Transcript
and for any of you who were here last year, you'd have seen me and Woon do the same thing, slightly older version of this last year too, right? So therefore, I am the opening act for Woon as it should be. He's the one who, he's the intellectual powerhouse who comes on stage after I do, right? Largely, my job is to sort of establish how AI no longer is an option. We keep hearing about, you know, how AI is everywhere and what's happening in AI. A couple of weeks ago was a very important week for AI India and AI in India. You guys must have heard about the AI summit that happened in Delhi, right? It sort of, for the rest of the world, it sort of was an eye-opener of what's happening. But first, and if you guys have seen me before, you can scroll your phones for the next 60 seconds because you'd have seen this, but when an Amazonian gets up on stage, they'll tell you a little something about themselves, which you won't find in their, you know, social media profiles, etc. So here goes mine. I'm neither an engineer nor a business graduate by education. My last institutional education was as a bartender. I majored in beverages. I was one of India's first certified sommeliers, right? So I get this question a lot. From a bartender to the head of data and AI for Amazon, which Diwali lottery did you buy? What happened, right? So much so that it almost became like an existential question, right? So, and if you have ever had the misfortune to hear me speak before, you will have heard me banging about this quite a bit, about data being the genesis of all innovation. And while I was doing research on this, here is what I came across, right? Any idea what this is? Don't worry, it's not a fair question, but this is apparently the first known record of data in human history. This is from approximately 1400 BC, what is now Southern Iraq, right? And it's obviously a clay tablet and like every other priceless thing, if you want to see today, you'll have to go to the British Museum in London, right? But what is more important is what is this record? So apparently the first known record of data in human history was a record of how much beer was given out to slave labourers. And that's when I figured alcohol and data has been fueling innovation for centuries. And serendipitously, I ended up at the right place at the right time. But innovation, innovation is a very bandied about word, you know, you sort of stick it onto anything nowadays. But if I asked you to close your eyes and visualise innovation, more often than not, this is what you would think of, right? This shiny Eureka moment where we had an idea that solved global hunger, that solved international peace. But actually, if you were to sort of zoom in, right, and this is what innovation looks like. This is these hundreds and thousands of data points that you have collected, whether it is through education, through experience, through nature, through nurture. And that allows you at some point of time, to connect the dots and come to that Eureka moment. So collecting the dots is as important as connecting the dots. And we are able to do it because the human brain is just this amazing machine learning instrument. And AI and specifically Gen AI is at the forefront of what we are trying to do in making your organisations, our organisations act like one single human brain, a unified human brain for everything, right? And what is the single most important thing to get generative AI and AI value from? Very obviously, you don't need me to tell you this. Very obviously, it's data. And today, if you look at what Woon will talk about, if you look at what I will talk about in AWS, right, S3 is at the centre of everything we do at data, right? You know, it's not, it's not just started off as a simple storage format. But now, if you look at it, it's at the centre of whether you're trying to do analytics, whether you're trying to do machine learning, whether you're trying to do AI and generative AI, it's right at the centre of stuff. And especially more so now, when you have things like the S3 tables, nobody talks about that better than Woon, so I'm not going to even attempt. Or as you know, machines don't understand pictures, they don't understand words, they only understand numbers which come out of your vector. And therefore, natively, it's a vectorised database. But why am I here? Right? I'm here to tell you, why is it important that you sort of look at your entire data foundation? And why is this partnership with Commvault so important? Right? And for that, I want to do something a little different. I said at the very beginning, AI everywhere for every Indian, right? And it's sometimes it's easy to get lost in these fascinating models, these AI tools, you know, all of these agentic AI utilities that we have. But I think the coming of age of any technology is when it touches our individual lives. That's when it really has become a technology that is going to change the world. And so here's one day I sat down, I said, I'm an average Indian. Do any of what me or my teams do? Does it really matter to me as an individual? And I sat down and I thought about a day in my life. And here's what I found. It was so fascinating that I thought I would share a day in my life with you. Right? So here goes, come join me. Right from the time that I am woken up by Alexa. And yes, I sort of every day wake up like this, looking like that. Right? To the razor that I should use more often. When I get the first shot of energy from my breakfast cereals, where I get my news, whether that news is in English, or it's in the vernacular, the gyms that I go to should definitely go to more often, or even how I track my health and my wellness. When I do any of that, there is an AWS data and AI customer who actually touches my life. touches my life. And then when it's time to go to office, whether I'm zooming through traffic in my shared mobility two wheeler, or maybe I'm actually driving my own car, right? Or even the tires that all my vehicles drive on. When I take a ride hail cab to the airport, find the quickest route there, on the way, upload my boarding pass to Digiatra. And who here doesn't love Digiatra? I mean, it's something that all of us Indians should be absolutely proud of. It allows you to zoom through these modern fantastic new airports in India, use all their absolutely brilliant facilities. And then finally get on to one of these new, new India's new airlines, which is creating a new story in the sky. And when I do all of this, there is an AWS data and AI customer who actually touches my life. Then I get to office. And I wish most days I was as enthusiastic as this on getting to office. But when I get there, what I noticed is all the applications that I use, whether it is my CRM systems, whether it is my mark tech systems, whether it is my accounting systems, and so, so, so many more, right? And even when, for example, I take this small little chai cigarette break, when I order for lunch, what I eat, where I order it from, how I pay for it, maybe get a little bank work done, check on what's happening to my investment portfolio, no great news there, nothing much to announce to you guys. But when I do all of it, there is an AWS data and AI customer who touches my life. And then after a long day at work, if you're like me, I like to put my feet up, have a little drink, have a little chakana with my drink, maybe Netflix and chill. And why just Netflix and chill? Every streaming service there and chill. Maybe book my next holiday. Maybe help the kids with their schoolwork, buy tickets to a movie date night. And then when it's my turn, when it's my turn to cook, it'll either be the two-minute meal delivered in less than 10 minutes. But honestly, who are we kidding? I'll probably open the app, go to my favorite restaurant and order in. And every time towards the end of the month, there'll be this one moment of panic when I realize it's the last day to pay my insurance premium. But hey, click, click, click, and it's done. When you're my age, you have that last pill that has been delivered in less than 30 minutes. Read a little, listen to some soothing soundscape and drift off in that wonderfully smooth mattress. So actually, right from dawn, and hopefully we'll get there at some point of time. Yes. Right from dawn till the time I'm done in the evening. Every moment of my life, there is actually an AWS customer who touches my life. Every moment that happens. That happens. And if you stop for a moment, think about a day in your life. I'm sure many of this will resonate with you. You'll find many of these small moments in your own personal life. Because AI is no longer just something we talk about. It's no longer in our labs. It's no longer in slides. AI is now in our boardrooms. It's in our factories. It's in our hospitals. It's in our classrooms. It's in your pocket. It's in our phones. It's in our bikes. It's in our cars. It's in our home appliances. It's in our kitchen. It's all over our houses and our lives. And therefore, doing AI is no longer an option. It's here. It's happening. It's happening in this country where it's being built in India for India and for the rest of the world. And therefore, it is now more important than ever that we figure out how do we safeguard our AI workloads, our data foundations, so that we can actually build better, more Vixit Bharat. It is my pleasure as always to call the main event of the show, my good old friend. I've never heard anybody speak more eloquently about our services than Woon. Thank you so much, Woon. Fighting as always. Thank you, Biswa. I can see Biswa, definitely AI is everywhere. I mean, this is the third time that I'm doing this with you. Like your slide, oh man, a huge upgrade. I don't know if I can match that. You know what? I'm not going to do this after you anymore because I don't think I can match that slide. Like I really feel bad about my slides. So anyhow, good afternoon, everybody. I hope you guys all enjoyed the lunch. Stay awake. It's actually about 2 a.m. for me. So we'll get through this. So first of all, I'm going to start with the problem statement right out of the gate. So if you think about it, data is really the cornerstone of all applications, especially generative AI applications. Having said that though, the large majority of the business leaders that we talk to, they tell us that they're not quite ready when it comes to data resiliency for their applications. But like innovations cannot stop. Today, if you think about it, enterprises are implementing and developing AI applications faster than ever. And when I think about implementing and deploying AI application, it is hard for me to come up with a better place than developing other than AWS. AWS provides you all the way from the infrastructure where you have silicons that are specialized to do training and inference. Not only that, they have data stores where you can actually store the data that your AI is going to be trained on, or it will actually process the data. On top of that, you have things like services like Amazon Bedrock, where you can safely and securely deploy any model of your choice. And then, like I said before, data is the core, it's the center. And at the center of that, there's S3. And it's actually pretty easy to actually see and think about why is that the case. S3, it is, I would say, the most efficient, scalable, and durable storage out there. And then it has so many integrations that once you actually bring the data into S3, you can actually extract a lot of value out of it, whether it is through data lakes, analytics, or, of course, machine learning and AI. Now, let me repeat it a little bit, because this is very important. And you see me repeat this a few times. Data is really at the core. And then for you to get a successful AI outcome, it really depends on having the right data. Data that is clean, data that is reliable, and data that is accessible. But honestly, what happens? Accidents happen all the time. This is not about if, it will happen. You have accidental deletions, you have operational errors, and even malicious attackers. Things do happen. And then let's look at it from a little bit of a different angle. On the left-hand side, let's say we have a few S3 buckets where you have your data lakes, which then gets consumed, piped into things like big data analytics that is used for customer segmentations, which fits into your AI applications. Then later, it gets fed into some of the AI agents that consume all that data. And let's just pretend that there's a corruption in that data. So what happened? That corruption gets propagated all the way down. So it's not about the data, but now customer segmentation, demand forecasting, customer services, everything gets impacted. So it's really, if you think about it, it's garbage in and garbage out. So really, what you really need to do is to have the right data resiliency strategy for that data. Now we've talked about data resiliency. Let's talk about some of the key points when you consider about data resiliency. First thing is about protecting the data, right? So it's not about hoping that things are not going to happen. That's not a strategy. Things will happen, and then you would have to go ahead and protect it. And then later after that, I would say it's more important than the protection. You would have to do the right way to actually recover from it. A lot of the times, once you have it backed up, people don't think about the recovery aspect of it. But then recovery, I'll argue, is way more important than backups. So essentially what it really means is that you should be able to test them and actually test them well. It's not about having this little, you know, small test, but really think about it very carefully and think about, you know, I have this deployment and come up with realistic scenarios. How long is it going to take you to recover all that data? So Poojan talked about customers like Atlassian. They have 30 petabytes of data. Imagine recovering that, right? And what does that mean in terms of RPO and RTO, right? Then really, then if you actually tie all of this together, then we're talking about scale. Like you are in the public cloud because really what you are trying to leverage is the elasticity of the cloud. Like you want to go from a few terabytes to a few petabytes, right? And then if your backup or data protection solution doesn't scale along with that application, that ends up being the bottleneck. And then at the end of the day, you end up paying the price in terms of RTO or RPO. Now let me talk a little bit about Clumio. So far I've been talking about a lot of problem statement and stuff like that. So Clumio founded 2017. Now it's about, you know, nine years, close to 10 years now. We started on AWS from the very beginning with the goal of creating a cloud native data protection solutions built on AWS to support the native data services in AWS. Things like, you know, S3, of course, DynamoDB, RDS, Aurora, EBS EC2, like many other services. That was the goal number one. We started on the public cloud to go solve, to protect the data services in the public cloud in a native manner. Now with that goal, basically allow us to take a lot of simplifying assumptions and actually go solve problems that matter for that one problem. Then allow us to implement very deep, deep, deep cost optimizations for us to be able to deliver, you know, TCOs that, again, Poojan talked about in the previous slide, in the previous session. Along with that is scale. In the cloud, you can really, really scale. Like you can have buckets with 60 billions of objects, 80 billions of objects, or you can have a DynamoDB table with about 900 billions of items in it. So for example, Duolingo will actually store a lot of their students' data in that DynamoDB table. And if that were to go away, that's not a good place to be. The scalability is key. And then the other thing that we did is, hey, well, we're starting from scratch. So what we need to do is come up with a deployment model that it should be zero footprint. We didn't want it to come to you and tell you, hey, deploy this, you know, compute on your environment. No. What we wanted to do is, from day one, it had to be a service. We had to be invisible. We should be able to protect all of your data, scale on demand, ask your application scale, but you didn't have to do anything. Like we didn't want it to, for you to actually deploy things and do things for us, but then the service will handle it behind the scene. All, again, it should be all transparent and zero footprint. And that's kind of some of the models that we had when we actually started Clumio at the very beginning. And then it was, you know, it was a long journey. It's not something that happens overnight. And then all that time, we actually worked pretty closely with AWS. Like I said, it is now almost eight going into the ninth and tenth year that we've been working very, very closely with AWS to come up with some of these solutions, both from product, engineering, go-to-market. We've been working pretty closely with AWS. And then one example is Amazon S3 table is something that Lizbeth talked about. So something we actually talked to the S3 product team about a couple of years ago. And then we talked about what they were working on. And then they told us that they were trying to implement what's called an S3 table. Think of this being as a special type of S3 bucket that is dedicated for Apache Iceberg. So we thought it was a great idea because from our customers pool, we would actually see Apache Iceberg being one of the data formats that is actually growing pretty fast on the S3 buckets. And then the fact that the S3 team saw that and they were trying to come up with a service that provides that, it provides basically Apache Iceberg more as a service, it made total sense. So we actually worked very, very closely with them. And then we ended up actually releasing a feature. So this was actually announced last re-invent, so put it last November. So it's basically a couple of months old now, or maybe a little older than that. But where today, basically Clumio is the only solution out there that will actually provide you Apache Iceberg backup, that is Iceberg aware. And that part is actually very important because I kept saying, hey, well, isn't Apache Iceberg a bunch of S3 objects that sits in their bucket? Why is it so important to be Iceberg aware? Why can't I just take that object and treat it as an S3 backup? In other words, why is S3 backup any different than Apache Iceberg backup? It's actually quite different. Because if you actually do the copy that way, like let's say you try to backup, so if you try to do things like replication, you're copying data without the format, like that copy process doesn't know that object contains Apache Iceberg metadata or data. So what it means is that once you actually do all the backup, you can bring all that data back. But if you actually try to use it, you might have to spend a long time to stitch back all that metadata. And then today, because Clumio can actually do this all knowing that it is an Apache Iceberg format, then we actually go ahead and fix all the metadata for you so that when it's ready to go, you don't have to spend a ton of time rebuilding the metadata. Again, coming back to RTO, because if you have an accident, you want to get going really, really fast. So that's kind of key to do things in a manner that is Apache Iceberg aware. So if you look at it, and just to give you an example, people used to, the analogy that I use is that you can actually back up your, let's say your database is at the disk level, but you don't do that because you can restore the disk. But then once you bring up the database, your database may not be functional because you took the backup at the storage layer, right? So it's very similar here. So then Clumio basically is the industry first solution that provides you an ability to create an ergo copy of your Apache Iceberg. And then some of the other additional benefits that comes with it is that in the case of Apache Iceberg, you could actually have different snapshots that you take. Essentially, every update to the table ends up being a snapshot. But at the same time, a lot of the query, because you're not, typical query, you're querying the last snapshot. You're practically, it is very rare that you end up querying a snapshot that is, let's say a year old, right? You're querying always typically the last snapshot. But then if you actually have a ton of snapshots locally on your primary, guess what? Your primary, the query suffers, the query performance. So then as a side effect, what we allow you to do is that we allow you to take some of the oldest, we'll take the backup of the whole, the newest snapshot as well. But then once you actually take the backup with the Clumio, we allow you to go back and maybe trim some of the snapshots on your primary. So then, because now you have all of the snapshots in Clumio, like the snapshots that created yesterday, but then the snapshots that you created two years ago. So if you have to retain them for compliant purposes, you can have them in Clumio while you don't have to take the penalty of slowing down your query on the primary, for example. And then another way that we actually work pretty closely with AWS is we are partnering with AWS. This is a program that is about a few weeks old. It's pretty, you know, pretty, pretty new. So like I said before, Apache Iceberg format is a format that existed for some time already. It's, you know, it's 10 years ago. But the S3 table is a new feature that got introduced about a year, close to a year now. So you can imagine there's more, there's a lot of Apache Iceberg in the standard S3 buckets. There's a lot of them because, you know, this has been building out for a long time. Now there's this new service that is provided by the S3 team. You want to probably, you want to probably take that and migrate it over into the S3 table, right? Because it gives you all the values. It gives you all that. Basically, they are doing the undifferentiated heavy lifting. So you can actually use Apache Iceberg the right way without doing a lot of the in front work yourself. But then you want to take that and migrate it. And we actually work pretty closely with the product team. And we realized that we can actually do, you know, we can actually help. Clumio could actually go ahead and help you. The way that it works is that it's really a glorified backup and restore because we support both types. We can actually take the Apache Iceberg, back it up, you know, the table that sits on the standard buckets, we will back it up. And then when it is time to restore it, we can restore it back as an S3 table. So then effectively what you achieved is a migration. Buckets that you had on the standard table, on the standard bucket, got migrated to S3 table. And all this, all while being Iceberg aware, all this while being very, very simple and automated. And the best part is actually for the program, it is free for the customer, obviously. And then I don't know you, I like free stuff. If you like free stuff too, take that, you know, snap that QR code and that's the program that allows you to do the migration. All right. So I'm going to do, I'm going to sum it all up. A lot of the stuff that I say is going to be repetitive, you know, it is important, right? So number one, AI application or traditional enterprise application, like data is the foundation, right? It is the core. It is the thing that you have to get it right. And you have to have the right strategy. Really, you know, just take away one thing. It doesn't really matter what model you use and what all of these things is garbage in, garbage out. You have to make sure that protect the data. And then once you want to do that, then when you come up with the recovery or data resiliency strategy, just think about, you know, is it scalable? What's the TCO? How fast can I recover? And all that data pipeline, it is now becoming very real, like accidents are real. And when it does happen, it is a real thing. That's something that you need to think about. Then think about scale because you're there, you are in the public cloud, right? So you must have billions of objects and, you know, billions of billions of items in the DynamoDB table. Think about scale. You want a solution that will actually scale with your application. Because if the data protection solution doesn't scale with it, then essentially, you know, either RP or RT or something will give. And you have to think of that. And this is my last slide. And maybe I will get Reswap back so that we can actually close it together. Thank you.
