>> Live from San Francisco, it's theCUBE covering Google Cloud Next 2018, brought to you by Google Cloud and its ecosystem partners. (techy music) >> Hey, welcome back, everyone, this is theCUBE Live in San Francisco coverage of Google Cloud Next '18, I'm John Furrier with Jeff Frick. Day three of three days of coverage, kind of getting day three going here. Our next guest, Sudhir, as the director of product management, Google Cloud, has the luxury and great job of managing BigTable, BigQuery, I'm sorry, BigQuery, I guess BigTable, BigQuery. (laughs) Welcome back to the table, good to see you. >> Thank you. >> So, you guys had a great demo yesterday, I want to get your thoughts on that, I want to explore some of the machine learning things that you guys announced, but first I want to get perspective of the show for you guys. What's going on with you guys at the show here, what are some of the big announcements, what's happening? >> A lot of different announcements across the board, so I'm responsible for data analytics on the Google Cloud. One of our key products is Google BigQuery. Large scale, cloud scale data warehouse, a lot of customers using it for bringing all their enterprise data into the data warehouse, analyzing it at scale, you can do petabyte scale queries in seconds, so that's the kind of scale we provide. So, a lot of momentum on that, we announced a lot of things, a lot of enhancements within that. For example, one of the things we announced was we have a new experience, new UI of BigQuery, now you can literally do the query, as I was saying, of petabyte scale or something, any queries that you want, and with one click you can go into Data Studio, which is our DI tool that's available, or you can go in Sheets and then from there quickly go ahead and fire up a connector, connect to BigQuery, get the data in Sheets and do analysis. >> So, ease of use is a focus. >> Ease of use is a major focus for us. As we are growing we want to make sure everybody in the organization can get access to their data, analyze it. That was one, one of the things, which is pretty unique to BigQuery, which is there is a real time collection of information, so you can... There are customers that are actually collecting real time data from click-stream, for example, on their websites or other places, and moving it directly into BigQuery and analyzing it. Example, in-game analytics, if in-game you're actually playing games and you're going to collect those events and do real time analysis, you're going to literally put it into BigQuery at scale and do that. So, a lot of customers using BigQuery at different levels. We also announced Clustering that allows you to reduce the cost, improve efficiency, and make queries almost two X faster for us. So, a lot of announcements other than the machine learning. >> Well, the one thing I saw in the demo I thought was, I mean, it was machine learning, so that's hot topic here, obviously. >> Yes. >> Is you don't have to move the data, and this is something that we've been covering, go back to the Hadoop, back when we first started doing theCUBE, you know, data pipeline, all the complexities involved in moving the data, and at the scale and size of the data all this wrangling was going on just to get some machine learning in. >> Yep. >> So, talk about that new feature where you guys are doing it inside BigQuery. I think that's important, take a minute to explain that. >> Yeah, so when we were talking to our customers one of the biggest challenges they were facing with machine learning in general, or a couple of them were, one, every time you want to do machine learning you are to take data from your core data warehouse, like in BigQuery you have petabytes of scaled data sets, terabytes of data sets. Now, if you want to do machine learning on any portion of it you take it out of BigQuery, move it into some machine learning engine, ML engine, auto-ML, anything, then you realize, "Oh, I missed some of the data that I needed." I go back then again take the data, move it, and you have to go back and forth too much time. There are analysis I think that different organizations have done. 80% of the time the data scientists say they're spending on the moving of data-- >> Right. >> Wrangling data and all of that, so that is one big problem. The second big challenge we were hearing was skillset gap, there are just not that many PhD data scientists in the industry, how do we solve that problem? So, what we said is first problem, how do we solve it, why do people have to move data to the machine learning engines? Why can't I take the machine learning capability, move it inside where the data is, so bring the machine learning closer to data rather than data closer to machine learning. So, that's what BigQuery ML is, it's an ability to run regression-like models inside the data warehouse itself in BigQuery so that you can do that. The second we said the interface can't be complex. Our audiences already know SQL, they're already analyzing data, these folks, business analysts that are using BigQuery are the experts on the data. So, what we said is use your standard SQL, write two lines of code, create model, type of the model you want to run, give us the data, we will just run the machine learning model on the backend and you can do predictions pretty easily. So, that's what we are doing with that. >> That's awesome. >> So, Sudhir, I love to hear that you were driven by that, by your customers, because one of the things we talk about all the time is democratization. >> Yeah. >> If you want innovation you've got to democratize access to the data, and then you got to democratize access to the tools to actually do stuff with the data-- >> Yes. >> That goes way beyond just the hardcore data scientist in the organization-- >> Yeah, exactly. >> And that's really what you're trying to enable the customers to be able to do. >> Absolutely, if you look at it, if you just go on LinkedIn and search for data analyst versus data scientist there is 100 X more analysts in the industry, and our thing was how do we empower these analysts that understand the data, that are familiar with SQL, to go ahead and do data science. Now, we realize they're not going to be expert machine learning folks who understand all the intricacies of how the gradient descent works, all that, that's not their skillset, so our thing was reduce the complexity, make it very simple for them to use. The framework, like just use SQL and we take care of the internal hyper-tuning, the complexity of it, model selection. We try to do that internally within the technology, and they just get a simple interface for that. So, it's really empowering the SQL analyst with an organization to do machine learning with very little to no knowledge of machine learning. >> Right. >> Talk about the history of BigQuery, where did it come from? I mean, Google has this DNA of they do it internally for themselves-- >> Yes. >> Which is a tough customer-- >> Yes. >> In Cloud Spatter we had the product manager on for Cloud Spatter. Dip Dee, she was, like amazing, like okay, baked internally, did that have the same-- >> Yes. >> BigQuery, take a minute to talk about that, because you're now making it consumable for enterprise customers. >> Yeah. >> It's not a just, "Here's BigQuery." >> No. >> Talk about the origination, how it started, why, and how you guys use it internally. >> So, BigQuery internally is called Dremel. There's a paper on Dremel available. I think in 2012 or something we published it. Dremel has been used internally for analytics across Google. So, if you think about Spanner being used for transaction management in the company across all areas, BigQuery, or Dremel internally, is what we use for all large scale data analytics within Google. So, the whole company runs on, analyzes data with it, so our things was how do we take this capability that we are driving, and imagine like, when you have seven products that are more than a billion active users, the amount of data that gets generated, the insights we are giving in Maps and all the different places, a lot of those things are first analyzed in Dremel internally and we're making it available. So, our thing was how do we take that capability that's there internally and make it available to all enterprises. >> Right. >> As Sundhir was saying yesterday, our goal is empower all our customers to go ahead and do more. >> Right. >> And so, this is a way of taking the piece of technology that's powered Google for a while and also make it available to enterprises. >> It's tested, hardened and tested. >> Yeah, absolutely. >> It's not like it's vaporware. >> Yeah, it's not. (laughs) >> No, I mean, this is what I think is important about the show this year. If you look at it, you guys have done a really good job of taking the big guns of Google, the big stuff, and not try to just say, "We're Google and you can be like Google." You've taken it and you've kind of made it consumable. >> Yes. >> This has been a big focus, explain the mindset behind the product management. >> Absolutely, there is actually one of the key things Google is good at doing is taking what's there internally used, but also the research part of it. Actually, Corinna Cortes, who is head of our AI side who does a lot of research in SQL-based machine learning, so again, the-- >> Yeah. >> BigQuery ML is nothing new, like we internally have a research team that has been developing it for a few years. We have been using it internally for running all these models and all, and so what we were able to do it bring product management from our side, like hey, this is really a problem we are facing, moving data, skillset gap, and then we were like, research team was already enabling it and then we had an engineering team which is pretty strong. We were like, okay, let's bring all three triads together and go ahead and make sure we provide a real value to our customers with all of that we're doing, so that's how it came to light. >> So, I just want to get your take, early days like when there was the early Google search appliance, I'll just pick that up, and that was ancient, ancient ago, but one of the digs was, right, it didn't work as well in the enterprise, per se, because you just didn't have the same amount of data when you applied that type of technique to a Google flow of data and a Google flow of queries. So, how's that evolved over time, because you guys, like you said, seven applications with a billion-- >> Yep. >> Users, most enterprises don't have that, so how do they get the same type of performance if they don't have the same kind of throughput to build the models and to get that data, how's that kind of evolved? >> So, this is why I think thinking about, when we think about scale we think about scaling up and scaling down, right? We have customers who are using BigQuery with a few terabytes of data. Not every customer has petabytes scale, but what we're also noticing is these same customers, when they see value in data they collect more. I will give you a real example, Zulily, one of our customers, I used to be there before, so when they started doing real time data collection for doing real time analytics they were collecting like 50 million events a day. Within 18 months they started collecting five billion a day, 100 x improvement, and the reason is they started seeing value. They could take this real time data, analyze it, make some real time experiences possible on their website and all, with all of that they were able to go out and get real valuer for their customers, drive growth, so when customers see that kind of value they collect more data. So, what I would say is yes, a lot of customers start small, but they all have an aspiration to have lots of data, leverage that to create operational efficiency as well as growth, and so as they start doing that I think they will need infrastructure that can scale down and up all the way, and I think that's what we're focusing on, providing that. >> You guys look at the possibility, and I've seen some examples where customers are just, like, they're shell-shocked, and you're almost too good, right? I mean, it's like, "We've been doing "Dremel on a large scale, I bought this "data warehouse like 10 years ago," like what are you talking about? (laughs) I mean, there's a reality of we've been buying IT, enterprises have been buying IT and in comes Google, the gunslinger saying, "Hey, man, you can do all this stuff." There's a little bit of shell-shock factor for some IT people. Some engineering organizations get it right away. How are you guys dealing with this as you make it consumable? >> Yeah. >> There's probably a lot of education. As a product manager do you see, is that something that you think about, is that something you guys talk about? >> Yes, we do, so I think I actually see a difference in how customers, what customers need, enterprise customers versus cloud native companies. As you said, cloud native companies starting new, starting fresh, so it's a very different set of requirement. Enterprise customers, thinking about scale, thinking about security and how do you do that. So, BigQuery is a highly secure data warehouse. The other thing BigQuery has is it's a completely serverless platform, so we take care of the security. We encrypt all the data at rest and when it's moving. The key thing is when we share what is possible and how easy it is to manage and how fast people can start analyzing, you can bring the data. Like you can actually get started with BigQuery in minutes, like you just bring your data in and start analyzing it. You don't have to worry about how many machines do I need, how do I provision it, how many servers do I need. >> Yeah. >> So, enterprises, when they look at-- >> Cloud native ready. >> Yeah. >> All right, so take a minute to explain BigTable versus, I mean, BigTable versus BigQuery. >> Yes. >> What's the difference between the two, one's a data warehouse and the other one is a system for managing data? What's the difference between Big-- >> So, it's a no-SQL system, so I will... The simple example, I will give you a real example how customers use it, right. BigQuery is great for large scale analytics, people who want to take, like, petabyte scale data or terabyte scale data and analyze historical patterns, all of that, and do complex analysis. You want to do machine learning model creation, you can do that. What BigTable is great at is once you have pre-aggregated data you want to go ahead and really fast serving. If you have a website, I don't expect you to run a website and back it with BigQuery, it's not built for that. Whereas BigTable is exactly for that scenario, so for example, you have millions of people coming on the website, they want to see some key metrics that have been pre-created ready to go, you go to BigTable and that can actually do high performance, high throughput. Last statement on that, like almost 10,000-- >> Yeah. >> Requests per second per node and you can just create as many as you want, so you can really create high scale-- >> Auto-scaling, all kinds of stuff there. >> Exactly. >> And that's good for unstructured data as well-- >> Exactly. >> And managing it. >> Absolutely. >> Okay, so structured data, SQL, basically large scale-- >> Yes. >> BigTable for real time-- >> Yes. >> New kinds of datas, different data types. >> Absolutely, yes. >> What else do you have in the bag of goodies in there that you're working on? >> The one big thing that we also announced with this week was a GIS capability within BigQuery. GIS is geographical information, like everything today is location-based, latitude, longitude. Our customers were telling us really difficult to analyze it, right, like I want to know... Example would be we are here, I want to know how many food restaurants are in a two-mile radius of here, which ones are those, how many, should we create the next one here or not. Those kind of analyses are really difficult, so we partnered with Earth Engine, Earth Engine team within Google with Maps, and then what we're launching is ability to do geospatial analysis within BigQuery. Additionally along with that we also have a visualization tool that we launched this week, so folks who haven't seen that should go check that out. One great example I will give you is Geotab, their CEO is here, Neil. He was showing a demo in one of the sessions and he was talking about how he was able to transform his business. I'll give you an example, Geotab is basically into vehicle tracking, so they have these sensors that track different things with vehicles, and then with, and they store everything in BigQuery, collect all of that and all, and his thing was with BigQuery ML and a GIS capability, what he's now able to do is create models that can predict what intersections in a city when it's snowing are going to be dangerous, and for smart cities he can now recommend to cities where and how to invest in these kind of scenarios. Completely transforming his business because his business is not smart cities, his business was vehicle tracking and all, he's like, but with these capabilities they're transforming what they were doing and solving-- >> New discoveries. >> New discoveries, solving new problems, it's amazing. I wonder if you could just dig at a little bit to, you know, the fact that you've got this, these seven billion activities or apps that you can leverage, you know, specific functionality or goals or objectives or priorities in those groups, and now apply those, pull that data, pull that knowledge, pull those use cases into a completely different application on the enterprise. I mean, is that an active process-- >> I don't think that's how people. >> Do people query? >> No, no. >> But how does that happen? >> No, we don't-- >> As a customer. >> As a customer completely different, right? Our focus in Google Cloud is primarily enabling enterprises to collect their data, process their data, innovate on their data. We don't bring in, like, the Google side of it at all, like that's their completely different area that way, so we basically, enterprises, all their data stays within their environment. They basically, we don't touch it, we don't get to access it at all, and they can know it. >> Yeah, yeah, no, I didn't mean that, I meant, you know, like say Maps for instance, it's interesting to see how Maps has evolved over all these years. Every time you open it, oh, and it's directions-- >> Yep. >> Oh, now it's better directions, oh, now it's got gas stations, oh, now it's where the... And it triggered because you said the restaurants that are close by, so it's kind of adding value to the core app on that side, and as you just said, now geolocation can be used on the enterprise side-- >> Yeah, yes. >> And lots of different things, so that-- >> Exactly. >> That's where I meant that kind of connection-- >> Exactly right, so-- >> In terms of the value of what can I do with geolocation. >> Absolutely, exactly, so like, that's exactly what we did. With Earth Engine we had a lot of learnings on geospatial analysis and our thing was how do you make it easy for our enterprise customers to do that. We've partnered with them closely and we said, "Okay, here are the core pieces of things "we can add in BigQuery that will allow you "to do better geospatial analysis, visualize it." One of the big challenges is lat longs, I don't think they're that friendly with analysts, like oh, numbers and all that. So, we actually will turn a UI visualization tool that allows you to just fire a query and see visually on a map where things are, all the points look like and all. >> Awesome. >> So, just simplifying what analysts can do with all these. >> Sudhir, thanks for coming on, really appreciate it and congratulations on your success. Got a lot of great, big products there, hardened internally, now-- >> Yes. >> Making consumable, it's clear here at Google Cloud you guys are recognized that making it consumable-- >> Yep. >> Pre-existing, proven technologies, so I want to give you guys props for that, congratulations. >> Thank you, thanks a lot. >> Thanks for coming on the show. >> Thanks for coming on. >> Thank you. >> It's theCUBE coverage here, Google Cloud coverage, Google Next 2018. I'm John Furrier with Jeff Frick, stay with us, we've got all day with more coverage for day three. Stay with us after this short break. (techy music)