.

1577.46 -> So some of these decisions were, of course, taken

1580.74 -> before we came up with the matrix.

1582.55 -> But in general, that was a very important asset

1586.25 -> for us to prepare the migration and also to steer

1590.03 -> the migration throughout the following months.

1595 -> It was mayhem and I mentioned it earlier,

1596.93 -> the stress was quite high because we are a startup.

1601.49 -> And that means that besides the actual migration

1605.68 -> that we wanted to finish within one year,

1609.37 -> we also got the deal with Samsung.

1612.75 -> And Samsung decided that they would

1614.66 -> completely base Samsung Pay on our systems,

1617.5 -> which was a huge deal for us.

1619.69 -> And we could just say no, just because we had

1621.45 -> the AWS migration also planned.

1624.7 -> The next thing was that, roughly the same time,

1628.12 -> we also got to deal with a company called Vivid Money,

1632.6 -> which is a spinoff of Tinkoff, a Russian bank.

1635.63 -> Also very ambitious and also big deal,

1638.09 -> very important partner for us, also nowadays.

1640.88 -> And besides all that, we had outgrown

1643.76 -> our core banking system, which was a core banking system

1647.14 -> based on SOAP, based on Windows setups,

1650.82 -> and it was slower than what we would actually need,

1653.99 -> in particular, looking at Samsung and Vivid.

1656.95 -> So we were actually running four huge projects

1660.3 -> at the same time and I can tell you, and Corona is,

1664.65 -> by the way, the name of our own core banking system.

1668.2 -> No one would know that couple of months, years later,

1671.57 -> we would end up in a situation

1672.77 -> where it means different things to people,

1674.45 -> but we just stick to it.

1676.5 -> But during this time, 2020, it was stressful

1681.44 -> but what I showed you earlier, having this central plan was

1688.04 -> definitely something that helped us ensure

1690.47 -> that we would be successful in moving over.

1693.6 -> So the core ingredients for the success were,

1695.85 -> as I mentioned earlier, the standardization.

1698.29 -> We established, internally, the whole cultural work

1700.59 -> in AWS Guild, so that people would have the possibility

1704.16 -> to play around with, you know, the tooling that is there

1706.86 -> in AWS, that they could exchange knowledge.

1710.76 -> I personally was talking, I think,

1712.87 -> every second week about AWS in 2020.

1716.45 -> People, like, when they saw me, they knew someone will

1719.02 -> like, he will now start talking about the progress

1721.19 -> in AWS and how important it is,

1723.06 -> but it was also important for the whole company

1725.72 -> to really understand that what we are

1727.13 -> doing here is important, even for parts of the company

1730.04 -> that might not interact with tech too often,

1733.92 -> but making sure that everyone understands that

1736.83 -> this is a huge step forward.

1739.49 -> And this is something where we need to support,

1741.38 -> where the whole company needs to support

1743.08 -> also tech and product to really make this,

1746.88 -> that was a core deliverable for this.

1751.74 -> We brought in external consultancy because if you think that

1754.7 -> you are the one who knows just everything best,

1756.39 -> that rarely works.

1758.66 -> So we got consultants in who helped us really understand

1761.28 -> whether the concepts that we came up with, for example,

1764.29 -> for the AWS accounts, the whole setup.

1767.69 -> Also we needed to have like, the standardization,

1770 -> of course, in place, also for AWS and in the future,

1773.47 -> someone who would make a cross-check

1775.49 -> and give us hints in how to improve things.

1778.55 -> And we introduced the buddy program,

1780.76 -> which meant for all the product teams

1783.78 -> that were supposed to move their own services over to AWS,

1787.7 -> because you cannot do 300 services through like,

1790.49 -> 5, 3, 4 or five people who would just

1793.15 -> like, take over this responsibility,

1795.39 -> you need to build up this knowledge within the teams

1798.52 -> and at the same time enable them really to do it themselves.

1801.35 -> And if they have questions, ask their buddy.

1804.56 -> So the four or five people who would usually do this,

1806.53 -> maybe in other companies were, for us, the AWS buddies,

1810.47 -> who would just be the bearers of more knowledge

1813.9 -> and who would sit together with the product teams

1816.42 -> and help them do the changes in case that they needed help.

1822 -> What it also allowed us to do is

1824.39 -> the definition of resiliency tiers.

1825.89 -> And that's something that really helps us tremendously now

1828.89 -> to boost the whole BCM topic within the company.

1832.67 -> Because if you're growing rapidly, compliance and so on

1836.12 -> also comes with requirements regarding risk,

1840.52 -> the concept or having different standards

1843.73 -> for different criticalities

1845.16 -> for different important systems has been something

1851.4 -> that allowed us to come up with a completely new idea

1855.36 -> on how we would deal in case of disaster.

1857.63 -> That's something that we, of course, I mean,

1859.32 -> this takes quite a while, that's something

1861.18 -> where we are still in the middle of, but this level

1865.01 -> of depth is something that just

1867.26 -> only the cloud allows you to do.

1871.08 -> So this, in the end, gave us the possibility

1874.64 -> to really move over to AWS with all our product services

1879.12 -> within the scope of, it's actually 11,

1880.68 -> it was 11 months, but 12 sounds better.

1884.42 -> It was 100 engineers who really built up this knowledge

1887.07 -> besides their usual tasks that they had.

1891.49 -> And the question was after we've been there,

1895.07 -> how do we bind this?

1896.45 -> How do we make sure that now as we got the migration,

1899.9 -> this is only the beginning, this is,

1902.45 -> what happens now is we enable the teams

1904.53 -> to do stuff themselves.

1906.23 -> So AWS helped us, or the two systems in AWS

1912.32 -> help us to be quick to scale our business quicker

1917.09 -> than before, because we removed the dependencies,

1919.81 -> the central dependencies to bottlenecks

1921.41 -> within the organization.

1922.78 -> But at the same time, you also need to steel this

1925.72 -> and give a certain direction to your teams.

1928.92 -> So the clear impact that we saw there was that,

1933.78 -> as I mentioned, the central bottleneck went away

1936.28 -> and that the teams themselves were also able now

1938.67 -> in the meantime, to challenge basically the status quo.

1945.15 -> Because before that, it was just clear there are

1947.53 -> certain things that you can do, there were just

1950.5 -> not other technologies there.

1953.27 -> With AWS, we have way more possibilities

1957.03 -> to implement systems and technologies, and by this,

1961.01 -> also to improve the quality

1962.65 -> and the reliability of our products.

1966.61 -> The operational changes, of course, are obvious.

1969.56 -> Before we moved to AWS, we had to create a ticket

1972.26 -> to get a new VM.

1975.61 -> That's gone, so the whole efficiencies

1978.11 -> in terms of what you do on a daily basis,

1980.5 -> just increased massively.

1981.76 -> And you can tell if you're just following

1984.33 -> our Slack channels, that the conversations

1987.12 -> that are happening in between the teams are less,

1989.147 -> "Hey, dude, I'm still waiting

1990.587 -> "for this change to happen.

1991.547 -> "Can you please do that?"

1992.94 -> But it's more a ping pong where they are

1994.63 -> challenging each other and that's really a huge benefit

1998.26 -> that we see in terms of how the culture changed

2001.42 -> by just going horizontally with multiple accounts in AWS.

2008.48 -> And just to make sure that you don't have to keep track

2013.61 -> like, active track on things,

2015.93 -> AWS also provides you with a certain set of functionalities

2019.34 -> that we also, that we are still refining, of course.

2022.32 -> That's also like, not a target that

2024.21 -> you just put for yourself and then you arrive there.

2027.41 -> It's an ongoing process.

2028.73 -> You're adapting the standards and the requirements

2031.71 -> for yourself on a constant basis.

2035.47 -> So in particular, the service control policies help you

2039.29 -> to really define how the systems will look like.

2042.47 -> The service control policies basically allow you

2045.56 -> to give blueprint of the applications that are open

2049.52 -> for your organization to use in AWS.

2052.8 -> And that's something that we enforced quite early.

2055.99 -> Tagging, just to ensure the visibility, is something that

2058.78 -> we also enforce in the meantime,

2060.27 -> to make sure that we just have the possibility

2062.39 -> to have an idea of what's happening in there

2064.3 -> and where the costs are going.

2068.04 -> And there is a lot of stuff that you can just predefine

2071.21 -> to define the outer shell of everything

2074.71 -> that is used within the cloud.

2079.04 -> Of course, alerting just as a very specific thing,

2083.22 -> everything that you do, we have our,

2085.84 -> we have two security teams and these two security teams

2088.61 -> get notified in case that things are not going as we wanted.

2092.23 -> So manual changes, there are some changes that need

2096.19 -> to happen manually.

2099.16 -> They get alerted in this case.

2101.77 -> Most of the changes, 95% are executed

2105.57 -> by Terragrunt and Terraform where we also have

2107.78 -> this centralized approach where we say for our principal,

2112.14 -> if you want to change something for your individual account,

2114.84 -> introduce it here, have it reviewed and approved

2117.48 -> by someone else, and ship it.

2121.88 -> And this gives you the possibility for the whole monitoring,

2125.64 -> having the eye in the sky for your costs, for your usage.

2131.88 -> Basically, the efficiency of your systems

2133.78 -> also helps you then to first identify

2137.91 -> where your costs are going, but in particular, as well,

2141.33 -> how can you improve your products.

2142.87 -> How efficient can you be?

2144.27 -> And where are the things that

2145.44 -> you can tweak and hence, improve?

2151.82 -> And if I now look into where we are now

2154.35 -> and what we want to do in the future,

2158.33 -> actually, like, one or two sentences to this,

2161.91 -> there are so many things that we still need to do with that.

2164.96 -> We want to dive in because this is just the beginning.

2168.51 -> We just arrived fully in AWS.

2172 -> We know that there are a lot of things

2173.61 -> that we can do to improve, be it Auto Scaling

2178.08 -> because auto scaling is not just this one button

2180.47 -> that you flip and then it magically works in a manner

2183.1 -> as you, as you expect it to do, you need to make

2186.89 -> conscious decisions what the criteria are

2190.9 -> for auto scaling and this takes time.

2193.26 -> And for this, we need to make sure that the teams have

2196.14 -> the time to really do this.

2198.14 -> And we want to mature it in tooling.

2200.01 -> So initially, we started out with an APM solution, X-Ray,

2207.02 -> that would help us to see what is happening

2209.12 -> within the systems, looking into how this solution

2212.44 -> will look like then in the future.

2214.18 -> That's something where we need to really spend time on,

2216.91 -> overall increasing quality, increasing the adoption

2220.23 -> of the functionalities that are out there.

2222.49 -> We clearly said we want to use the cloud

2224.68 -> to really leverage the benefits of the cloud

2227.59 -> and not to build up just some new data center in AWS

2231.68 -> that wouldn't really make sense.

2234.7 -> And with that, I thank you for giving me

2239.13 -> 40 minutes of your time.

2240.37 -> And if you have questions, I'm happy to answer them.

2245.18 -> Thank you.

2248.298 -> One quick question is, oh.

2251.09 -> One quick question is in the release cycle, right?

2254.41 -> So there is always a concern about the risk.

2257.39 -> So what do you do for like, analyzing the risk

2260.61 -> and the blast radius of the change?

2267.079 -> So your question is what we do?

2273.46 -> Ah, okay.

2277.85 -> So the question is what do we do

2280.55 -> to reduce the blast radius of a single release?

2286.67 -> If you're releasing more often,

2288.45 -> you're introducing smaller changes.

2291.41 -> That's like, this one truth that led at some point

2294.77 -> in time to the agile work methodologies.

2299.46 -> That's what we do, we try to keep the releases very small

2303.09 -> and at the same time, it's a service-oriented architecture

2306.77 -> in general, which means we have different dependencies

2310.38 -> between the services, of course,

2312.44 -> but if one of these services goes down,

2314.66 -> it is usually only limited.

2320.411 -> The incident is only limited for this specific product

2322.89 -> or this specific service, or the services

2324.87 -> that are directly depending on it.

2326.99 -> So with the set up, the overall set up,

2330.18 -> where we have multiple accounts for individual teams,

2334.13 -> we are reducing the blast radius also for, in case of like,

2337.51 -> a security incident or something like that,

2340.38 -> to really only the domain of a respective team

2344.32 -> or of a respective product.

2350.27 -> What was the tool for the cost analysis?

2354.15 -> That was Grafana.

2356.93 -> Anything else?

2359.38 -> Yeah.

2361.52 -> Is everything a single instance

2363.29 -> or multi-tenant set up?

2364.55 -> Or do you have like, per-client separate production running?

2369.54 -> We have a platform which means we have

2372.49 -> one, big infrastructure that is used

2375.41 -> for multiple of our partners.

2377.07 -> We don't deploy individual technologies for our partners.

2382.37 -> Okay?

2384.98 -> All right, thank you.

Source: https://www.youtube.com/watch?v=65JUKWXPQXQ