Purpose-Built Databases for Modern Applications - Level 300 (United States)

Hello and welcome to the AWS Summit online. My name is play license and on the head of the database business in the asia-pacific and Japan region Fallout public-sector customers. Today we're going to be talking about Peppa still databases for building modern applications. What we going to go through is looking at. What is? I'm on an application. So give you an idea of what we do. Find that as, and then wouldn't go through and look at why customers should be considering how to spell databases and then looking at all the different iws databases that can help you do these

modern applications. At the end of this presentation, that should give you a very good idea of what a new application. What database or database it. As you can use to build for those applications. So what is a magnification? First of all, let's talk about a traditional application. If you think what we've been using for the last twenty years or so it's basically a 2T or 3T application, will you have a client connecting either to a website or application server and then that connects to the database or you still using client-server? It's simply a client

connecting to a database. Now for modern applications, what we're seeing is that does a lot of different requirements. So you going to see that if you're building something that's hopefully going to be very successful and it's either a mobile application online. You could have millions of users, you also could have terabytes or petabytes of data associated with that application. I could be application logs that is generating, lots and lots of data with Olivia uses or it could be a combination of these thirties logs as well as things like video or audio and picture files that are taking off.

Lots and lots of Dale. Then, if you look at like, how we're going to look at the point globally, let's say you have a customer today, whose business is built in the USA or someone else who has built a business in Australia and I want to expand into Europe. They're going to have to look at the architecture of that application to make sure that it can form and to the next Point. Have enough performance and latency to make sure the customer experience as good as they expand across the globe. And then you looking at the number request. So if you've got a very successful application and

that's going to have millions of users using it, then you're going to have a very consistent high, right? All the requests coming in, but let's say you go to smaller application, but it could be very thirsty. Let's say something like learning management system. West units are going to come into that system especially at enrollment time as well as exam time. And of course, the normal Daily Business of coming in the morning. Looking at their classes and then at the end of the day logging off, Sorry that you're going to different types of profiles. Way, you could have Dusty or sustained

high levels of import from uses causing millions of requests and then looking at the different access. Now, traditionally, it's probably just a computer or browser. That's been accessing your applications now, call us when you're looking at building a multiplication, but you also go to take into account things, like all your teeth, and then, even devices that we haven't even thought of. Yet that a connecting into your applications, and then I'm coming back to that Alan ASL, learning management system. You don't really know when it comes

to enrollment. I'm how many students are going to come in and enroll in your particular course, sorry. But you need to be able to scale that out and especially if it's going to be able to be a provision. Those resources in multiple places across the wild without having to worry about resource constraints. And then, we want to be able to pay as you go. Sorry. Say that again. You got all the students that come in for the enrollment and then everything quiets down, we don't want to keep paying for all of those resources. So, we want to be able to turn off the resources and then only pay

for what you actually need. Now, if we're going to go get the permit development perspective, we going to be able to have apis and be able to separate the application components and the storage especially when it comes to the database. So we won't be able to have compute and storage in different areas. Now looking at Amazon and Prime day, this is a good example of a massive event in turn we, where we have to scale and have lots of lots of uses coming in, over a single event. Now, across all Amazon's 442 fulfillment centers. We have a huge

amount of data that needs to be processed and that's across 48 Hours, going around the world for all the different cities and countries. All primed I made 7.11 trillion, holes to the dynamodb API over that. Peeking at 45.4 million requests per second Amazon Aurora also supports Network off Amazon Fulfillment centers, with open 1900 instances of Aurora postgres with 148 billion transactions and they stoled 690b of data and transferred over 360 of data for Amazon

Prime day. So that's an example of a multiplication and the scale that you could actually achieve and have to manage. So, we want to build things with purpose. And if we go back and look at the different types of use cases that you have, you probably don't need to have one database that is going to do everything for you. It is probably better to use something that is specific for the task at hand. So here you can see different types of vehicles that are purpose-built for that job and that's what we're asking you to consider when you're building your applications. Just like with prime day at

amazon.com where we started my DP and Aurora postgres for different use cases to help serve our customers. So instead of building a monolithic application, perhaps you should start considering building microservices. So if you look at the way that most customers have built their applications then they going to have all of their code in this one monolithic executable now some applications and they have gone down with our route and that's kind of like the first step. The problems with Soil Service on side of Architecture is

that you typically need in the Enterprise service bus to orchestrate all of the components and it's complicated microservices you break those components into specifics of his Services, which even have a purpose-built function for that individual service. And behind that service, you going to choose the right technology for the code that you're going to write. And it could be particular language besides don't natural Java or could be a combination of those together with the databases that you need to support that specific service. Team has the full control over that technology that

they do that with. Now, the thing that is different with saw and Microsoft's is that with Microsoft says, you're going to be calling everything with apis over HTTP, so it's much more Loosely, coupled and doesn't need something like an Enterprise service bus to hang it all together. So why would you consider purpose-built databases for this month? Application design with microservices Will traditionally you've had a relational database from the 1970s when called first to find that model all the way up until. Now the great majority of databases that are out there, a relational

databases around the 2000s. When the internet started becoming popular, we have the.com, boom, people start to realize including Amazon that there was problems with scanning is relational databases to fit all of the needs for this month applications and that's where we develop the Dynamo technology and release the paper in 2007 and that became dynamodb in 2012. So what you're saying, there is that we just created a purpose-built database to solve a particular problem. There from that. We also had other databases such as Cassandra, react accounts, DB mongodb,

and oil different flavors of nosql or new type of database. That came out. So now there's a lot more choice for you. When you go on field, one database architecture has to choose from all these different databases. And there are many many out there and fight, there are hundreds. So there is a good choice for you to go. And look at for these more Network actions. So, if you're looking at why you want to feel these purpose-built databases and why do you want to use them again, it comes down to this Kyle performance and availability for the specific task. So, if you think of that girl,

relational database that you threw everything out before, you know, it, maybe it's supposed to dress on now. That's great. But if you use a specific date of ice, that was able to handle some documents and manipulate them. Well, I'm sure you're going to get Benes. Kyle performance and availability for that particular solution, rather than trying to get the relational database to do everything. So Capital One is another customer who has adopted this purpose-built database technology. So they use Amazon IDs or relational database service for that transactional State Management applications.

And then when I want to look at analytics for their web logs, then they putting that data into Amazon redshift where they have to do a groupings, all those web logs and then do reporting on them. And then looking at consistent low, latency Amazon, dynamodb for the use of data and the mobile app is going to be what they use. They're so the game, that's another customer just like Amazon with the Prime die. Capital One has got their own specific requirements and a few different different databases for their specific use cases. So what we really want you to do, is to look at the

use case and choose the right technology. So to help you with that, we have to find a number of different categories. So here you can say that there is good. All relational as the first one then we book key value, document in memory graph, x, series Ledger and why call him. And at a time us we have a series of different databases that match each of these types of technology. And what we're going to do now is we going to go through each of these different types of Technologies to give you a quick overview to help you understand what each one is all about. Sorry,

Amazon Aurora. It is a my sequel or postgres compatible. Relational database built for the cloud. So this is something that we protected for the cloud using our own AWS services and we build very specific technology for the storage to which is storing your data based on the database logs. So it's not actually using a file system for the rights. It is rotting, the database log changes into the storage service but then the reeds are reading off a filesystem. So this is a very different way of using this technology. So great, if the architecture, the way that we stole the data is

across three different designs with two copies in each of those facilities. So six copies of your data for very high durability and resilience from a performance and scalability perspective, you are able to actually get up to five times more report for Aurora, my Sequel and up to three times. With Aurora postgres, this is not running a single querimit. You can run in your laptop and it's super fast and then you put it on, Aurora, anyone, go to say, hang on, it's not as fast as my laptop with talking about running, hundreds of thousands of different transactions

throughout Aurora. Technology, to give you that to report that you simply would not be able to do on a normal instance, on AWS without a specific technology. So if we looking at security, of course, at AWS security is a day Zero priority for us. So you going to have Security in transit and security at rest and we will backups will be encrypted as well. And then of course, it is, part of the obvious manage family where you're going to get all the things like provisioning patching, upgrades and backups and so on will manage for you as well. So, what I want to do now is to

give you an actual demo of how a roar works. And then after that, we'll come back and start talkin about some of the other database services at iws. Welcome to the team. Here is the audience console and on here you can see that we have two databases running already. I want to go to do is quickly create another database and show you how easy it is to get started. So there was the standard create option as well as the easy create option, where can I use the EZ create election? And then look at my Amazon Aurora database for my sequel 5.6 compatibility and will create a dip test instance

of reasonable size server for that you need to enter the details for the cluster identify. So he's going to put a custom name in there and then you could change the miles to you tonight if you want and doesn't order generator auction for a password. But if you want to put it in, you is I'm click that and then you can put the details in there, I'm going to stick with the automatic one, click on create database and then that's going to take us to the screen where we can see all of the databases that I currently running. So this is the cost of its being created and then of the some

information of the talk to tell you what Happening as that's getting created. Now you'll see there's another series of databases and cost is here. And this is part of a global database. So this is running Aurora postgres and this particular day by SIA is going to write that in a Rita and that's in the Arctic region and you can see across different Azz is in Oregon now but here's this Alec Foster has got a read a note in it that is in the Singapore region. So what we've done is created by global database connected between Oregon and see

if we click on that, that will open the Singapore console so you can see IP Southeast one here. So that's for Singapore. And if I got it back and show you then this one is the organ console here. So, let's have a look at this. Singapore Foster, we've got the existing details above and then we want to check the monitoring. Now, what we really want to look at here is, what is the time? It's taken for the date of the proper guy from Oregon to Singapore. So this is the one you want to look at DB replication lag in milliseconds or click on that. So you can see Clara. And you

can see it's roughly about a hundred twenty, five seconds latency on average to bring that died from Oregon to Singapore. So that's giving you the capability to read the data in Singapore, for your application uses there, as well as give you a disaster recovery option. In case something happens in Oregon, so close that down and then close the Singapore tab as well and go back to the organ console. So you can see now that the device has been created so you need to get out of the password details from there for this particular class. And you can say the right side is now

being created, so the cost is done but then they write a note is getting created. So, hopefully that's giving you an overview of how to quickly create an Amazon Aurora. My sequel database, as well, as how we actually build out Global databases and the capabilities that we have with Aurora postgres for that matter. Thank you. Well, hopefully, that Gemma has given you a good of you all to Amazon Aurora, and it probably is giving you a good introduction in terms of how you access outdated by Services through the AWS console. Now, I'm going to move on to other database services and the

first one I want to talk about is Amazon buying my TV. Now remember, I reference Amazon dynamodb with Amazon and Prime day, as well as the Capital One. So they are customers out there are absolutely depending on this for business-critical applications and at AWS and Amazon it is really critical for all the infrastructure that we run. Many of our services depend on Dynamite eBay to make sure that they are running themselves. So you can trust that this is a super reliable database and no, I mean, that it does Skylander perform very well. So you can start with a very minimal Shreveport,

requirement and then of you application become successful, you can increase that through Port. All we have orders scanning capability where you can actually have done in my TV. Play provision expense report. Depending on the request that are coming through is always going to be single digit millisecond. And I've seen this with some demonstrations that we've done what we've actually double the capacity and then double to the game and then reduce it afterwards. And the response time is consistent throughout that process that is very very hard to do with a traditional relational

database. So you'll see this as something that is very compelling to start off with if you're building a new model application. Now there's a service architecture so you don't have to worry about managing service or ec2 instances and it also allows you to provision through put on each individual table, they security is that is in place now, is much improved. So, remember dynamodb has been around since 2012. So if you looked at this a couple of years ago, as a number of different features that have been introduced including improved security as Well, as capabilities around things like time to

live and the last one there, on the slide Global date. I access. Now what we have is dynamodb global tables that allow you to have a table and replicated across to another eight up, your estrogen. And you can write in both of those and have data going and streaming in both ways. So if you're building Global applications don't be is going to be a great way for you to build these capabilities out. The next one is Amazon documentdb. Now this is a mongodb compatible engine and it is built with the Apache 2.0 open source mongodb,

3.6 apis. So essentially what we're doing is we are emulating those apis and giving your application a response that it would expect from a mongodb application. That means that your applications. Today you can simply take them from pointing at mongodb and point them to Amazon documentary be and I will continue to work as they were a previous late. So it's a great flexibility to bring this into a managed service on AWS as well as that. If you've used mongodb and tried to scale it up and then perhaps even scale it back down. You know, that process is quite trying. Now, with

Amazon documentdb, we can actually spend not breathe replicas up to 15, Barry simply and then when you don't need them, you can actually turn them off again. So you going to be able to get Much better service for you to scallop and scale down with Amazon documentdb. Then if you are running mongodb yourself. And I think this is something that's a really big pain point, for many mongodb, customers that I've talked to. So I'd highly recommend you start looking at Amazon documentdb. The next one we're going to talk about is Amazon elasticache. So this is two different engines men

pasty and redis cement iced tea is pretty much a straight for cash. Do you still strings? You pulled the diner in you get it out. It does allow you to actually Sky all across a number of different notes but fundamentally it is a pure cash. Where is redis is more of an in-memory DB and what we're seeing is that most customers are adopting read us these days and some of the capabilities that it has included attama corporations, which a really important for things like count is to make sure that you are getting the actual instruments that you want. And then pubsub, which is quite complex and

well-implemented to allow people to use things like Hughes and then there's ordered sets now. It sets a great for things like leaderboards. So if you're going to game and you want to see you at Webb school is then that can be already ordered in memory from the register device. I just pulled out and slide in your application for could be something similar. Articles. Again, that's an ordered set. That you're going to pull back and bring that into your application. So this is really good. We do see cashing being used as a kind of its own data store. Now, especially for things like session

management and even put more complex, these cases. But one of the other use cases, we see is protection of a relational database. And that means that you can put in memory cache in front of your database, get a lot of the Reeds from that and protect your relational database behind that so that you don't have to scale that relational database as much. Now, this is really good because if you think of things like Annie, we've got product catalog that mostly read only. It's going to be so much pasta to your costume is Antonio protecting your relational database but you're improving the user

experience as well. The next one is Amazon Neptune. Now this is a graph database. The easiest way to probably think about graphs is if you think of Facebook and what they refer to as the social graph, you can think of you and then your friends and their friends. And there's quite a few different relationships between all the way into dependencies of that little own the post say, you're posting and interactions on those posts and comments and relaxed. And so on that is a quite complex graph. But there's other grass out there for looking at analysis of things, like purchasing patents, and

then giving recommendations or recommendation engine, also looking at things like fraud. So perhaps, if you're looking at a bank situation for credit card applications and you wanted to text people who submitting similar addresses or similar to post in 1/10 of a viable information across different applications. Then graph database is a very good for these types of use cases, as well as things like, knowledge grass. So what we see here that Amazon Neptune is able to query, Billions of these different relationships and give you that very far as compared to a relational system. And again,

this is a fully managed service using open-source and Open Standards. So, we've got tinkerpop and w3c IDs as the graph and models if you like and then the languages that you going to be acquiring with Gremlin and Sparkle, The next one is Amazon Prime. Strain time is really difficult to manage in a relational database. You can do it, but you're not going to get the kind of performance and the operations that you can do if you have a purpose-built database for time. So I think of some of the use cases here manufacturing plant or a oil

refinery way. You have things like that, temperature, sensors, and pressure sensors, the constant that you are analyzing against is time. And in the variations are the temperature readings and the pressure readings, and you might have requirements for hot days are to analyze is the temperature sensor going out of range and Reporting on that very quickly. But you also I might want to look at things over a longer period of time and compare things and look for anomalies. Now normally is in these types of environments where pressure and temperature is getting out of whack, might be a sign

that things like maintenance needs to be done to prevent fight is going on. So a lot of this is much easier to do when the whole design of that database is around time. And that's what I was on time. Spring is all about, it's available in preview today and hopefully later this year, it will be generally available for you to use in production. The next a device is Amazon, Quantum legit. Advise otherwise known as qldb bank statement or you got your credits and debits and your balance and you hope that that is immutable. If you could prove

it, that would be great because then you could have no augment with the bank about the transactions that are coming in and I would do it. No, definitely. That that's his not being changed her, that is what is happening with Kotb. You're able to do that, you were able to have an immutable Crystal graphically verifiable journal and Ledger. Now, what does allows you to do is some pretty interesting, use cases relating to financial ones idea for a similar like a bank statement or it could be a government grant system where you were giving out grants and you want to be able to track

and record all of those details. But it could be able to say something like, Becca registration details and managing the ownership of that vehicle over time, making sure that nobody can change those records. And then another example, which BMW is doing is they're recording all The maintenance records of their vehicles. So the second hand or not can see all of the records and prove that that has been the maintenance done on that particular vehicle. And the last one we going to talk about today is Amazon. Managed, Apache Cassandra service. So if you are in love with

open-source Technologies, and you love your Cassandra database, and you don't want to use some of the other database as at iws, then by all means, use this service to implement, your Cassandra database has. Now this is a supporting, the Cassandra query language. So again, similar to what we were talking about, with Amazon documentdb, the way that you can use this service is just point your existing applications at it and it we have implemented the apis and we'll give you the responses that you expect. So this is going to allow you in a socialist model to scale, your Cassandra databases, and

Mike managing that in aws-3 easy. So I'm looking at our approach, remember, we talked about that application Technologies and architectures a changing that customers are moving from monolithic Technologies into Microsoft offices and when they're in that microservices environment for each of those microservices, they get the freedom to choose the different Technologies for the languages. They want to ride him as well as the databases or data stores that they want to use and help you in that environment. Make those choices for the right database than a w. S has

a series of managed services that we've been through with her. It's for relational model which is still a great way to start if you're familiar with that or with no sequel models, when things like document models, as well as graph and Ledger. As we just saw now, all of these services, and it's really important to understand the benefit of this, you don't have to worry about having to provision physical Hardware. You don't have to worry about installing the software patching a software doing upgrades, and then things, Bad cops are taken care of for for you as well

as recovery and invent all the highway. If I Leo storage by someone so we want you to concentrate on doing what you do best, which is building the applications, that light your customers. And let us take care of managing the infrastructure and the databases behind the scenes. If you do need to get that, we do have two technologies to help you with that. The first one is the database migration service and this is all about moving data from one database to another. So you can bring a postgres database into Aurora postgres with the

database, migration service with minimal downtime. But you could also bring an Oracle database and then point that at a another relational databases such as my school. For example, you could bring a nosql database in such as mongodb and then bring that into a document eBay. So the database migration service is going to allow you to move between different types of databases with minimal. Define from, I'm from ac2 and into a minute Services. Then there is a schema conversion tool. Now, if you need to move things like let's say you were migrating sequel server but you'd like to

move that too. But you need to transfer a schema definitions at able definitions that data types as well as t-sql and convert that to a p, o p g c equal than all that happens within the scheme, a conversion to fully automatic. You're still going to do something on your work but it is a great way to get started and make that process much easier. So if you do want to get started and then we have a couple of links head there. First one is a WSDOT amazon.com databases and that's going to give you a whole heap of information to show a documentation. I get started guides

videos and also two things to help you get up and going Now, if you do want training and certification, then there is self-paced learning that you can do online. There is in class options as well. For specific areas that you want to learn about a WS and there is now a certification. Course for databases. So where is before you might have done your solution, architect certification. Now you can do your database certification as well. So that brings me to the end of my session. Thank you very much for attending the iws summer online and please fill in the feedback form. Thank you

very much.