 Thank you Good afternoon You know it's such a big room and few of us we can just talk one-on-one. I think You know I've given a lot of talks In my career and never in a movie theater. This is a interesting experience here So today's talk is in before I dive in first of all, how many of you have heard of Redis? That's pretty good. And how many of you are using Redis in some way? Okay. All right You know Redis said I'm gonna have a video at the end It'll give you a good perspective on how some very very large global companies are using Redis and in today's talk It's really about how do we go from here to the future? I know this conference is all about artificial intelligence but underpinning that is a large number of technologies that make it happen and and in the world we live in is It's really a world of instant right what we do today is So much more than what we did a year ago two years ago our productivity He tends to go up exponentially every year with new tools and technologies and the expectation on us keeps on going up The bosses we have expect us to do more in the same eight hours of work, right? Anybody doing less today than they did last year? Probably not And and our expectations in our daily lives are all about getting what we want when we want and to get it immediately Right, we don't have the patience if you're watching a movie in Netflix We want that to show up immediately the menu it better be to our personalized Requirements if you're ordering a cab Uber you want it to show up in the next two three four minutes And you want the responses to be immediate if you're doing shopping on Amazon you want that to work and and and anything in life today is Really if it's not instantaneous, we don't have patience for it our attention span has going down every day There are studies that say you know anywhere between eight and maybe down to two seconds is our attention span these days And as the generations the Millennials and the younger generations come in their attention spans are even shorter So how do you achieve that men meet the expectations of the consumer? When it's becoming so difficult everything has to be within within that instant that you're looking for you know We're all used to this you click on something and then you wait for the circle and then something appears, right? No longer. I don't think we have the patience for this anymore if it's slow It might as well not work, right? We want it to work and and and it's not just a matter of experience There's actual real revenue that can be impacted We all know Amazon is the largest e-commerce player in the world and and they have something called a prime day Amazon prime day earlier this year they had a Hiccup in their operation for several hours. It cost them hundreds of millions of dollars So it was not that the people could not connect to Amazon But the service and the experience got slowed down to an extent where the transactions were not completing at the time And they would time out and so from a user standpoint Slow or disconnected is the same thing and so the revenue is at risk. If if that instant experience is not delivered You know in our estimate there are four major macro trends in the marketplace that are driving What we experienced today and what we can expect to experience tomorrow in five years from tomorrow and And and here are in in no particular order. So first is quantum computing You know this has been around and talked about for about 30 years now But finally quantum computing is getting to a point where it is going to become a reality in the next year Maybe two years. So quantum computing is is a way of processing data Based on quantum physics. So the data is stored in what's known as quantum bits or qubits And and there is a threshold of 50 qubits the moment the industry can hit 50 qubits You will have a computer that can process better than any super computer can today with less with fewer resources Intel has announced earlier this year that it research did a chat test ship with 49 qubits So we are almost there We are the cusp of making quantum computing a reality and the moment that happens What we are used to in our daily lives on our cell phones on our computers and even in the kind of super computers I think we'll be taking to a whole new level thus delivering brand new experiences that we have not seen so far The next one is persistent memory You know, we all know that RAM is where the processing happens calculations happen and We know that disk hard disk are where data gets stored, right? So if you have Oracle database it sits on hard disk and you retrieve the data you process and do the calculations in RAM The problem is RAM is fast, but very expensive and hard disks are cheap, but very slow, right? So how do you bridge the gap? What persistent memory is trying to do is give you the performance of RAM at The price point of a hard disk, right? So it's trying to bridge that gap and and we are getting very close We do we work very closely with Intel and Samsung the companies that are leading the initiatives are on persistent memory And and we're getting to a point where within the next six months three months You will see this is becoming reality and and that has is a game changer because you can store Incredible amounts of data and get performance at almost as good as RAM The third thing is 5G so Europe certainly has been a leader in LTE 4G and and and now moving to 5G and And and we on the cell phones are used to getting a megabits per second five megabits per second If you're lucky, maybe up to 10 megabits per second, right? What 5G promises is and the tests have some early tests have shown we can expect it to a gigabits per second type of a Throughput on our cell phones now we may not get a gig, but but if you get seven hundred five hundred four hundred megabits per second That's a game changer already today. We're used to watching videos We're used to doing almost everything in one and we're pretty good and the experience is good But think about if you now had the capacity there was a few hundred times what you'd get today and we're getting close to that And finally what this event is about is artificial intelligence You know earlier I was doing an interview somebody asked me is artificial intelligence is you know Is this is a too much hype and is it's going to sort of die down and are we going through the the trough of? disillusionment and and Gartner's terms and and and I don't think so I think all the way from you know complicated things like facial recognition to autonomous Autonomous driving vehicles to even things like the recent refresh Google has on its on its Gmail is it gives you you know Recommendation it fills out the answers for you It gives you answers that might be appropriate and you can just click on it and you're done Those are all examples of algorithms machine learning algorithms that are supporting AI underpinning underpinning AI and and I think there is no facet of our life that will not be touched by AI So if you take all of these things together and you think about what's possible in the future between communications Processing algorithms you have a world that zero latency is really becoming real You know you almost have to just think and it happens. That's the vision we have for for for the future Bringing it down a notch to the applications now. How many of you are application developers? Okay, and how many of you are sort of in the DevOps that you operate applications Or integrate various applications together, you know regardless of the three environments you're in You are being influenced by certain brand new paradigms that might have been around for a while But but are becoming reality today. The first one is cloud native and microservices You know, this is where instead of having a monolithic application You are now getting an application that's developed with very specific use cases very specific services that together Make it a single application again going back to the uber example You might under uber may be a hundred microservices running a geospatial microservices a credit microservice a financial service microservice You might have on credit card balance Microsoft your profile microservice and so on right all of those together Give you the experience that you get when you're using so that's the microservice The second piece is cloud is clearly a reality, right? Soon we'll see, you know greater than 50% of the workload is going to be in the cloud But but it's not going to be a single cloud. It's not going to be just a public cloud It's likely to be a private cloud public cloud, you know running underpinned by AWS Google Azure and perhaps others and and you might be running in your own enterprise cloud private clouds of your own What will not happen is everything from on-premise is not going to move to the cloud Some things for governance reasons for how things operate in your company will remain on premises And so this hybrid world of cloud and on premises is likely to be status quo and the applications have to deal with that and finally You know the the thing about we want when we want we also want where we want, right? And this is not just about are we moving from you know? I live in California. I'm in Madrid today and and I still want to have the same experience on everything I do regardless. I don't want any degradation of services Similarly when you're thinking about developing an e-commerce application a communications application Your consumers and producers of data on that application are likely to be all over the place So how do you architect your application such that it is giving the experience? That is what you expect regardless of the location and there is no downtime and you get the zero latency future So let's let's go a little bit deeper on to microservices and cloud native you know what what this graphic indicates is Two to two visuals of some of the largest applications right Netflix and Amazon and you can see there are hundreds of microservices that underpin and an Amazon e-commerce and an environment or in Netflix Video delivery environment these microservices are very loosely coupled. You know, they don't rely on each other They don't necessarily are dictated. So the the the reason microservices aren't play is because it makes the management the flexibility of upgrading the ability to add new capabilities So much more flexible than than you ever had in the past You would see software applications be updated every six months Maybe every 12 months in some cases every two years because the reason was if you have an application with a million lines of code and that has to be updated think about their Regression and the QA cycle that has to happen, right? Now with microservices the app is broken down where you can update sub-components of the app and you can see now Amazon and Netflix they update their apps every single day sometimes Multiple times a day and it's not that they're updating everything but they peek picking up these microservices in doing So these loosely coupled lap allow you to to deliver the experience and to give you this better experience every single day You're getting something new and something fresh This is a Simple representation if you have an app that's supported by a number of microservices Each microservices it has an instance of a database that it talks to and if you think about the example I gave earlier where a microservice is a geospatial microservice a second microservice is for your credit card information a third microservice is for your profile information and and you can layer all of those and Each one has a specific need and requirement from a database and it's possible that each microservice will use and apply Distinct database to achieve that objective now take an example of a hundred microservices supporting an application And if each one had a unique database instance, that could be a pretty complicated It could be a nightmare to manage and so the need for databases that have a Multi-model approach becomes very important. So you don't have this variety of database that you see at the bottom here The next piece is multi-cloud and hybrid I talked about it earlier You know when you develop the applications you want to make sure they run on any infrastructure You don't want to get locked in by AWS or GCP or Azure You want to have the ability to run on any and if possible Do it in a way that the workload can be shared across multiple clouds the multi-cloud environment Large enterprises that we sell to and we work with are all looking for multi-cloud setting because they don't want to get Locked into a single cloud and second very important piece is going back to the fact that the workloads will not shift to the cloud in its Entity is you want a hybrid ability to serve cloud and an on-premises environment and in starting at the largest enterprises Where this is a requirement But I think as we go down over time Every company is going to want to have the flexibility if they don't get the service of the pricing from AWS You should be able to move to to Azure or be able to work in New York They work on on on a AWS infrastructure in Madrid Perhaps they work on a GCP infrastructure right and that flexibility is really important if you're thinking about building a global application The next piece is availability everywhere right so a geo-distributed multi-availability zones and and just making sure that when you have this distributed environment It is built in a way that it can resolve the conflict if you have multiple sources Which are writing to the to the infrastructure and certainly being read from different places the chances of creating their conflicts are very high And so you want the database technology the network technology in the entire Service is sweet to be able to deal with that conflict resolution in a real-time basis and in foundational to all of the things we've talked about is Data you know the world of application is almost flipped around It used to be that you would build a large application an ERP or MRP or you know one of those applications And then you would figure out what the source and the sync for the data was and where you would store the data What the retrieval pattern might look like what the eviction policy might be? What data is supposed to be hot what data can be remained cold? You know you would think about that after the app was developed the world is really flipped and the right way to do it Is to think about what is the cadence of change of data? What's that velocity? What's the volume of data you expect today in five years from now from the app will continue to be used and And what's the kind of you know the presence of the data by geography by time? There are many attributes that really are data-driven attributes that should then inform and drive your architecture for the application You know we talk about cloud native data, and what does that mean one is the data will be stored in multiple ways So you want to support the data in a way that it's it's schema less and it can support the polyglot persistent that you saw the Variety of models that the data has to be stored in in the databases in this new microservice environment You want to make sure it's a self-oriented service architecture? So you can self-service and make changes over time you want to make sure that you know it used to be where The changes were done so infrequently that the data processing had to be done at a much slower cadence today You don't need to store every single piece of data You might process the data and immediately evict the data because the processing done informed the analytics Transactions done any move on you don't have to store everything if you decide what type of data you want to store What type of data is simply for processing to inform your immediate interaction analytics and so on and then move on and that Eviction and and ingest policies become very important So all of these things need to have the flexibility and the ability to deal with a multi cloud environment a hybrid environment and in an agile manner So this takes us to the the core of things where data resides and that's databases We all know you know data database 1.0 was oracle and my sequel and postgres and all of that And they continue to be the largest players, but those databases are not suitable for unstructured data They're not suitable for data that changes very often and hence about ten years ago No sequel came into play companies like MongoDB Redis Labs and others came into play and and these databases allow you to deal with higher velocity with all data types And are are more prevalent and ready for fought for today's environment And now as we move to the next level Can some of these databases evolve and morph to meeting the zero latency future We talked about can they give you the response times that you need can the right and read of this at the same instance Can they deal in in multiple use cases? Can they operate in the cloud and can they operate on premise all of these things come into play as? The desire to meet our consumer expectation of instant becomes real So we have a vision that all data must be instantaneous Doesn't matter where it's coming from doesn't matter how frequently is changing Doesn't matter who is producing it and how many people are consuming it the data must be instantaneous to everybody and so this is where where Redis is and Notice, you know about 20 percent 25 percent of you have heard of Redis So let me just share a little bit Redis is a open source in-memory database It's been around for about nine nine ten years this in 2019. It's a 10th year anniversary You know one very simple measure in today's environment is the containerized world so Docker is a standard for containers and On Docker hub just about three weeks ago Redis surpassed a billion launches It is the most containerized database of any database in the world a billion launches that is a you know There is no better data point to to say how popular Redis is in today's modern Application architecture the microservices architecture containerized world Redis is is number one The second piece is the developer community absolutely loves it for its performance for its simplicity and for its extensibility This survey that's run by Stack Overflow for the last two years has voted Redis by about hundred thousand developers worldwide Voted Redis as the number one the most loved database Anybody wants to guess which database would be at the bottom, which is not here the most hated database I Heard Oracle so this is not meant to if you work for Oracle This is not meant to disparage you Oracle is the number one in terms of the number of deployments However, you know the complexity of dealing with an a legacy architecture traditional database It's it's very hard for the developers the performance is not there and the complexity is just as too much So it's on the other end of the spectrum DB engines is a site that tracks 340 some databases around the world and Redis is ranked number seven on DB engines and within the no sequel category It is ranked number two behind MongoDB. So just to give you some data points, you know On the viral interest and the love for Redis among the developer community What we have done is Redis labs is we are the company behind open source Redis We invest quite a bit in with the developer community But to serve the enterprises we have built on top of open source what we call Redis Enterprise It is it is a enterprise great solution that can be built and deployed at scale It can be built and deployed in a highly geo distributed environment It can be delivered. It can deliver high availability with with no data loss. It runs on all clouds So we try to put in things that are very important for enterprises as to think about delivering this zero latency experience to the end users and And going back to the microservices world You know, you think about if you had to deploy a graph database a key value database a document database You know and try to do for each microservice category You have you would be saddled with a large amount of technology that you would need to have their competency in your companies to deal With it becomes very complex to manage becomes very difficult to train the people and to keep that up so what Redis and Redis Enterprise have done is is Offered a number of things that are inherently natively built into it and we've done that in an elegant way using something called Redis modules What you might hear about this multi-model approach by other databases But they simply utilize an API to do calls with other Functionality and that's how they offer the multi-model capability in that effort. They lose time. They lose performance They do the translation so there's inefficiency what Redis and Redis Enterprise have done is is these are all based on the Redis core and these modules can work on the same data that resides in the database and Pull in depending on the use cases that has been utilized at that time and can apply the services For graph for key value for document as as you move forward So it gives you the agility to bring in the use the capability you need for the use cases You're trying to serve and gives you the ability to lever the same database without it becoming a highly complex environment to deal with So it's sort of best of all words flexibility without being saddled with the complexity that comes with other databases Talked about availability on various clouds So it can run in a multi-cloud environment on any single cloud as well as on the right-hand side You can download it as natively as a software and run it on any bare metal that you want You can run it in a containerized environment if if Kubernetes is your orchestration layer Which it like it most likely is you can run natively on Kubernetes Or if you're running pivotal infrastructure with pks or red hat with the open shift So it gives you really the ultimate flexibility on deploying a Redis Enterprise database regardless of what your infrastructure looks like and and I wanted to share a little bit about about the scale and And and and the performance and what happens to most databases is you can have good performance when you've got a small scale As you increase the scale the performance degrades very very quickly in most databases What this chart is indicating is that regardless of the number amount of workload of the ops per second Operations per second you're trying to get you are not going to take a hit on the latency So here it shows all the way up to 50 million operations per second Continuing to be operating at sub millisecond latency and that's absolutely Incredible there is no other database that can do that So if you're trying to do an e-commerce app or a communications app that's deployed in a global manner That's got this volume and an SLA of you know running millions of operations per second And you want to deliver the zero latency experience Redis is the way to do it the other thing on the horizontal axis you will notice is the number of nodes So this is the infrastructure cost that's associated with your database and because the database if Redis is highly highly efficient Very low footprint you're able to get this level of performance with very small So even at 50 million operations per second, which is very high end You only need 26 easy to instances on an on an ad of blue as infrastructure What that translates into is a dramatically lower cost of operation when you're running Redis as opposed to other databases And and what we have done is you know talking about the multi-model environment Regardless of what you're running on Redis whether it's you know, it's your standard no sequel capability you're running search graph streams or Serving machine learning algorithms in conjunction with spark or some other framework You are able to get performance that several magnitudes higher than alternate options All of it because the core model of Redis is the same and these Functionalities can layer in and leverage that performance that comes with lettuce without having to sacrifice or do or do API calls Which makes it slower and and as resource intensive The next piece is we comes into play when you've got active active environments that ensure that you know, I talked about Conflict-free a free environment if you've got a geo distributed application you have deployed your your Source and your sink might be very different what what's known in our terminology master and a slave might be very different And they may change roles over time. How do you ensure that? Service is not disrupted. How do you ensure that you're getting local latency? Regardless of where the source of the data might be or the right is happening. How do you make sure that happens? We have something called active active based on CRDT's that Prevents from the data from the from that takes care of the conflicts in the data and allows you to get local latency in a highly distributed environment Next piece is consistency So in the database world, this is this is a big deal So relational database like Oracle talk about strong consistency no matter where you're writing and reading you're always consistent data When you move to a faster environment where the data is moving at a much faster clip and you're trying to achieve the Conflict-free resolution and you're trying to do the multimodal aspect It is incredibly hard to achieve strong consistency in many use cases. It doesn't really matter So what we have brought to the table is what we call tunable consistency ability to get strong Eventual consistency which means that you may not be immediately consistent But over a period of time you will be consistent as the data changes and there's also an element of Causal consistency, which means the consistency is informed of what happened prior to the to the step You're you might be in at the moment and between those two capabilities It gives plenty of flexibility for the app developer to achieve the consistency It needs in a geo-distributed environment regardless of the scale and the volume of data we're talking about The next piece is you know I talked about one of the big paradigm shift was persistent memory now before that happens what we have done over the last two three years is Redis operates in RAM. That's its native state. However, there is a cost associated with RAM more expensive than disk So if you need the performance and you don't need all of the data to be in RAM We allow we have something called Redis on flash where you can put in hot values and keys and RAM and put the cold data On flash and and you're able to get about 70% of the performance that you would when you put everything in RAM and and the cost savings is dramatic Right, and you don't have to go to a low performance environment like a hard disk You can still back up stuff there and and and have replicas and so on but but for things that require the performance You can balance your workload between RAM and the SSD So that's the second step and finally as I said we're working very closely with Intel and Samsung So when they go go live with the generally available persistent memory Redis will operate on that as well So it gives you tiering of memory at it from an and in a very intelligent manner that doesn't degrade the performance Yet gives you the perform the the cost savings that that comes with some of these newer technologies Now I want to shift gear a little bit to talk about what are people doing with with Redis and And you know, you might have heard is it in a database for transactions? Is there a database for analytics and and with Redis being multi-model being that it can be very very high performance It has the ability to operate in an analytics environment It has the ability to run in a transaction environment and certainly in an operations environment as you can see from a large number of very Simple use cases to very complex use cases One of the things that's very unique with Redis. That's at the bottom It says translitical or h-tap those are terms from from the analyst firms But but what that is is the same workload or the same data set can be applied in utilize for Running your analytics algorithms. It can also be applied to execute your transactions But for the two things to be simultaneously done is where the complexity comes in and most database simply can't do it So what we are able to do with Redis is you're able to Inform the transaction based on real-time analytics. So what does that mean? That means that let's say you've got x amount of data You can go to the run queries on the data do the analytics and based on the results of the analytics You can then inform whether the transaction should be executed or not a very simple example might be a credit card transaction Now when you swipe your credit card There is 10 to 11 steps that actually happened on a visa mx or mastic or whatever you might be using and And and those are trying to do a number of things at the same time and you want that to be instantaneous Right, you know my patients to wait there So what Redis is able to do is look up all of your profile your transaction history your geo Where are you located? What was the last few transactions you did your credit balance on the credit card all of that and if the answers are Appropriate then the transaction can be executed and Redis can help you run the session of the transaction as well So trying to do that all in a in a real time where you can get this zero latency experience is as you can imagine Incredibly complicated Redis is able to do that and and and you know three of the top four credit card processes are our customers Doing exactly that so that's an example of a use case and and there are many many more Very large customers that that utilize Redis in in different industries We have over 8,500 customers with Redis Enterprise that have deployed this at large scale Serving tens of thousands hundreds of thousands in some cases over a million transactions or operations per second We've got customers that are running several terabytes of data running on Redis enterprises some examples So instead of going through each one of them I'm going to play for you a three-minute video. It'll give you a good sense of what some of these big brands are doing with Redis So hopefully you get a good sense of the breadth of applications where Redis fits in and has been utilized with large enterprises We have a booth here if you have any other questions You can experience Redis by going to this URL Redis labs comm get started. There's a free service You can experience, you know within a five minute set up an instance of Redis and Experience a performance and zero latency work that Redis is able to do. I think we have a few minutes for Q&A Okay, maybe one or two questions And on the theater it fall asleep anybody awake Well, feel free to stop by our booth. It's just around the corner. We have a whole team technologists Folks that will love to answer your questions. Thank you very much for your time