Transcript for:
Disruptive Changes in Banking Technology

foreign [Music] we weren't supposed to tell you about it but we tell you the truth that we came here to hand the best mind to our Dream teams yeah and drink to good coffee I guess a lot of you know this guy uh almost 20 years ago he just made this sentence banking is necessary but banks are not so let's check whether this quote is still available and related to our present times my name is Eric my name is piot and now we jump to agenda we tell you about a short story of vodino what was the future of banking and what changes are happening now on our eyes and also what technology challenges we have in the bias API area in infrastructure and also in the course perspective okay so what do you know the short story the company was founded in 24 years ago in 2018 we got the founding from one of the biggest private equity in the world Warburg Pincus and after one year of development we just create the fully working operational banking platform and it was ready in June 2019 just to start start serving the banking services for friends and family face so this was just the beginning after five months of very rapid transformation we migrated one of the oldest the brand the Belgian Belgian Branch one of the oldest bank in the world just to the cloud and this is the funny story but the journey to the cloud of this bank took them almost 500 years uh in 2020 we became a Google partner we became a Google partner in three areas in technology reseller and services if you'd like to you can just scan the QR code and read the the whole story in 2021 uh when the pandemics run through the world and we were closed and shut down in our houses we successfully helped ion to open new new markets one in Poland one in Germany and in 2000 2002 we started working we started working on a bus so banking as a service area and we are able to serve the Technical Solutions for our clients so this is this is it and okay so now I I can present you what disruptive change in banking area is happening now in the past if you want to open account take a loan you must go to the bank the bank the branch of this bank to do this current if you are want to open account tag along you open your favorite shopping app and make payment take a loan for your grocery shop and what's means for technology company in the past if you deliver such solution to banking area you you implement project mouth years and now if you want to add a new payment methods change something in the learn process you can do this in the hours or days and what is important what they know provides all solutions to take this journey from the bank business and end user customer I guess you remember those quote from Bill Gates uh for his this was a future for us it's just a current state so let's see how it works in the in the market you can see four examples of uh not banking applications but they are embed the banking services in the Europe in the europes so starting from the left unit plus is a company that are helping you to invest your money in ETF portfolios and using those ETFs wallets you can also buy a donut or just go to the coffee coffee shop then money flow this is the application or business that helps other businesses with funding like funding for income or funding for uh for for Lending three come three count is the application that allows you to split your bills across your friends when you just join the lovely dinner at the restaurant and you are just using psd2 interface to to split split the bills across each other and the last one Metro financial services this is the application I guess maybe you know metro metro is one of the biggest wholesale reseller in the world and they are offering their services for their for their SME clients that means that they are offering uh the couple debit card just helping correct a business I mean hotels restaurants and Catering for paying for shopping in in their shops and what is important although for examples are just part of portfolio of our clients technological clients of vodino so what technology technological challenges we had during those four years of a journey the first one the bus apis as you can see as Spirit mentioned as probably you also know we started with a bank so we call it Ion Bank era and we didn't have any external clients external partners that means that we didn't have to expose our apis to external external work this was quite uh quite easy for us just to do the development internally right now because we had to change we need to provide a well-documented publicly available bus API so this was the quite challenging thing and we are we are offering right now for us what piotr also said in integration with our bus apis it's quite it's quite important to be as easy as changing subscription and Netflix and not in the not that hard if you will go to the traditional bank and open the account in branch okay now we jump to infrastructure you can see on the slide how big our infrastructure is current we are running on our all environment 20 000 ports nine thousand deployments 60 plus uh Cloud SQL instance and VMS so our infrastructure is not so small maybe not so big what is also important current we are our deployment are in three gcp regions Belgium meter line and Frankfurt Frankfort is only needed for kafka's keeper uh and what is important that we offer services current in four countries Poland Germany Sweden and Belgium next slide please and build a bank in the cloudless challenge but build the bank will be compliant with regulator is more challenging currently platform must be compliant with European banking Authority and we prepare for you three uh three bullets three topics where we address these things first of all one of the requirements is that data center must be geographically separated in case of Google regions it means that we must be ready to outage of one gcp region it means that we implemented our solution we must think about this how to replicate data between gcp regions on application layer on database layers storage layer and we did it and what is important this is not only on the paper because also the same regulator require that at least one a year we should prove it then we delivered it and we do this second thing is that we are working in regulator sector and also regulator needs to know what if we must took our all infrastructure from Google Cloud to another Pub another public out or on-premise so we must prepare some kind of mapping Services between different cloud provider and on-premise and and doing this we also thinking what will be if if it will be necessary and in our cases we decided to build our monitoring alerting logging system based on elastic stack to have this opportunity to easily switch between gcp another cloud provider on on-premise and the last one is networking uh in gcp is quite too easy achieve this even easier than on-premise because we use shared VPC and in one host project we manage all networking topics firewalls etc etc and all services related with this network and we have separated environment production and non-production and this is really separated because in traditional World nobody wants to pay uh the cost of two expensive firewall for production non-production and in cloud in gcp is quite too easy to achieve this and also our third main services for application layer gke we use private gkey it means that nodes have private IP addresses and also Cloud SQL we use in private mode it means that nodes and instances are in the private Network so now let's go to costs I know that talking about costs might be quite boring especially For Engineers but at the end of the day we are still we are all of us are working in the businesses and we need to think about costs and income and we need to compare those two values so all the I guess all the cloud providers will tell you that pay-as-you-go model is probably the best option for you and probably it might be but there are few things that you need to think of the first one do not rely only on a cloud cost calculators on the right hand side there is a Google Cloud calculator and you can see the cost for compute engine for gke you can see there is one position and one standard two and it costs around 54 for 54 euro per month and probably this is the final cost that you will be thinking of and this is not fully true because on the bottom you will see all those simple all the small positions that you will find on the billing and sometimes for for bigger environment governments those simple positions like storage like Network might cost you might cost you a lot and usually the the people that are starting working with clouds just forget about those positions the second thing do not rely only on a cloud billing why because in our case we implement the the rule trust but check what does it mean we had a situation with Google that we had a cloud SQL instances and in console we saw the one type of of disks under these instances but we were charged for something different so this was the mismatch of course after the discussion with Google support this was this was removed and replaced but still if we wouldn't check it we'll just paying for different types and we we will even get the the worst services that we expected to have so uh in the last example I guess very very uh important at the beginning of our journey one of our data scientists wanted to use Google Cloud Google Maps apis just to decode addresses from uh from strings to geolocalization positions and he use places API just to decode it and in documentation he found that using one special endpoint it's potentially free so here execute this API for about hundreds of thousands addresses and after a month we get a bill for about 20K Euro on deaf environment environment so this is the this is the case that if we wouldn't check it if we wouldn't just verify and then get to the Google and says guys what's up we were charged for something that is potentially free yeah this was a mistake there was a mistake in documentation and so on and so on so you need to look at every single position very carefully do not rely on on your Cloud providers because they also they might also do some mistakes so the question is how to improve it how to implement how to implement the proper cost model just to avoid such situation that we had we suggest you to based on our example just to implement two cost models the first one on the left hand side we call it proactive or bottom to top that means that for every single application build cost mode cost calculator so put every single component from your application and then try to think how your application will be scaling what will be cost driver how you will be growing and based on this you can try to predict or estimate the cost for now for future for for next for next years so this is the first thing of course this might be quite challenging for smaller application or easier or smaller environments you will just put your simple simple position and that's it for more complex applications with lots of dependencies this might be quite challenging quite challenging situation the second thing reactive so when you will get the first bill the First Billing get the invoice get every single position and try to allocate those costs across your business units or even to applications and then I'm the the core banking care banking owner so my my responsibility to keep this product as low as possible so I need to know what is the cost calculator and what are the final costs and I try to compare those two calculators if the discrepancy between those two two values are below five percent from our opinion then we can say that calculator is working properly and we can estimate those costs and we will not be surprised that after one year and when the business will be growing will pay you know 100 1000 K or even even more so this is the the thing of course if you've got the bigger discrepancy you need to review whether your cloud provider is lying probably not or your Cloud calculator is not working correctly so this is the these are the cost and the examples and periods if you can explain how we deal with cost cuttings okay so we present with the Eric few examples how we can cut the cost and have some benefits implement this first thing uh you you have production non-production Dev environment in this way is to save the money turn machine type on non-production environment from normal one to preemptive or spot and what is and you save around 70 percent the money and second thing is if you are wondering how you can destabilize non-production environment and make some uh house banking tests is good idea because in Bri if you choose preemptive in Google it means that you don't know when Google turn off the VMS you don't know it and and this is important to and discounts banking test is very important in Cloud to be ready that environment is unstable and if you do this and was you will have some problem with application you should ask is this application is ready for cloud or no next one second idea is as in in your house you turn on light if you don't use it make the same in Cloud it's quite too easy to do this and you can save the money and if you are green and want to decrease the global carbon footprint you can make it and also there is other thing you need to review your components architecture I guess every single year at least every single year at the beginning for for core banking we decided to use cloud spanner database from Google it's highly scalable globally distributed highly available globally distributed horizontally scalable database and it's pretty simple to use but of course it costs especially if you don't have so big scale so after uh after a year or two we decided to move from cloud spanner to Cloud SQL we also redesigned our application architecture so this was the natural way or natural path but still we review what are the potential changes in terms of Google portfolio whether we can decrease some cost based on doing some maybe potential small small changes so in our case we we estimate that the cost reduction will be for around 30 percent from spanner to to Cloud SQL but it depends on the components it depends on your on your situation okay and the last thing about Cloud cost if you really don't have time to make some changes or Implement something you can save the money do nothing only you must sign the commitment with your cloud provider as you see for compute engine from Google if you sign the contract from one year you can save 37 percent if you sign for three years you can save 55 percent and the same is in Cloud SQL 25 and 52. but signed the contract from the year is quite good and saved because when you sign the commitment you must commit that you use resources in specific gcp region it means that if your business Journey was changed will be changing geographically you must think twice if you want to assign the commitment for three years because in case if you sign the commitment today for three years and two years from now you decided to take your resources to another region you must pay for previous region and the new one and it will be very hard to achieve some good TCO for your solution so we were talking about most of negative aspects or maybe some challenging things so let's talk about one positive thing in terms of cloud we had a discussion about running some performance tests in in cloud in our environment and we know that doing some such things in traditional way might be not possible so we were supposed to run our environment for about 7 million and customers so it's quite quite huge uh quite quite huge platform So In traditional way in traditional data center on on-prem just pinning up such huge resources it's even sometimes it's not even possible because of many things first of all you don't have enough resources available in your data center or even if someone will would like to pay for this and you will order this you will need to wait for delivery installation and so on and so on and Cloud there are no such things of course you need to discuss with your provider whether in your region or Zone there they've got enough resources we had such a situation once in the past but for not so big even for the 7 million and customers it's not so big volume your cloud provider should allows you and you will just spin those resources and just run the test so it's pretty easy and once again remember about costs because there might be some small factors that you will forget run environment for 7 million customers and you will be you know what okay so thank you very much for listening you can reach us on LinkedIn and Twitter and if you'd like to if you would like to chat with us directly we will invite you to our booth you can get this stickers you can get the stickers to get some nice gifts so thank you very much grab your coffee and see you at the booth foreign