Google Cloud Next '17 – Day 2 Keynote

>> URS HOLZLE: Good morning, everyone Good morning and welcome to the second day of Next So, as you know, we’ve been building and operating a hyperscale Cloud for a very long time, and our goal throughout this time actually has remained the same, namely to make it possible for you to build great applications and to be very productive as a developer or as IT while we deliver this on a global scale with unprecedented reliability, performance, and security And this Cloud powers all of Google’s services, including the seven billion user services, so YouTube, Gmail, Search, and now it also powers a GCP, which itself is a billion user application, meaning that every single day the users of GCP, the customers of GCP, connect to over a billion individual users every single day And what makes this possible is that we started from first principles We designed every element of our infrastructure so that you could be uniquely productive and that you can enjoy the performance that we created Now when designing this from first principles, you have to go and actually optimize every single element, so from efficient data centers to custom servers to custom networking gear to a software-defined global background to specialized ASICs for machine learning We’ve been living this Cloud at scale And, in fact, in just the last three years, we spent almost $30 billion on capital expenditures alone Now one example where this all shows is our network It’s probably the largest global network today Analysts put its traffic at between 25 and 40% of global Internet user traffic And as a GCP or G Suite customer, you benefit from this network because your traffic travels on our private ultra-high-speed backbone for minimum latency And 98% of the time we hand off your packets directly to the end-user ISP because we interconnect directly with almost any ISP everywhere in the world So you won’t see any congestion And fewer handoffs mean more throughput, lower latency, and better security So, in fact, we have a global network presence in 182 countries or territories in the world For comparison, the United Nations today has 193 Member States And so to carry this traffic to pretty much everywhere in the world, we also need to cross oceans And so nine years ago Google became the first non-telecom company to build an undersea cable That was the Unity cable from the U.S to Japan And since then, we’ve built or acquired fiber capacity, submarine fiber capacity, pretty much anywhere in the world so we have a redundant backbone to pretty much any place So, for example, last year we turned up what today is the highest capacity submarine cable in the world from the U.S to Asia And in just a few months, we’ll turn up an even higher capacity cable between the U.S and South America Now one of the ways we use this network, of course, is to connect GCP regions to each other Last year we announced that we’re building 11 new Cloud regions Of these 11, Oregon and Tokyo are already live, Singapore and Norther Virginia will go live in just the next few months, and Sydney and London will come on shortly after that And today I’m excited to announce that we are adding three more regions to the set, namely the Netherlands, Canada, and California And all of these will come online either this year or next and bring the total count of GCP regions globally to 17 and the total number of zones to 50 Now our technology leadership extends from hardware to software If you think about Containers, MapReduce, NoSQL, TensorFlow, Serverless, Kubernetes, they all originated at Google and they’ve been powering our services, including GCP, for a long time

And recently we added yet another of these ground-breaking new systems to our lineup so you can use it on GCP, and I’m talking about Cloud Spanner Spanner solves a long-standing problem in databases Until now, you could either use a database, a SQL database with nice transactional semantics, but it was really hard to scale it beyond one or just a few machines and very complicated and very expensive Or you could use a no SQL system like Bigtable and you get infinite scale, horizontal scalability, but you no longer have strong consistency guarantees, and so you’re pushing the complexity into your application Neither of these two choices gave you a truly global system because your replicas needed to be pretty close to each other in order to get good performance And Cloud Spanner, which is a globally distributed database service, solves these problems So it scales horizontally like a NoSQL system and thus it can take pretty much any transaction loads, but it also has these strong consistency guarantees and the relational SQL semantics of traditional databases And as an extra bonus, it can also be globally distributed while managing high performance We use it for hundreds of mission critical applications inside of Google, including AdWords, and now it’s available on GCP as well So how does Spanner do that? I don’t really have time to go into that but I’ll show you one example So Spanner uses a timestamp service, a highly accurate timestamp service, to keep the copies of your data in sync This timestamp service is powered by atomic clocks because it needs to be really very highly accurate So that means we literally run atomic clocks in every one of our data centers to make Spanner possible But with GCP, since Spanner is a service, you see none of that complexity In fact, Cloud Spanner is fully managed and serverless so that you don’t have to worry about machine types, capacity planning, replication, software updates, really anything And that’s what we try to provide with all GCP services, the as much value as possible with as little administration as possible And to show you how Cloud Spanner really works, I am pleased to introduce Greg DeMichillie >> GREG DEMICHILLIE: Thanks, Urs Thanks, everyone I hope you like live demos because this is the first of four live GCP demos you’re going to see in the keynote today And I couldn’t be more excited to have Spanner be the first one To give you a picture of how amazing Spanner is, let’s start at the beginning, just creating the database Those of you who have actually tried to deploy a multi-region database that’s highly scalable, highly reliable, and highly reliant, know that there’s like a zillion steps You’ve got to configure machines, you’ve got to set up replication, you’ve got to worry about failover, you’ve got sharding, what happens if I have network issues, and then there is the operational side of just keeping the thing running, having to have whole staffs of people who do nothing but take care of machinery With Cloud Spanner, we turned that into one machine Robert is going to bring up the dialogue to do that He names the instance, he chooses whether he wants it to be a single or a multi-regional database, and then he says how many nodes And in this case, he picks 42 because that’s always the right answer So that’s it We’ve now created a multi-regional database that has amazing capabilities But that’s getting started Let’s look at Spanner in an actual real-world scale example Imagine you’re a ticket seller You want to sell tickets to events worldwide to customers worldwide Now obviously it’s pretty important that you sell a ticket once and only once, right? And it’s also pretty important that customers get a good performance experience wherever in the world they are buying their tickets In this example, we’ve deployed Cloud Spanner into three Google regions – U.S., Europe, and Asia – all connected by that private network that Urs told you about We’re also running workloads that emulate customers buying billions of tickets In this case, we’re selling at about half a million tickets per minute through this system, all through a system that is maintaining consistent version snapshots, full asset support for transactions, full, strong consistency That’s the power of Spanner as a distributed base, as a distributed database Our customers get a low latency experience; we get good old traditional SQL consistency

Now to prove that this really is a database of some size, let me show you the schema behind this database You can see it actually is a SQL database There are columns that are joined together And if you look at that ticket table, it has three billion rows in it So while we’re doing this, in fact, we’re running all this traffic against a real SQL database In fact, we can actually run SQL queries against that database while it’s being hammered with all that traffic Here we have a query And if you look at the bottom part of the where clause, you will see that we are looking for tickets in the U.S that are still available as of 10:00 PM and how many seats they have So these are our top five events with unsold tickets Now what’s important here is what you don’t see What’s remarkable is how unremarkable this SQL looks If you have deployed a sharding system, your SQL is a mess because you’re trying to have to do joins across multiple shards of your database In fact, it’s probably not even human writable SQL when you use those systems Spanner takes care of all that But you know real-world applications aren’t static; they change over time What happens when your needs change? Well, first of all, Spanner lets you make transactionally consistent schema updates with no downtime, which is critically important when you’re running systems that your business depends on And what happens if you get a wonderful success and your needs grow? Well, with a traditional SQL database, you better put a PO in for a brand new server, you better figure out how to configure the new server, you better figure out how to migrate the data, you better figure out how to do cutover Or if I’m doing scale out, well now I’ve got to add a whole new complex middleware layer that’s incredibly difficult to maintain Even that takes time and adds operational risk In Cloud Spanner, Robert is going to go back to the console and show you how we add capacity to a Spanner database We simply go from 70 nodes to 99 nodes That’s it We’ve scaled out That’s the power of a horizontally scalable SQL database And, of course, since we use Spanner so much internally, we made sure it handles large database Robert is going to switch over to the monitoring console here and what I want you to look at is that graph on the bottom left We’ve been running this whole demo against an 80 terabyte database And if you look at the top right graph, when we loaded the data in, we loaded the data at an average of 800 megabytes per second, and we peaked at about a gig per second of loading data That’s amazing That’s crazy fast So that’s Cloud Spanner You create a highly reliable performant, fully managed, multi-region database in seconds Cloud Spanner is SQL You get interoperability in the power of SQL queries, you get ASIC transactions, plus you get advanced features like on the fly schema changes Cloud Spanner is fully managed We never had to worry about setting up replication, sharding, failover The system actually even tunes itself automatically over time So, as you use it, it learns your query patterns and it optimizes its own performance characteristics And Cloud Spanner is built for scale We’ve been running our largest production systems on this for years So with that, I’ll turn it back over to Urs for a little bit more of the keynote. Thanks >> URS HOLZLE: Thank you, Greg So if you’re used to traditional databases, it takes a while to absorb that I just want to summarize because Spanner is truly something new It is a SQL database, a traditional SQL database, with the same semantics, that just scales by changing a value on the admin interface It scales through thousands of nodes and has the same strong consistency guarantees that your applications have come to depend on And, of course, as you saw, easy to administer, does replication automatically, multiple locations worldwide, all without any setup It has been immensely popular inside of Google, and so we’re very, very excited that we have it now available on GCP Now from databases to compute Many of our customers have demanding workloads, financial risk models, movie rendering, scientific computing, large ARP systems, and so on And so starting today, GCP VMs will come with up to 64-cores and 416 gigabytes of memory And like our other VMs, these VMs are available as pre-emptible VMs So if you have some time or flexibility in your computation, you can use them for batch workloads and get much lower prices We’re not stopping here So, later this year, you will see even higher core counts and memory sizes of a terabyte or more In fact, when it comes to hardware innovation and to

optimizing the hardware for the best performance in the Cloud, we’ve been working very closely with industry partners throughout the industry to specialize these systems and to accelerate the innovation in hardware One of these partners of course is Intel Please welcome RaeJeanne Skillearn from Intel to talk about our partnership. RaeJeanne >> RAEJEANNE SKILLEARN: Hello >> URS HOLZLE: Thank you for joining us, RaeJeanne It’s great to have you here >> RAEJEANNE SKILLEARN: Yes. Thank you >> URS HOLZLE: We started to work together quite a long time In fact, I don’t quite remember even how long ago it was Can you tell us about the history? >> RAEJEANNE SKILLEARN: Yeah It actually started in 2003 with our Intel Core 2 processor Google was able to give us in-depth performance analysis and real-world benchmarks and enabled us to tune our processor for their unique workload Fast forward to 2008, we started working on the Intel Xeon processor, the 5600 at the time And since then, we have optimized six generations of Xeon processors specifically for Google’s environment Now in November of 2016, we took that collaboration and we expanded it Your Diane Greene and our Diane Bryant announced a strategic alliance between our two companies We are now collaborating on Hybrid Cloud orchestration, security, machine and deep learning, and IoT edge to Cloud solutions >> URS HOLZLE: That’s right We have had six generations of custom processors and that required a lot of work We actually learned a lot from that Can you tell us how we worked together on this and how this actually works in practice? >> RAEJEANNE SKILLEARN: I can And it starts early It starts in the absolute earliest phases of our architecture and CPU process development We take the Google feedback every step of the way and we incorporate it and we iterate and modify our processors We know that to truly create the best technology that is performance optimized, it really relies on the software as well That’s why both of our companies are investing in the TensorFlow and Kubernetes open source projects Intel is actually going to contribute code to both of those projects so we’ll dramatically improve the performance of both Kubernetes and TensorFlow on our Intel Xeon processors and our Intel Xeon Phi processors >> URS HOLZLE: So three weeks ago we launched the newest server, Intel Xeon server processor codenamed Skylake on GCP, but yet you can’t buy this processor anywhere and Intel is not planning to launch it for quite a while >> RAEJEANNE SKILLEARN: For a while >> URS HOLZLE: How does that work? >> RAEJEANNE SKILLEANER: Well, first I would like to congratulate Google on being the absolute first to have Cloud services on our next generation Xeon processor, Skylake >> URS HOLZLE: Great Thank you. Thank you >> RAEJEANNE SKILLEANER: There was a trick to doing this First we had to accelerate production readiness of a targeted set of features that are a little bit different than our broad feature set, all the systems in the software that we have to qualify and validate for the broad general availability Intel and Google started an early definition on a custom skew all the way through significant joint investment in the validation of those systems into production in your environment It was a team effort many steps of the way, at massive scale and in a very aggressive timeline for both of us >> URS HOLZLE: Well thank you for being such a great partner We really enjoyed working with you >> RAEJEANNE SKILLEARN: Thank you We enjoy it We look forward to tomorrow >> URS HOLZLE: We’re very happy that our customers benefit from this decade of collaboration between us and Intel Because we’re pushing the envelope in so many directions on performance that we have to really, really work very differently with vendors Skylake offers great performance for compute-intensive workloads We are very, very happy that Skylake is available at first on Google Cloud Now let’s go from performance to economics Clouds, as you know, are supposed to be elastic and free from capex or capacity planning, yet some Cloud providers force you to pay up front for three years to get the best price But if you have to buy a VM for three years, then how is that better than buying your own server for three years? There’s no flexibility And, in fact, a recent study showed that Cloud users waste on average 45% of their spend on resources that they bought but can’t use That’s caused by many factors One is these three year leases that force you to predict your future perfectly And let’s face it; none of us can do that We end up with stranded resources that we actually

can’t use Similarly, we are forced to buy – like on premise – we are forced to buy servers in fixed sizes, and so we end up overbuying on some dimension, and we still pay for it even though we can’t use it And last but not least, you pay for the full hour even when your test run runs for only ten minutes So you pay for compute time that you’re not even using All of that adds up to 45% waste Now we believe that it should be easy to get the best price, and so we have solved these sources of waste In fact, in 2014 we introduced automatic sustained use discounts that give you discounts without any long-term agreement So as soon as you use a VM for more than a quarter of a month, you start seeing savings And if you use the VM for the entire month, you get a 30% discount But you’re still free to stop using a VM at any time So sustained use discounts brought the Cloud back into Cloud You can change your mind at any time but you still get great prices And you don’t need to accept VMs whose shapes come in powers of two So with other providers, if your application needs say 20 cores and 50 gigabytes of RAM, you might have to buy the next larger machine size that might have 40 cores and 160 gigabytes of RAM And so you end up paying literally twice the resources that you need But not on GCP, because with custom machine types, you can dial in exactly the configuration you need so you pay for exactly the resources you need Today, over 20% of our core hours on GCP are for custom machine types and our users save an average of 19% through the customization And we even help you save money with our right sizing recommendations That’s a service that looks at your memory and CPU usage and then suggests the best virtual machine size And last but not least, GCP has permanent billing for VMs So if you use a VM for 11 minutes, you pay for 11 minutes It seems pretty logical So GCP and only GCP is truly an elastic Cloud You only buy what you need, you only pay for it when you actually use it, you get automatic discount if you use it for extended periods of time, and we even alert you when you appear to be wasting resources And when you put all of this together, you save an average of 60% relative to what you would pay on other Clouds So given the complexity of Cloud pricing we just went through, it’s not surprising that the same study I quoted says that 53% of Cloud users say that optimizing and controlling spend is their top problem in the Cloud But not with GCP Because with GCP, you don’t need to create an entire new ministry in your company just to get the best price because our flexible pricing structure lets you enjoy a Cloud as it was meant to be – on demand, pay as you go, paying only for what you need But today it gets even better because we’re introducing another way for you to save – committed use discounts So in exchange for a one or three-year commitment, you receive a discount of up to 57% billed monthly, no upfront payment But these are not inflexible reserved instances that lock you into a particular instance type or family that force you to pay up front. No They only commit you to an overall volume for your compute and memory So you’re not locked into any particular VM size You can change individual machine types and VM shapes at will You can change the number of VMs You’re only committing to the aggregate volume and not the details And if you’re not sure how much commitment to make, you can start smaller because you still get sustained use discounts and all the other benefits on any of the usage above your committed usage And so you’re not facing a huge unit cost cliff when you exceed your forecast Because we believe that when you move to the Cloud, capacity planning and cost planning should really become a distant memory and not your number one headache So on GCP you are saving money automatically with no regrets, no spreadsheets, and no PhD in economics needed to manage it all In fact, one way where you see this flexibility is our GPUs Many of our customers already GPUs for a variety of

applications to speed up simulations, transcoding, deep learning, computational chemistry, finance, and many more But on GCP, you can add a GPU to any VM configuration and start it up in under a minute, much faster than our competitors So that means that if you want a GPU with lots of memory, you can get that Or if you want a GPU with little memory but lots of cores and an SSD, you can get that, and you don’t waste money on your GPU because you have to buy some other stuff with it And, of course, GPUs also have per minute billing so you don’t end up paying when you’re not actually using it So to help you learn more about this, I want to introduce one of our customers who is using GPUs as a critical part of their business Please welcome Ashok Belani from Schlumberger Hi, Ashok Great to have you here >> ASHOK BELANI: Thank you for having me >> URS HOLZLE: So Schlumberger moved to Google platform last year What drove you to adopt Google? >> ASHOK BELANI: At Schlumberger, we’ve been leaders in high performance computing for many decades You see the size of our compute clusters growing on the slide We bought the first grade back I think in the mid ’80s and we moved to massively powered PCs in the ’90s And then we were actually one of the first companies to jump onto very high GPU to CPU ratios back in 2007 or 2008 What we think is that we can go to the next level of compute possibilities working together with Google >> URS HOLZLE: So you clearly use massive amounts of compute Tell us what you’re actually doing with compute and those GPUs >> ASHOK BELANI: So we are looking in the subsurface for our geologists who are exploring for oil and gas deposits So we look very deep into the subsurface, tens of thousands of feet We use acoustic data and dynamic data of different kinds Here you see on the slide behind me a vessel which has a huge spread of sensors which are being pulled behind it There are millions of sensors on a spread which is ten kilometers by two kilometers Sometimes we use two of these boats which would cover half the surface area of San Francisco And, as you see, the amount of data that we generate has been increasing over the years significantly And then we use high performance computing to create images of the subsurface for our exploration efforts >> URS HOLZLE: So tell us how your experience has been How has it been working with Google? And tell us actually what we’re seeing here >> ASHOK BELANI: So we see an image here which actually has been – the data has been acquired over about five months This data was actually processed in Google over a period of three months And you see the subsurface which is actually 20,000 feet into the ground and it is 50,000 square kilometers off the shore of Mexico – bigger than the size of Switzerland >> URS HOLZLE: Our mountains are taller No need to go on the ground >> ASHOK BELANI: I think the interesting thing about working with the Google Cloud is that we are able to create basically a cluster which is suited for that particular problem because of the algorithms we are going to use on that problem Within a matter of minutes, we are able to mix the CPUs, GPUs, high GPU ratios, memory close to the processor the way we need for particular algorithms to work in a very good way This is a big advantage that we get out of Google But I would say another advantage is Urs Holzle himself I think Google should be very proud of having someone like this >> URS HOLZLE: Okay. Yeah Let that be stricken from the record That was not in the script >> ASHOK BELANI: That wasn’t in the script We certainly cooperate very well with the engineers in Google We actually created a center here in Menlo Park so that our engineers could work together with Google very closely And we are only at say 10-15% of the way on this journey to make Cloud very efficient on our applications We think we will be able to weave in things like big data type of technologies or analytics or machine learning In the future, they will be woven into our applications and I think we will definitely be able to achieve the next level of computing in oil and gas And I think we serve all the oil companies in the world So as we go on this journey, they will come together with us

>> URS HOLZLE: Thank you very much It’s really been great working with you >> ASHOK BELANI: Thank you very much >> URS HOLZLE: Thank you, Ashok All right So when you use Google, you use the same security infrastructure that Google uses And we’ve been investing an enormous amount of effort into that security You can see an overview of the many layers here I don’t really have time to talk about that today Today I would like to show you just a few highlights But we have a detailed security design, white paper, and we have a one-hour session that do real justice to this topic So we put a lot of effort into security starting with physical security For example, in this single Google data center campus we have over 175 security guards on staff 24/7 in addition to countless cameras, motion sensors, iris scanners, and so on Many of our locations are not individual data centers but data center campuses And so this very high level of physical security is amortized over hundreds of thousands of servers And thus on GCP, low prices don’t mean low security standards In fact, to protect the security of our hardware, we put a security chip on all our new machines to serve as the basis of trust for that machine’s identity So this is a custom chip designed by Google so we know exactly what it does and it helps us protect servers against tampering even at the bios level; or in this particular example, it helps protect the bios of the networking card that we built The chip, in fact, is so small that I’m actually wearing one on my earring here So here it is So now you know that you are watching the authentic GCP keynote So this hardware chip helps us authenticate the hardware And then on top of that, it helps us authenticate the services that we run When services call each other, so the services that implement GCP, they must mutually prove their identity to each other and use certificates for that and the binaries are signed and cryptographically signed so that we can verify we are running the right binary And then on the storage side, GCP’s durable storage services encrypt all data before it is written to the physical media And on the networking side, all Internet traffic from and to G Suite or GCP are protected with strong encryptions and multiple layers of service protection Now that covers the data center side But a system is only as secure as its user accounts Today, phishing probably is the number one security problem for enterprises, meaning someone is trying to trick your users into providing their password and perhaps their one-time token But your G Suite and GCP account is already protected by a sophisticated abuse detection system to thwart those attackers that are trying to guess your password or trying to use a stolen password But with our optional phishing-resistant second factor, we can provide a very strong defense against phishing No other Cloud today provides you this protection against what is probably the number one security problem in enterprises For further end-to-end security, we also ensure that the user’s client device is secure So Chrome, Chromulus, and Android are designed ground-up for security, all featured Cloud management and frequent over-the-air security updates And our client operating systems feature a hardened boot So they actually have a chip similar to this to verify that they’re booting the correct software They have encryption of storage by default and enterprise grade management And with such a strong security stack and nearly zero administration cost and a wide range of models to choose from, it’s no surprise that Chrome Books outsold Macs last year >> URS HOLZLE: Yes, thank you So today we are announcing several new security features that we’re adding to GCP and G Suite First, some tools to let you secure your data

Data loss prevention is an API now available that lets you discover PII and other sensitive data in your content and take appropriate actions such as redacting it You will see a demo in a minute The engine behind this API is the same engine that powers the data loss prevention feature in Gmail and Drive, so you get consistent results everywhere Next, Cloud Key Management Service is now generally available It lets you manage encryption for your Cloud services because Cloud KMS protects your data and secrets that are stored at rest and it can automatically rotate your keys for you And then we also added tools that let you safely access that secured content once it’s been secured The first one is the Identity Aware Proxy or IAP It is available in beta and it enables you to configure secure controlled access to your applications So you can enforce a who can see what access control at the application layer So you don’t need client software, remote access VPNs, firewalls, network configurations You just deploy IAP with a single click and you’re going to see that in a second So IAP acts as a smart front-end in front of all of your Cloud applications It’s built on top of the GCP load balancer so you benefit from the transport security and the scalability of the Google front-end service And last but not least, security key enforcement is now available to GCP users as part of G Suite Enterprise So this feature lets you enforce security key use for all members of your domain That means that all of your applications – access to all of your applications is now strongly phishing-resistant Here to demonstrate some of these features are Greg DeMichillie and team >> GREG DEMICHILLIE: Thanks again, Urs You know, today the task of taking a corporate application and making it available to your employees outside of your network is just painful You’ve got to set up VPNs and nobody likes those They’re hard to set up They’re hard to configure And in the end, you end up with this perimeter-based security, right, with a semi-hard exterior and a soft underbelly, which isn’t nearly as effective as modern device multi-factor authentication So we’re going to give you an example of how easy it is to use the Identity Aware Proxy to take a typical corporate enterprise app and make it available to your users Now the specific app we’re going to use in this case is as enterprise as it comes It’s Oracle E-Business Suite running on Google Cloud platform And as a company, we want our users to access this application whether they are in the office or they are out on the road So let’s see how simple that is to do with IAP Neil is going to click the button to turn on IAP He’s then going to tell us what domain he wants to publish that application for And now users in that domain and only that domain can access the application So let’s test it out He’s going to flip over to a browser tab Now the first thing is there is no VPN software installed on this machine We are just navigating to the URL Now first he’s going to try his personal Gmail account which has not been authorized for this application Sure enough, he is denied Now he’s going to try it with his corporate account which has been authorized and he gets the username, he gets the password, but now he will get prompted to enter his second key authentication So if you look on the sides, I think we’re going to project that up in a second Well, okay, trust me, he pushed – there it is He pushed the security key I think you know what that looks like And now he’s taken directly into the application without having to have another additional login step needed Not only is this easier for the developer, it’s easier for the admin, and lord knows it’s easier for the end user rather than hassling with a bunch of VPN software, and it’s more secure because we’re not trusting anybody who just happens to come in through a VPN tunnel So that’s IAP The other thing Urs talked about was data loss prevention Let me show you how Google can help you secure some of your most sensitive data Most companies today have a policy around data minimization, the goal being to minimize the amount of data that you collect and store to only the data that’s actually needed to run your business But that’s easy to do in words but hard to do in practice Data has a way of leaking out all over the place So let’s look at a concrete example We have an enterprise application that a support agent might use to chat with your end users And in the chat you can see Neil is a support agent and he’s asked Alice for some information to verify her account

Well, Alice has way over-shared We just wanted the last four of her Social She gave us her full Social, her phone number, and a picture of her Social Security card Now obviously we want to store this chat, right, to see how our conversations are going or how the agents are working, but I can’t store that in a database I’m going to have half my company now knowing Alice’s Social Security Number So how do we do that? DLP makes it easy When he ends the chat, we will use Data Loss Prevention to identify and redact sensitive information So now you’ll see that we’ve – yes We recognize over 40 different types of sensitive info types In this case, we’ve identified Alice’s name If he scrolls down a little bit, you can see we found her phone number, her Social Security Number, we replaced all those with red dots, and even we went into the picture of the Social Security card, found the Social Security Number there and redacted it Now this chat can be used for analysis Now just to prove this is real, Neil has a webcam over there and we’re going to try one more example I was in the green room a while ago and I found this credit card sitting there and it said Urs Holzle I’m just kidding Neil has a sample credit card with a valid number but it’s not Urs’ card I trust DLP I don’t trust 10,000 of you with cell phone cameras So he’s going to put the credit card number on it, he’s going to hit the button, we’re going to use DLP, and it will find and redact the credit card Now this isn’t just dumb OCR You’ll notice it didn’t block out the expiration date or the name because we told the API that the only information we wanted to eliminate was credit card numbers So you can control your definition of PII, what matters to you So there you have it Identity Aware Proxy We made it super simple to take a corporate application and make it available to our end users without the hassle of complex network configurations And with Data Loss Prevention, we were able to help it make sure that you minimize the amount of data that you collect so that’s one less headache you have in terms of compliance or regulatory or really just running your business And with that, Neil and I will turn it back over to Urs again Thanks, everybody >> URS HOLZLE: Thank you Thank you, Greg So, as you just saw, it’s really easy to use the IAP Proxy to control access to your applications But we’re already working on the next version based on a principle that we’ve been applying to our own corporate users for a while Because we view every access decision to resource as not something that’s just about the user credentials and maybe their second factor but really about something that should be based on the context around it; for example, the state of the user’s device, their location, and so on So we call this Context Aware Security The user’s context determines access, not just the network they’re on or who they are The context that we have today with the IAP Proxy is just the user identity and the security key – so the strength of authentication But you can expect our future versions to use a richer and richer context over time to better secure access to your Cloud applications or your G Suite Now I am thrilled to introduce Brian Stevens to tell you more about how customers adopt GCP. Brian >> BRIAN STEVENS: Good morning Thanks, Urs So, public Cloud has absolutely exited the early adopter stage It’s now a shared platform, available to everyone from start-ups to the world’s largest enterprises And it’s quickly gone pan vertical from financial services, healthcare, industrial, government, everybody is choosing the Cloud And when they pick GCP, why do they come to GCP? A lot in common They want the world’s best security They want to be on a flywheel of continued innovation, all available through an API They want the best in the world data and analytics And the most important thing, they want the tools and the platforms that their developers love I’ve spent a lot of time working directly with customers And when you actually zoom out and you try to look at patterns, we’re actually seeing three distinct things when they come to GCP The first is wholesale migrations of their existing workloads from on-premise to GCP, which, to be honest, was a little bit of a surprise The second is building Cloud native applications

That includes start-ups as well as some of the world’s largest enterprises now And the third is just coming to Cloud to get the richest set of data analytics and machine learning So our first Cloud service was back in 2008, Google App Engine, almost nine years ago, and it was an incredibly powerful platform as a service, possibly too advanced for its time We’re going to talk about that a little bit more soon But most enterprises either can’t or don’t want to rewrite their application architecture just to move to the public Cloud And what’s amazing is that even without that rewrite, the drivers for moving to GCP are still incredibly compelling They get the amazing security that you just saw, they get to reduce their capex, they get incredible reliability, Google’s pretty good about reliability, performance, they get it for free, look what you just saw with Skylake and what’s happening in the network, and they get this improved efficiency So I actually kind of loathe, to be honest, industry conferences, but there’s this one CIO conference that I go to every year, and it’s put on by Excel, and it’s a small, little, intimate venue with about ten CIOs And two years ago the top things when we went around the room that they cared about was security, what kept them awake at night, and mobile handsets The two are really related A month ago when we went back again, same group of CIOs, they went around the table and the top two things that were keeping them up, security still, and then shutting down data centers Every one of them was shrinking their data center footprint I felt a little weird because – and then it was my turn and I was talking about we’re doing the opposite Like look at what Urs just showed you So it’s kind of clear where these workloads are going So, ideally, you want to be able to shift the workloads with as little change as possible because you don’t want it to be a lot of work just to move to Cloud, and the more change you introduce, there’s more risk But that’s just the first chapter Once you get to GCP, often what happens is chapter two, and that’s when they actually look at how do we re-factor? How do we use things like a managed database? How do we use something like Spanner? So our call to action at Google is how do we make that as simple and easy as possible for them, and that includes technology, processes, and people One thing that we did is we recently added right from the Cloud console the ability to migrate a virtual machine to GCP This is more than transcoding an offline image It’s actually the live migration of a running server to GCP What I love about it is that it’s hypervisor agnostic, so it works on bare metal, it works on Hyper-V, ESX, KVM, even from another Cloud Also, we’ve been making big investments in Windows because we want to make GCP as great for Windows developers as it has been for Linux and Open Source developers Our goal wasn’t just to be an okay Windows platform We want to be a great Windows platform, perhaps the best Windows platform So we already have support and pre-built images for many flavors, a Windows server, as well as SQL server And we also support active directory running in the Cloud, and you can integrate that with your on-premise domain controller But it’s really important for us that for developers and Windows developers that we meet them exactly where they are We don’t want them to have to change how they do things just to be able to take advantage of Cloud That’s why we did the Visual Studio integration that we’ve done with GCP so that you can actually deploy dot net apps from Visual Studio and then just manage all of your Cloud resources And we’ve also integrated with hundreds of cmdlets for PowerShell right into our Cloud SDK And so now they can be very comfortable managing all of their Windows-based Cloud projects So today we’re announcing the general availability of SQL Server Enterprise and that actually includes support for a high availability as well as clustering Also, a beta dot net and that will be available in both App Engine and Container Engine And to help people on this journey, we’re announcing a new Windows partner program And so we’ve partnered up with a number of top specialists that actually have great Windows expertise as well as GCP expertise So they can help customers on their journey to move Windows environments to GCP So now Linux or Windows, my SQL Oracle or SQL server, and as of today also a beta of Managed Postgres, developers can really easily migrate to GCP with minimal refactoring of their application stacks So what’s the best test for moving to the Cloud with minimal friction? Probably to be able to do that with zero downtime

It seems pretty impossible, but Evernote actually did it last year They moved their entire software infrastructure from their on-premise data centers to GCP – 200 million customers depend on Evernote every day – and they did this migration in 89 days with zero downtime And so Lush, a British cosmetics company, they began their migration to GCP They started last year in September and it was critical that they finish in time for the holiday season and they did it And so here, I would like to introduce Jack Constantine from Lush to tell us about their journey Hi, Jack >> JACK CONSTANTINE: Hi, Brian >> BRIAN STEVENS: So 22 days I don’t think anybody is going to believe you It seems impossible How is that even possible? >> JACK CONSTANTINE: Yeah So in September we started discussing doing the whole migration The actual migration was from December the 1st until December the 22nd So it’s not just a little bit business critical; we’re talking peak trade time Yeah, so it was a huge deal I mean one of the key things, I couldn’t have done it without the in-house engineering team that we have, some of the guys in the audience I wouldn’t be up here if it wasn’t for all the hard work those guys put in, the hours, the dedication, the focus I think they deserve a round of applause We found ourselves in a bit of a tricky spot We were in a contract that we weren’t really comfortable with We wanted to be able to actually have a look and see what else we could go for The contract ran out on the 22nd of December, hence the reason that we had that hard deadline >> BRIAN STEVENS: Lucky us >> JACK CONSTANTINE: Yeah I’m a little bit of a risk-taker myself, as you can tell Basically there is very little bureaucracy in Lush, so the ability for me to be able to actually make that decision was quite fast and then the team just powered through It was a really exciting time We were really, obviously, so pleased with the result >> BRIAN STEVENS: So were the like any sort of challenges with the migration, technical inhibitors, process? >> JACK CONSTANTINE: Well I think, like with any migration, there are always technical challenges You’ve got to worry about getting your data from one place to the other and the amount of it and are you going to make sure you’ve got consistency We moved 17 websites from all over the world with customer data, with order data, product data Obviously you want all of that to be completely up and stable But I think one of the main things really from my perspective was the kind of commitment to actually achieving it I think sometimes people can – obviously when you’re going to throw a kind of hard deadline like that, it can sound a bit unachievable But I always like to think it depends on kind of which reality you’re looking from I like to think that it’s realistic; you may think it’s not realistic. Right? I think it’s realistic so I think we should try and do it So keeping that focus on people actually kind of believing that this is a goal we can achieve in the timeframe and not letting people start to put the blockers up and go, oh, we’re going to have to delay this, we’re going to have to delay that All of a sudden you watch everyone – I mean, yeah, everyone gets very tense, but also you achieve a lot You actually get through it and the things you need to get done get done >> BRIAN STEVENS: Anything that could have made it a little bit easier? >> JACK CONSTANTINE: Oh, definitely I think the awkwardness of our previous supplier and the fact that it was a very closed environment made it very difficult for us to get visibility of everything we wanted to be able to move over >> BRIAN STEVENS: Open wins again >> JACK CONSTANTINE: Exactly Openness is something that we absolutely cherish in Lush Obviously you guys do at Google, which has been great for us And the other thing that we had, we had this great – one of my colleagues had this great conversation with a Google partner on the phone and it was about a week before and they were obviously getting a bit scared, oh, is this going to happen, and they were on the phone The partner said what is Plan B? My colleagues said Plan B is to make Plan A happen >> BRIAN STEVENS: That’s a great line >> JACK CONSTANTINE: That’s it. That’s it And we made it happen >> BRIAN STEVENS: That’s cool Urs, the last speaker, was the one who recruited me to Google I was already sold because I believed the future is public Cloud and just fascinated by it You have to be a technology company to win this and just everything that Google has been investing in for years in technology But the surprise for me when I actually got here was the culture, right, the ethics, the diversity, the focus on inclusion, sustainability And in our chat yesterday, it sounds like there are a few similarities in your culture that you and your parents have >> JACK CONSTANTINE: Yeah. Yeah So my parents founded Lush over 20 years ago and there are a lot of ethical values that we’ve built throughout the

organization, and we pride that in kind of everyone we go through, supply chain when we’re buying ingredients, and fair trade, and looking at the best quality ingredients, we look at that with our packaging I represent the more digital side of the business, obviously It’s a very interesting time and Lush is looking at its digital future and understanding where we go and how we navigate through the landscape I think there are a lot of similarities with Google in terms of the openness versus closed, the cultural elements It’s been great to be able to – one of the reasons we were so eager to do the migration in that time period was because it felt like by moving over to the Google Cloud platform we would be aligning with our ethical values on a much higher level Things like the renewable energy and the open mentality, all of those things, we’re looking at the moment around digital ethics policy Actually, just before we did the migration in November, we did a global campaign to support keeping the Internet on, especially in countries where the government may shut the Internet down because they don’t want to encourage communication Obviously kind of the reverse of what we wanted to happen with our migration Thankfully we also kept the Internet on when we migrated our websites But, yeah, the synergy between Google and Lush is great I’m really excited about even the things you’ve been showing today and being able to work together much more on prototyping new ideas, having that flexibility I spoke to Urs earlier and he was saying about that whole kind of engineer-to-engineer dynamic And my team straight away, they were absolutely buzzing We’ve only been working with you guys for the last 3-4 months but the energy is huge It’s great >> BRIAN STEVENS: Well, my wife and daughter have always been big fans, but I’ll say I’m a convert now as well Thanks, Jack >> JACK CONSTANTINE: Thanks very much >> BRIAN STEVENS: That’s a good story So Evernote and Lush are great examples of a complete migration to GCP from either another Cloud or on-premise But for the largest enterprises, the move to Cloud isn’t a point in time It’s going to be a perpetual state to run in a hybrid environment And what we don’t want is we don’t want this steel curtain between public Cloud and private It should really feel like this data center extension, albeit this really amazing data center extension So cornerstone to that is Virtual Private Cloud, VPC That really allows enterprises to build these really nice integrated hybrid environments It gives them a completely private virtual network running inside of GCP It used to be that everything running on Cloud had public IP addresses Now with VPCs you can have a completely private environment, private IP addresses, private DNS, and full control And you want to be able to also control what applications outside of your VPC can actually have ingressed in as well as to make sure you have full control of anything running inside your VPC that you allow to talk to the outside world And on top of all that, you always want full auditability and full telemetry so that you really see everything that’s going on, even inside of these managed services They should not be opaque Also, you don’t want application in data silos forced on you just because of this move to public Cloud That was really one of several drivers behind the acquisition of Apigee So with Apigee, you can actually put really elegant APIs in front of technology so you turn them into consumable services and that allows you to integrate applications in these services whether they’re built on Cloud and integrate them into on-premise into your enterprise or on-premise and integrate them into your Cloud applications So soon Chet Kapoor, the VP of Apigee, is going to be on stage and he’s going to go into far more detail on how they’re helping in building connected business platforms So Docker has been this amazing thing It’s been this amazing gift to the industry what Docker has created because it’s really the first time that you’ve actually been able to build these consistent application stacks that run across hypervisors, different operating systems, and different Cloud It’s not perfect in terms of compatibility but it’s really setting us free to do some amazing things And it’s because of Docker that Kubernetes is so successful Because now with Kubernetes, you need a control plane, you need to be able to manage and orchestrate these container-based environments and that’s what Kubernetes does So it’s really quickly become that de facto operational model And because it’s open source, we’re seeing customers, enterprises run Kubernetes on-premise and then they’re running our managed service for Kubernetes, the container engine, on Cloud It gives them this single operational model that’s

entirely consistent across hybrid environments or they can integrate them all together and run a single control plane on GCP to even manage the on-premise world Serverless is a really important concept Developers shouldn’t need to think about managing infrastructure Servers should be provisioned automatically and just sized to the workload And engineered right, it’s more reliable, it’s easier, it’s more efficient The spoiler alert is the servers are still there When you need them, they become plentiful And when you don’t need them, they just go away It’s really been a design approach Serverless is not new to Google It’s been a design approach across many of our major services On the compute side, App Engine and Container Engine are serverless On the database side, Datastore and Spanner you don’t see infrastructure, and also with BigQuery Each service consumes no servers when there is no load and they all scale out horizontally for when you need more horsepower And today we’re announcing the beta of our newest serverless offering – Cloud Functions So functions are simply fragments of code and they get applause That’s great But what developers do with these fragments of code is they connect services together and they plug them into a growing corpus of events across GCP and that’s how they tether it in But what it does is powerful It lets you take this generic Cloud that’s good for everybody and you can personalize it in a way that is meaningful to you During the alpha, we saw people do some pretty amazing things One example that I love was one company actually looked at event logs and then they keyed off certain events in the event logs and they actually automatically filed bugs in JIRA I think that was pretty cool And then we also see people doing PII scanning They plug in for the PII they care about, and any time an object comes into GCS, they can actually scan for it So this ability to be able to extend GCP in a way both for developers as well as system administrators are virtually endless So I mentioned earlier that back in 2008 Google App Engine was really this pioneering serverless and it was ahead of its time The core promise to developers still remains the same – bring your code and Google will handle everything else It empowers applications to scale from one request per day up to millions of requests per second And so when you actually liberate developers from managing and patching servers, dealing with scale, dealing with load balancers, version rollouts, managing databases, they can create great things That’s what’s been happening at Snap, Home Depot, and Philips Lighting And internally to Google, App Engine has been around a long time So our corporate IT, we have thousands of App Engine-based apps in production that Googlers depend on each and every day So beginning today, we’re delivering a promise to Google App Engine but we’re making it available to entirely expanded developer communities It’s a focus on more openness, developer choice, and expanded portability So out of the box, we now support seven popular languages – Java, Ruby, Go, PHP, Python, C#, and Node – or you can now, for the first time, bring your own runtime, bring your own framework to App Engine Flexible Environment As long as it runs in the Docker Container, it now runs on App Engine Thank you So I can’t think of where serverless would matter more but to mobile developers Do you know a mobile developer that wants to manage operating systems? What they want to focus on is building a great user experience for their users So we’ve been working really closely with the Firebase team, which is Google’s mobile application development platform, which supports iOS, Android, and Web They actually jut surpassed a milestone last month In the last 11 months, they now have one million active projects on Firebase So, really amazing momentum Today we’re actually making Firebase even more powerful because we’re integrating it closer so that you can access GCP resources right from Firebase The first thing we did is we integrated Cloud storage into the Firebase SDKs, so now you can access any GCS bucket anywhere in the world right from Firebase So you get this direct to mobile upload and download for every Cloud storage user from a mobile platform One thing that people love about Firebase is it has this amazing built-in analytics capability and that’s what developers use to understand their users Now we’ve actually integrated that so you can actually take the analytics data in Firebase and then drive it into BigQuery

and that allows even more analysis and really understanding your users in realtime And we’ve also integrated the Cloud Functions that we just talked about with Firebase, and so that allows Firebase developers to extend their backend logic in crazy ways They can send push notifications, transform data to machine learning, define custom business logic, all with just a few lines of code In fact, the integration with Cloud Functions has been the number one feature request of Firebase developers over the last year And then finally, we love lawyers almost as much as we love developers, so we’re actually working to extend GCP’s terms of service to cover many of the Firebase products So we’re not there yet but soon you’ll be able to bundle much of Firebase into essentially the same contract that you use for GCP So we’ll be making a lot more integrations over the coming months Greg is coming back now and he’s going to show you how we use GCP and Firebase together to really modernize corporate IT apps. Greg >> GREG DEMICHILLIE: Thanks, Brian Brian told you we’re going to take an example of an existing on-premise application and we’re going to modernize it with Google Cloud Now in this case, the app that Chris and I are going to work with is an old ASP.net application Well that’s not an ASP.net application That’s an ASP.net application This is an application used by an insurance company So the adjusters go out in the field, they take their digital camera, they take pictures of the accidents, they put it in their SD cards, and then they upload the images into the application It’s also got an API on it so that you can manage all the images that are associated with a given claim Now we’re going to start bringing this application out of the ’90s and into the modern era We’re going to start by giving our agents a good mobile application so they don’t have to lug a laptop around Firebase makes that super easy You can build a mobile application without having to be a backend expert In this case, what Chris is showing you is the line of code that integrates Firebase with Cloud Storage So this line of code allows the camera to take a picture, that picture then automatically gets uploaded into a Google Cloud Storage bucket So to show how this is going to happen, Chris is going to take a quick photo with the phone. Yep >> CHRIS: All right, everyone, say cheese I love having audiences do that >> GREG DEMICHILLIE: So that photo now is being uploaded and it’s being stored into a Google Cloud Storage bucket Now to show that it really is in a GCS bucket, we’re going to switch over and we’re going to list the contents of this bucket Now the Windows users among you are going to notice that he’s using PowerShell That’s because he’s using Cloud tools for PowerShell It gives Windows developers a first class experience using Google Cloud platform with the services that they love And sure enough, there is the picture that was just uploaded today So now we’ve got a claim picture in GCS, how do we get it to this old legacy application? Apigee and Cloud Functions make that simple The good news is our legacy app has an API, as I mentioned The bad news is the app was written over a decade ago so it’s a SOAP API So it’s not really very friendly to a modern restful developer like Chris Apigee, however, is an Enterprise API Management platform We’re going to start by using it to create a proxy to convert this legacy SOAP XML into modern JSON So here is the Apigee console We’ve connected it to the WSDL of the service I can’t believe I’m saying that in 2017 And on the left side, you see the SOAP XML, and on the right side, Apigee has automatically created a much friendlier, easier for the developer to use, JSON version of it Now Apigee has a lot of other capabilities So, in addition to doing that, Chris has applied a quota so that we don’t flood this poor service with millions of mobile agents all trying to upload photos at the same time It automatically provides throttling We’ve also put authentication in so that only our agents and only those mobile applications can upload the data Now Apigee has got tons more capabilities; I am only scratching the surface But for our purposes, that’s good enough for this application So we’ve got images in a GCS bucket, we got our application that’s now got a nice modern API on it, how do we connect the two? Well I could deploy a VM, but that’s overkill, right? All I want to do is copy a little file from GCS to an API Why should I pick an operating system and have to patch the OS and deploy a big, heavy VM image? Cloud Functions allows me to just write a snippet of Javascript This Javascript is listening to that GCS bucket

So any time a file gets added, this snippet gets invoked And if he goes down, you’ll see he is using the post method, that nice, restful interface onto our web service So now when Chris takes a picture, we should see the picture go all the way through Now I don’t have a car crash on stage, Chris I think you’ve got though some sort of a next best thing for us? >> CHRIS: Actually, what I am prepared to do, Greg, is I’ve got some scale model vehicle replicas here that I am going to use to simulate a high speed collision scenario >> GREG DEMICHILLIE: This is scaled down in a digital scale up Yeah, very good >> CHRIS: You ready? >> GREG DEMICHILLIE: Yeah, go for it >> CHRIS: Vroom, crash. Ah! >> GREG DEMICHILLIE: Please tell me you didn’t misplace the phone So he’s going to take the picture of, oh, the humanity, and it gets uploaded, and now he’s going to switch back to our application, our corporate app He’s going to hit F5 because it’s the ’90s and we have to refresh And there we go, there is our application We started with a legacy on-prem app, we used Firebase to build a mobile application that didn’t even require a backend expert, we wrapped the old SOAP API in modern JSON, and Cloud Functions was this wonderful glue layer to connect everything together But one more thing This application started as an on-prem application and we’ve wrapped it with all this amazing Cloud stuff, but it’s still sitting on-prem Let’s fix that bug, too With just a couple clicks, Chris will start the Migration Wizard In about 45 minutes, we will live migrate this running service into Google Cloud platform where it runs on Google’s amazing infrastructure, network, operations, and reliability And now we’ve really taken this app and brought it into the modern era Now I’ll turn it back over to Brian. Thanks >> BRIAN STEVENS: Thanks, Greg So, great companies use data to react faster to the market, build novel new products, it even changed culturally how they work Breaking down data silos to see bigger and better pictures of what’s going on in realtime is really super important And so for GCP to enable that, we deliver an end-to-end managed analytics platform It now spins storage, data warehousing, ingestion, you saw PII cleaning, ETL, batch and streaming modes, visualization, and even machine learning It is expanded and extended by a great set of partners So Google is actually pretty good, right, about being a data-driven company But to support that, it was all about Dremel Dremel was the purpose-built analytics engine inside of Google that we depend on There is a white paper out on it It surfaced as Google Cloud BigQuery for users outside of Google It’s great because it scales horizontally in realtime It can ingest millions of rows It can process trillions of rows a second And you really need that for a lot of new applications to have realtime decision-making Think about like what’s happening with data ingestion and IoT and social So it’s great for realtime results across ever-changing data sets As part of that, it’s a big part of our data analytics platform What we’ve been doing is we’ve been actually connecting this sophisticated data analysis that’s inside of BigQuery with rich data sources such as advertising platforms So today, to make that easier, we’re now seeing the Google BigQuery Data Transfer Service What that does is it automates the transfer of data from SAS applications into BigQuery but it does it on a scheduled, managed basis So today we have connectors for AdWords, DoubleClick, YouTube Analytics And then once the data is inside of BigQuery, that’s where further enrichment happens You can integrate the ads data with weather data, geo data, your sales data So it makes it really easy for marketing teams to become empowered and they can build marketing analytics warehouses on GCP But the process of actually bringing data into a data warehouse can still be cumbersome I’ve seen stats that these data scientists spend 75% of their time just like dealing with these disparate data sources and cleaning them up so that they can be homogenous enough to relate to each other And we want to make that easier That’s why today we’re introducing Cloud Dataprep Dataprep is this intelligent data service that allows you to visually explore and clean your data so that it can be integrated into a BigQuery environment It’s as simple as using a mouse cursor You can hover over an attribute in a JSON object and you can decide that you want that attribute to be a top level column in BigQuery Or we’ve seen people take a single field that has a location address and the Dataprep service allows them to break that apart

and say I want separate fields, separate columns for state and Zip Code And it’s really smart It actually uses machine learning itself, because often in data, there is data quality issues as well And so based on machine learning, it suggests transformations to your data to make it cleaner So all of these need not be one time ingestions and transformations orchestrated under a Cloud data flow The aim is that you actually get out a batch and you’re building a streaming data analytics pipeline We’ve integrated with Data Studio out of Google’s analytics team What that now allows is you can actually visualize, create these really rich dashboards and charts almost automatically Or you can integrate all your data that’s now in Cloud with a lot of our machine learning So you can actually train whole new models and then use them for recommendations and predictions Oh, with all of this, never will you be managing infrastructure You won’t be installing software You won’t be integrating software You won’t be managing performance and scalability You’ll just get right to the mission at hand – analyzing data So, enough talk Greg is actually back to live demo again how all this works >> GREG DEMICHILLIE: Thanks, Brian This is my last of the four promised live demos Before we start, I would be remiss if I didn’t thank Chris, Neil, Martin, and Robert who have been helping with building these demos Would you thank them for me before we get too far into this? Thank you So to give you a picture of how this data platform puts together, we’re going to use a scenario of an advertising agency that puts ads in the back of those screens in New York City taxis, if ever you’ve been in a cab recently in New York There’s an ad that plays and you can interact with it To give you the picture first, this is a simplified architecture of this data solution The taxis are sending their position in realtime, where they are in the city, as well as their destination, how many passengers they have, and what sort of interactions the customers are doing Pub/Sub is then ingesting that data We’re using dataflow to process a data pipeline to handle and process all that data We’re also archiving the entire historical data in BigQuery which is a really powerful tool for you to do ad hoc analysis of all of your historical data And at the end, we want to make a visualization So why don’t we actually see the pipeline in action? So this is the dataflow visualization tool You see Robert has hovered We are ingesting a little over 15,000 new elements per second into this database of all of this information But how do you visualize all that? Google Maps is super powerful but I would hate to point 15,000 updates per second at my Chrome browser So dataflow actually allows us to down sample that event stream to give us a realtime visualization of taxi positions This is that visualization This is showing where all these taxis are in New York City at any given time Now what if I want to look at one particular customer who has placed ads with us, a particular agency? We can filter that dataflow and now the visualization will update so that we’re only seeing ads that are being placed for a specific store In this case, the Acme store, it looks like, is on the Upper East Side And sure enough, we see that we’re placing ads roughly geographically near the store But right now we’re just using a dumb rule for this It’s just a very simple rule Where is the store? Where is the taxi? Place the ads Can we do better? Can we automate this to start to take into account all the various data that we have about this? Let’s start by using BigQuery to sort of explore this dataset The query Robert has here looks at all of the taxis coming from the airports and tries to see how many of them are traveling near restaurants Now note here he’s querying this historical data even as we’re ingesting new data at 15,000 events per second So BigQuery gives him the ability to have an always up-to-date realtime view into his data In this case, we see that roughly 17% of our taxis are from the airport and near a restaurant That sounds like an insight that maybe we ought to do some ad targeting there And if you look at this query, you can see it really is a non-trivial query and BigQuery is handling all that for us So right now we are manually configuring these ads I think what we want to do is look at a way that we can do this in an automatic fashion Now we could similarly search for every different customer, but how do we build a generalizable and flexible model? That sounds like a job for machine learning Machine learning is tailor-made for taking large chunks of data and building models that give you insights out of that So we’re going to build a model that takes into account all of the data we have Where did the taxi start? Where did it end? How many people were in it? What time of day is it? Where is it going? First, we’re going to start by using, as you see in this

picture, we’re going to use Dataprep to take all that historical data, clean it up, make sure that it’s ready to use, and then we’re going to use it with ML Engine So this is Dataprep The first thing to notice is, across the top, Dataprep has automatically inferred the schema from my table I didn’t have to tell it that Now if you look at that passenger count column, you’ll notice that it looks a little funny The histogram up at the top includes some elements where it says there are 10 or 11 people Now I don’t know about you, but I’ve been in a New York City cab and I’m pretty sure you can’t actually put 10 or 11 people in a cab So that’s a common data bug Dataprep allows me to fix that in a number of ways I could delete the data I could replace the 11 with a 1 on the theory that it’s a fat finger mistake And Dataprep allows me to build simple recipes that clean up my entire data at scale It also can do things like if you look at our location column there, it’s a composite column It’s longitude and latitude comma separated I would really like to have a latitude column and a longitude column Dataprep allows me to build a recipe that splits columns with composite data into separate columns so that I now have a better schema to work with Now this just scratches the surface of how Dataprep can take this really tedious job of preparing data for analytics and make it much faster and must more reliable Once we’ve done this, machine learning, as you heard from Faye Faye, allows me to build a model on my laptop, upload it to Google, and deploy it at scale Let’s see how that actually works If you scroll down in the dataflow pipeline, you’ll see we’ve added an ML engine call So, now as our taxi data is coming in, we’re calling to machine learning with the data and saying help us choose the best ad to target for this particular user So let’s switch back over to the map with the deployed machine learning model and now you’ll see that we start seeing some interesting combinations When there is one passenger coming in towards Midtown, the taxi is currently in Chelsea and it’s on its way to Midtown, we’re presenting an offer for Manhattan Bagels because it’s a single person, maybe he or she wants a bite to eat If we pick a taxi that has more than one person in it – if you have a taxi with more than one person in it – there it is – in this case – that’s still not right That’s a one person one But they’re heading into a different location and instead we’ve chosen a winery If you pick a taxi with three people, the system would automatically recommend discounted tickets to Broadway, for example, as multiple people coming to Midtown in the evening So that’s Cloud ML making suggestions based on all our available data We used Pub/Sub to ingest a huge stream of data at scale We used dataflow to build a pipeline to give us a good analytics system We used BigQuery so that we always had the complete historical data available We never had to deal with sample data And at the end, we used Cloud ML and Dataprep to build and train at scale and model What’s important here is what you didn’t see I didn’t deploy a virtual machine I didn’t deploy patches I spent my time actually in the data That’s the point of Big Data is to spend the time in the data and not in the machines that take care of the data That’s what Big Data on the Google platform does So I want to thank you all and I’ll turn it back over to Brian >> BRIAN STEVENS: Thanks, Greg So people that know me know that I loathe buzzwords, and so I’m probably going to offend a few people in the audience, but digital transformation is absolutely one of the ones that I hate I think in part it is because I’ve never seen even two people even agree on what it even means But what I do know is this, is that markets are incredibly competitive, each year more so than past, and the new companies that actually take advantage of new technology without having any legacy move really quickly to either disrupt or create whole new lines of business The companies that win are going to be the ones that shed the mundane, take advantage of state-of-the-art technology, and actually put people to work, their intellectual horsepower, and they’re creating new end-user value for their customers That’s what we want to do at GCP is enable you on that journey So thank you and it’s my pleasure to introduce Prabhakar Raghavan, the leader of our G Suite team >> PRABHAKAR RAGHAVAN: Thanks, Brian It’s so exciting to be here What is productive work? It’s a systematic progress from a huge multitude of choices down to a decision or an outcome Whether it’s figuring out where to file the document that I just got, choosing the words for my next email, or at the epic end,

choosing all the pieces that come together in a grand symphonic work, in every instance we are going from chaos and ambiguity down to an outcome You will notice that the three examples I just gave you range from the mundane to the sublime The mundane end was where do I file the document The sublime end was the grand symphonic work Machines are really, really good at taking care of the humdrum down there, but it takes humans to do the truly creative work At G Suite, we are obsessed with the idea that computers should constantly raise the bar on what they can get done at the mundane so that our users, your employees, have more and more time to focus on truly creative work In a study of several million of our Gmail users today, we find that one in eight of their email replies is actually machine-generated and they take those machine-generated replies and send them off and they’re good So that’s a case where we are raising the bar on what computers couldn’t do five years ago but today they can take care of that and leave humans to focus on truly creative work And so what that means is while our computers may not have written Beethoven’s Ninth Symphony, maybe we could have freed him up to write nine more symphonies At G Suite, this has been an obsessive pursuit for over ten years Today, three of our apps are on over a billion smart phones We have over three million paying businesses that use the G Suite But I will say this has been an evolution, a journey for us, because we have gone from applications crafted for consumers and then outfitted for enterprises after the fact to having an enterprise first focus I’m going to give you a couple of data points for that One thing, we have begun to do early adopter programs with our best enterprise customers so that we don’t just build a product and push it out the door We work closely in the final months of development with our top customers and partners We take their feedback and refine the product and get it just right before we deliver it Here are three examples of recent early adopter programs that we have done I call these out for a specific reason Each of these apps is something that was built for enterprises only These were not consumer apps ported to enterprises There is no consumer pedigree here The Jamboard, App Maker, Google Cloud Search, every one of these is an enterprise-only product You will see the Jamboard in a bit We don’t expect too many consumers to be buying those Now at the center of productive, collaborative work is content I am thrilled to share a statistic with you today Google Drive today has over 800 million monthly active users and it’s on a tear to hitting a billion It will soon become Google’s latest billion user product We think of Google Drive as the premiere personal storage solution, but we also see an opportunity there where it can evolve from being personal storage to serving the needs of the enterprise for file sharing And as we go through this journey, I will be making five announcements now to represent steps along the way First, Team Drives goes into general availability today As you can tell, this has been one of the most demanded features from our customers It lets teams easily share content in their drive and manage the sharing Now once the content is in a drive or in a team drive, we

would like to make sure that the enterprise needs of compliance, of archiving, litigation holds, e-discovery, all of that stuff is taken care of So I am pleased to announce general availability for Google Vault for Drive content So once the content is in the drive – okay, I see a few people that are excited about that as well Thank you Google Vault for Drive has arrived, general availability All right, you say But if you’re a prospect, you look at it and go that’s all very well for content in the Cloud but most of my content sits in on-premise file servers What do I do about that? I’m excited to announce that we are acquiring Vancouver-based AppBridge AppBridge is a company that builds connectors from on-premise file servers to siphon the content up into Google Drive The fourth announcement I am making today around Google Drive has to do with the following Once the content is up there, how do you make it accessible and easily manageable for someone who has got a Windows laptop or a Mac? No more sync clients No more checkboxing which files you want to save No more worrying about how much hard drive space do you have Drive File Stream takes care of all of that seamlessly and obliviously so you don’t have to worry about all of these minutiae and can work as if Google Drive, the entire Cloud, is connecting to your laptop. Great Now once all of that content is up in the Cloud, it opens up the potential for Google’s machine learning magic Think of the following You come to your favorite file depository and you know there is a file you’re looking for but you’re scratching your head did I create it, did Joe share it with me, what key words does it have? In our studies, you spend something like 40 seconds on average trying to figure out the right filters and restrictions and browsing and keywords before you actually get to the file You shouldn’t have to There is one file you are looking for Google’s machine learning magic will build a predictive model for who you are and your activity and serve up that file before you even ask for it We call this Quick Access and today it is generally available for both Drive and Team Drives, Android, iOS, and the Web So that is the cache of announcements I had to make around Google Drive Now that’s all about content, but really where do the people come in here? So for this next segment of announcements, rather than my keep talking, we’re going to do a bunch of demos around teams and meetings for you I’m going to call on stage my colleagues Scott Johnston and Jonathon Rochelle >> SCOTT JOHNSTON: I’m Scott Johnston and I’m going to show you how we rebuilt Google Hangouts with a focus on making teams productive And to do this, we’re going to visit a company called Cloudy Coffee which is in the midst of launching a coffee bean with 100X the caffeine of a normal bean This is a bean that I desperately needed this morning and I think will be very popular in the market We’re going to start by looking at the new Hangouts chat, completely rebuilt to be an intelligent messaging app for teams Now, Cloudy and our G Suite customers already use Hangouts’ direct messages every day to keep work moving forward But in the new Hangouts chat, we’ve added rooms Rooms are a central place for team and project discussions To help look at this and help us walk through this, I’ve got Mandy at the controls >> MANDY: Fired up and ready to go So I’m going to bring up Cloudy Coffee Room in the new Hangouts chat >> SCOTT JOHNSTON: So the first thing you will notice in this room is that discussion, realtime discussion is threaded, and this allows me to separate the lunch conversation from the work and allows teams to dive deeply into discussion without fragmenting other discussions Mandy, why don’t we move a bit lower in this room and look at some other stuff? So here we see Nicole has posted a brochure for our launch That brochure is in Docs In the new Hangouts chat, Docs and Drive are deeply integrated

and the room manages the permissions for you, so anybody who is a member of the room now or in the future always has access to work with that file So now we have a central place to discuss project and team information We know that search is critical, and so we have built a powerful search directly into the Hangouts interface >> MANDY: So on top of free text search, you can also filter by people or types So you can always find the content in your rooms Let me take a look at slides Ah, and there’s the sales forecast that Patrick put in >> SCOTT JOHNSTON: Perfect Teams work with a myriad of tools today, and so we have created the Hangouts platform that lets third parties deeply integrate with Hangouts and Team rooms The platform supports a wide range of capabilities from lightweight scripting with our Google Apps script so you can automate team workflows quickly all the way through to intelligent bots So let’s look at an example Cloudy Coffee uses a sauna, a work tracking product, to stay on top of their launches and know who is working on what and when >> MANDY: Okay So right from inside the room, I can assign a task to Mike, who I like to assign all my tasks to And with one simple click, task is created >> SCOTT JOHNSTON: Great We’ve teamed up with a number of companies like Zendesk, ProsperWorks, Box, and more, to deeply integrate their products into the Hangouts platform And let me show you actually how we’re using the platform itself to integrate our own products We’ve built an intelligent bot we call Meet Meet uses natural language processing and machine learning to automatically schedule meetings for your team >> MANDY: I’ll ask Meet to find a time for us – for the people in the room today >> SCOTT JOHNSTON: Perfect >> MANDY: Oops, maybe I’m – I may have had too much of that coffee >> SCOTT JOHNSTON: Too much of the coffee? Yeah >> MANDY: Oops >> SCOTT JOHNSTON: So what’s happening here is Meet is going to go out, it’s going to look at all our calendars for members in the room, it’s going to find the optimal time for us to meet this afternoon, and automatically book it in Google Calendar >> MANDY: Actually, let’s move that meeting to tomorrow >> SCOTT JOHNSTON: Okay So here we are seeing with simple conversational commands we can do something that otherwise used to take many, many steps Wait, if you’re moving that meeting to tomorrow, that means all of you are going to have to stay until tomorrow for the demo. Is that okay? Are you guys good with that? I can lead a sing-along We have snacks. No? >> MANDY: Good point, Scott Let me just move that to right now >> SCOTT JOHNSTON: Okay So we’re going to ask Meet to schedule a team meeting right now And before we jump into this meeting, let me talk about meeting technology today I look around and I see us landing rockets on rafts in the ocean and it’s still so hard to get people into a meeting Our customers sometimes spend ten minutes getting a meeting ready Somebody doesn’t have an account, the meeting system sent them 72 phone numbers and you can’t find the code And so this is what we obsessed about when we rebuilt the Hangouts meeting experience What does it mean to obsess about it? It means no plug-ins required One click and you’re in We dramatically reduced the code size and optimized the experience so that you’re instantly in the meeting, you have less CPU fan, and your video and audio quality are dramatically improved All right, I’m done with my rant >> MANDY: Great I think we have the entire team already one >> SCOTT JOHNSTON: All right, great >> MANDY: So when you click in, you’re instantly taken to what we call the green room where you can check out and make sure you’re ready for the meeting and then join Hey, guys Welcome to the keynote >> TEAM: Hey >> SCOTT JOHNSTON: Here we are So what you’re seeing here is the new Hangouts Meet, our enterprise solution for video meetings So, sure, it loads fast All right, yeah So it loads fast, it performs well, and I’m sure you’ll get to try that when we launch it today But there’s also friction getting into meetings in other ways What about that consultant that doesn’t work for the company, doesn’t use G Suite? I don’t know why you wouldn’t use G Suite, but some people don’t I hear No problem Vroon is a consultant that is helping us with this coffee campaign, and with a single link he has jumped in >> MANDY: So I’ve just accepted Vroon’s knock and he’s instantly there Hi, Vroon It also looks like Mike has joined from the road >> SCOTT JOHNSTON: And that’s because Hangouts Meet with every

meeting now can contain a dial in code so that you can connect and participate in the meeting even when you don’t have a data connection So there is a lot more I could talk about with Hangouts Meet, but what I want to stress is that the focus was to really, really get you into the meeting as quickly as possible so the technology would get out of the way and you could focus on real work I don’t want to spend more time taking because we’re in a meeting So now that we’re actually in the meeting, why don’t we talk about the ad campaign for this new bean? >> FEMALE SPEAKER: Yeah, I have some ideas that I can white board but I don’t know if you guys are going to be able to see it >> JONATHAN ROCHELLE: Stop Please stop You said white board? >> SCOTT JOHNSTON: Yeah, is there a problem? >> JONATHAN ROCHELLE: And you’re trying to reduce friction in the meeting? That’s a great idea >> SCOTT JOHNSTON: What’s the issue? >> JONATHAN ROCHELLE: No, really I mean, it’s 2017 Nobody’s going to be able to see it And a five person meeting is bad enough We’ve got 5,000 people here; they’re not going to be able to see the white board Oh, wait Actually, we can just have you point the camera to it, right? Yeah, we’ll do that >> SCOTT JOHNSTON: Yeah >> JONATHAN ROCHELLE: And you can’t save it There’s no way you’re going to be able to save it >> SCOTT JOHNSTON: Can’t I write don’t erase? >> JONATHAN ROCHELLE: Yeah, how about that? Write do not erase on it That will work >> SCOTT JOHNSTON: Let me guess >> JONATHAN ROCHELLE: Or snap a picture >> SCOTT JOHNSTON: This thing they rolled out here, do you have a better solution? >> JONATHAN ROCHELLE: I think we might. Yeah Let’s talk about that So what we need is not to waste the time once we’re in the meeting, right? We got there really easily Let’s not waste that Let’s get work done That’s why we meet, to collaborate and to work But there’s never been a great tool for that The white board is close What we need is all the virtues of the white board It’s fast, it’s simple, and it’s freeform, but we want that with something where all the friction is removed That’s why we created Jamboard Jamboard is a white board in the Cloud, in the meeting room, and beyond your meeting rooms You just pick up the stylus and you start thinking, communicating, and working with your team So let’s see what that really means So T.J. is going to write some ideas that we’ve talked about for this new Cloudy Coffee ad campaign, but T.J won’t be working alone here >> FEMALE SPEAKER: Hey, guys Now we can’t see the Jamboard >> JONATHAN ROCHELLE: All right I think we’ve got a solution for that, though So wait for it The Jamboard knows that there is a meeting going on and automatically presents a prompt, and with one click T.J can join the meeting and present the Jamboard to the meeting So everybody remote, everybody on the meeting can see what he’s doing on the Jamboard Hey, Patrick Patrick, you guys can see the Jamboard now, right? >> PATRICK: Yeah >> FEMALE SPEAKER: Yep, thanks >> MALE SPEAKER: We’re all good >> JONATHAN ROCHELLE: And by the way, we have an opinion on the stylus that T.J. is using The Jamboard stylus is passive You don’t have to charge it You don’t have to pair it You don’t have to dock it And when you lose it, it won’t cost an arm and a leg because it’s passive It’s intuitive and simple Most importantly, the Jamboard still knows the difference between that stylus and your finger so you can write with the stylus and erase with your finger naturally and intuitively So, guys, why don’t you help us out here? Help T.J. out and take a minute and let’s brainstorm some activity So you know the ad campaign that they’re working on, sometimes you use those sticky notes if you’re in the same room So the team is going to use the Jamboard companion apps and anyone on the team can now add content and help T.J with his brainstorming So whether it’s from their lap or their phone or their tablet, from an Android or an iOS device, the companion apps let them participate and get work done to add their ideas, to make a point, to organize what is already there It’s the same magical collaboration experience hopefully you’ve gotten used to in Google Docs Now we’re all at the white board So Lucy in Liverpool or in the room over there, or Bereen in Brazil, or Tae in Taiwan, this is teamwork This is actual work happening Progress here and now while we’re in this meeting So let’s keep working We’re looking at where to target this ad campaign So let’s see another feature We’re going to look at a billboard location So how about New York City? Let’s go big I’ve seen a lot of billboards in San Francisco this week, actually But we’re going to go big and go in Times Square So the power of the Web, the power of Google is right there for T.J. or anybody using the companion apps, finding relevant, functional, beautiful information to add to the Jam and your Jam comes to life So T.J. is using Search and Maps on the Jamboard to add useful content for the team, and Bereen and Tae and Lucy and others are adding other content So they’re working on visuals What’s the branding? What is the branding feel we want for 100X caffeine coffee? I’m afraid to ask So they’re all using Search And Jamboard is also integrated with Google Drive So let’s try something else You see somebody actually added using the companion app a slides deck that we’re working on the financial model for Cloudy Coffee So spreadsheets, presentations, and documents, and anything from

Drive can be added to the Jamboard from the companion apps so the team can see it, work on it, and keep it. Excellent Now imagine we had more than one Jamboard Imagine we had more than one And we actually do We have actually somebody in New York on this meeting, one of our best graphic designers I think we’ll work on a custom logo for this Cloudy Coffee So Elon is working on a Jamboard out of our New York office You can see actually the avatars – T.J., if you could open up that frame organizer – you can see the avatars of where everybody is while they’re working on the different frames in the Jam And Elon is working on something and we’re seeing it change live here Everyone that’s on the companion apps is seeing the same thing Thanks, Elon Awesome work So now I think we’re pretty much ready to send this off to the ad agency So T.J.’s handwriting is not actually the best That’s probably the best I’ve ever seen, T.J But let’s make it a little better The power of the Web and Google and the machine learning capabilities of Google are available for T.J to make his handwriting really good too Excellent. And when you’re done, please don’t snap a picture of this. Okay? You don’t need to Everything from Jam is saved in Google Drive so you can pick up the progress next time That was just a brief overview of the new Hangouts chat, Hangouts Meet, and the new Jamboard, built with the enterprise power of security, intelligence, and scale that you would expect from G Suite and designed to bring teams together in several ways If you want to learn more, please participate in our breakout sessions or come see us in the sandbox Back to you, Prabhaker >> PRABHAKER RAGHAVAN: Awesome job, Jon and Scott Thank you for that The final announcement we are going to make today thinks of G Suite kind of in the same way you think of your teams They work really well together but they also need to work outside of the team We decided that it was time for Gmail to have add-on capability because people were wasting simply too much time going from their primary work surface, which could be Gmail, digressing over to a distraction, and then by the time they come back 20 minutes are gone Rather than my doing the talking and explaining this, I’m going to bring on stage a partner, Intuit, who will explain what they did with Gmail add-ons Please welcome to the stage EVP and CTO of Inuit, Tayloe Stansbury Welcome, Tayloe Take it away >> TAYLOE STANSBURY: Thank you, Prabhaker And good morning, everyone I am delighted to be here If I could just say a word or two about Intuit, our primary brands are TurboTax and QuickBooks And with QuickBooks, we have some 1.8 million online and mobile users around the world Now it turns out that about a half million of those users also use Gmail It seemed natural that with this add-on capability we would want to integrate the two so that those users could have a more smooth flow between those applications So let me give an example Here we have a customer whose name is Craig He’s a gardener He really loves making people’s gardens beautiful And he’s sitting in a cafe and he’s reading his Gmail on his phone and up comes a message from his customer Sarah who says she’s delighted with the work that he’s done on her garden and she’s wondering how she should pay him He’s thinking, wow, this is an opportune time to invoice her Now what it used to be is that he would have to jump out of Gmail, go into QuickBooks, log in, find the place to do an invoice But instead what we did with this add-on integration is that we have a QB icon that is at the bottom of his Gmail He simply clicks on that, it infers context, single signs him on to QuickBooks, drops him right into the place where he can fill out an invoice And as you can see, it is excerpted from the email the name and address of his customer Now all he has to do is fill in the rest of the details for his invoice and off he goes Remember, he’s still in Gmail He’s never left the application and he’s been able to fill out

the invoice entirely in there Boom. He sends the invoice and he’s done It was that easy Now when he’s done with that, he can actually quickly look at how his invoices are doing with his other customers in QuickBooks still inside the Gmail app, see how that’s going, and get right back into Gmail so he can finish his coffee and go off to make other people’s lawns beautiful So, to summarize, it was really easy to do this integration We wrote once and deploy many across Android, iOS, and the Web We can’t wait to get this to market It will be later this year And this is just one of many integrations that we expect to do between QuickBooks and the G Suite We already announced one earlier around the calendar and we expect to have more over time Thank you >> PRABHAKER RAGHAVEN: Outstanding Outstanding, Tayloe Thank you for being such a great partner in this All right I’m just going to finish with a wrap-up of the announcements we saw today So here are the announcements we went through today Team Drives, general availability; Vault for Drive, GA; AppBridge, Vancouver-based company, acquired for building connectors; Drive File Stream and early adopter program; and Quick Access for Team Drives across all platforms We saw the new Hangouts chat that Scott demonstrated That goes into early adopter program today Hangouts Meet in general availability That’s the video conferencing piece The Jamboard We will begin taking orders soon and we should be generally available in May at a price point just below $5,000 for the board with an annual subscription fee of $600 for the service And finally, you saw Tayloe showcase the Gmail add-ons, and we’ve been working with a bunch of other partners to build out exciting new add-ons That goes into development preview and we hope many of you will come and join us on that exciting journey With that, I want to thank you for all your attention Next, welcome my colleague Chet Kapoor >> CHET KAPOOR: Thanks, Prabhaker Hello, how is it going? >> AUDIENCE: Good >> CHET KAPOOR: That was a little weak I just wanted to let you know that we have extended the keynotes We are going to have another two hours, so you should just get really comfortable This is going to take a while, so you might as well enjoy the show My name is Chet Kapoor It is tough to follow Prabhaker It always is Until recently, I was the CEO of Apigee and have been now at Google for 3 1/2 months and it’s been great I want to do a special shout-out to a badass woman pioneer in the tech space, Diane Greene It is great to work with her and for her and the opportunity to continue to do it for quite some time to come Oh, she’s actually in the audience Sorry, I didn’t realize that It is awesome to see so many customers and partners here and so many others that we will soon have an opportunity to work with We thank you very much for giving us the opportunity to serve you, and we hope to do that for many, many, many years to come In addition to all the great innovation, all the great innovation you’ve heard, one of the conversations that we have with the G Suite for the companies that you work for and board members is about how the Cloud enables innovative business models, new ways of you creating and changing your business that you wouldn’t have been able to do before And we think a large part – a large part of creating innovative business models is actually working on ecosystems In fact, Gartner actually in 2017 CIO agenda talked about how some of the top performers of companies that they track actually participate in many digital ecosystems Interestingly enough, not all ecosystems are the same There are many different kinds of ecosystems And generally when we think about ecosystems, we say, you know, it must be public ecosystems But there are many different kinds and all of them affect

innovative business models So let’s spend a couple of minutes talking about them T-Mobile is a great example – great example of an internal ecosystem Cody Sanford, the CIO, has taken their core components like billing, like customer, like product, and assigned senior leaders to them These senior leaders are responsible for making sure that these core components are available as APIs with SLAs; by the way, only for the internal folks at T-Mobile In addition to making them available as APIs, they also have to modernize the stack underneath the APIs And so what this is helping T-Mobile do is deliver products and services at a pace that they have never done before We’re going to see so much more from this team soon ABnet is one of the largest distributors of electronic components and solutions Recently, they wanted to get into the Cloud solutions business They set up a Cloud ecosystem They now have 5,000 resellers – 5,000 resellers And their Cloud business is growing over 1,000% – at a pace of 1,000% year-over-year Industry ecosystems are actually well understood The stakeholders in healthcare IT have all come together These are the providers, these are the pairs, these are the device makers, they are all coming together to create FHIR, which is an interoperability standard to securely exchange patient information FHIR is picking up momentum I love the Ticketmaster story Ticketmaster actually has a billion dollar API internally and API that actually processes over a billion dollars on an annual basis The aspirations are to become the operating system for live entertainment worldwide They want to create a billion dollar business in a public ecosystem They want all of you to be able to use Ticketmaster functionality in the apps that you create; any apps, whether it’s in your car, whether it’s your mobile device, or any other kind of app that you create We think they are well on their way What’s common across all these ecosystems is they are all focused on one common goal That is to create innovative business models We think to create innovative business models you need three different parts There are three pillars to creating innovative business models You need to create and join ecosystems, you need to be able to connect to apps, to data, and to devices internally and externally, and be able to leverage all the great innovation happening around you So let’s spend a couple of minutes talking about each one of these The question that all of you have to ask and every G Suite has to ask, every board member has to ask, is your enterprise ecosystem ready? Because the right ecosystem platform can help your enterprise with keeping an inventory of products or solutions if that’s what you do, apply pricing models to these products, different kinds of pricing models, consumption-based, many others, be able to work through multiple tiers of partners because it’s going to be B to B to C in some cases, so multiple tiers of partners, and the most important thing, to make your product very easy to discover and easy to buy for end-user customers Orbitera is a multi-tier Cloud ecosystem that helps ISVs and enterprises take their Cloud-ready solutions and help them distribute and sell them We’ve had great momentum with Orbitera recently We are now over three billion transactions per month and the growth continues APIs are well understood and everybody knows that it’s the cornerstone of every digital transformation happening in any and every industry APIs come in different types There are APIs as a service This is how software talks to software And now with microservices, you’re going to have thousands of APIs in your enterprises And then there is APIs as interactions This is how your software talks to the physical world, whether it’s a mobile device, whether it’s Google Home, or whether it’s your car, all through one API APIs as products is just getting understood Technology companies actually understand it well Now enterprises are starting to get it as well It is about taking a business focus to your API

Thinking about your API as a business and bringing everything that you can think about, product management, revenue, channels, thinking about how the usage patterns are, thinking about having a road map for the API that you publish, thinking about all the different version and every engineering schedule that goes with it So there is a lot happening with APIs as a product with some of our enterprise customers as well Obviously an API platform needs to cater to the entire spectrum of APIs And in addition, it needs to be very secure, it needs to scale, it obviously needs to be multi-Cloud, it needs to work in your private Cloud as well, and it needs to be able to support an hybrid architecture At Apigee, we know a thing or two about APIs We process billions of API calls in our Cloud You process billions of API calls using our technology in your Cloud Thank you very much for the customers that have bet on us Hopefully we get a chance to serve more of you soon Leveraging innovation is something that start-ups get because they’re born with constraints They start every day and say what is our core value? What am I going to do that is going to be world class? Everything else, they don’t outsource but they go and partner for It’s something that happens really, really well in the tech industry And now it’s good to see enterprise companies do that as well Audi, Crate and Barrel, and many others are taking maps APIs and making it part of the customer experience that they are delivering every day It’s not just about the maps API There are thousands of APIs AccuWeather’s weather API, Twilio’s telephony API, and on and on that are available that you should try to leverage to accelerate your ecosystem As Google, we pioneered the product API space Google Maps has been around for over a decade It now has close to three million daily active users And you’ve probably heard in the last two days the number of APIs that we are going to bring to you all with one simple goal is to accelerate your journey as you think about more innovative business models So we take these three pillars and bring them together and call it the Connected Business Platform A Connected Business Platform is about creating and joining ecosystems, internal and external ecosystems, all kinds of ecosystems It’s about connecting to apps, data, and devices in your data centers, in your firewall, as well as externally And it’s about leveraging all this great innovation that is happening in the bazaar Digital is happening today You work for a company that has either been disrupted, is getting disrupted, or is going to get disrupted It’s going to happen with every industry Every analyst is writing about it, every strategy group is writing about it, and more importantly, you are living it We think the time to act and to go off and think about creative new business models is now As Google, we have 7 one billion user apps that we run every day We have created multiple ecosystems and have joined many, many more We want to take our experience and meet your requirements for today and partner with you for the future Please come by our showcase It’s right here outside the third – on the third floor itself Come and take a look at the great examples we have We have many enterprises that are building different kinds of ecosystems and we are happy to talk to you more about helping you with your journey Accenture has been a great partner We’ve participated in many Accenture tech vision reports The one in 2017 is phenomenal It talks a lot about ecosystems, AI, and many other things But I thought this quote was really interesting because it talks about how companies need to partner with their customers, a long-time pet peeve, the best product managers are your customers and the employees, and this requires a cultural shift So to discuss this, I would like to invite Gene Reznik to the stage Gene, take your time

They already know we’re going to take another two hours >> GENE REZNIK: Twenty minutes Twenty minutes >> CHET KAPOOR: That’s all right we said, right? Gene, thank you very much for joining us >> GENE REZNIK: Pleasure >> CHET KAPOOR: Would you introduce yourself? Tell us a little bit about how Accenture thinks about ecosystems And then most importantly, what are you seeing in the market? What are enterprises thinking about and doing with ecosystems? >> GENE REZNIK: Yeah. Yeah So I’m responsible for Accenture’s ecosystem and ventures What that really means to us – Chet, as you said – every industry, every enterprise customer is being disrupted or is doing a disruption What we really wanted to do is coming from the strength of Accenture is really to help our clients on that journey And due to our partnership with Apigee that we started 5-7 years ago, and now we’re very excited to continue it as part of the Google family And really what we see is this concept that you brought up, externalize innovation To compete, you need to work much more broadly And fundamentally, innovation is the operative word with many of our clients Now innovation is a culture, innovation is a set of business processes, and I think what our clients are really demanding for is for innovation to be also not hindered by integration I think this is where the technology and what technology enables really needs to reinforce, the velocity, the creativity, the business models that are really a lot of the Fortune 100, 1,000, really want to set up to take their business to the next level That’s really the excitement for us We’re really disrupting with them and helping them evolve and really empower their businesses in the new >> CHET KAPOOR: That’s awesome So the one thing that is different about the 2017 tech vision report than the others that we participated in and work with you on is that you come up with this concept called people first >> GENE REZNIK: Right >> CHET KAPOOR: As we talked to the G Suite and about their journeys, the cultural ramifications are quite significant Can you tell us a little bit more about this people first concept and how you’re thinking about it and how you’re talking to your clients about it? >> GENE REZNIK: Yeah I mean, over the past couple of years, I think we’ve all realized that it’s really about the people, the employees, the innovation, the culture, but fundamentally it’s probably the hardest thing to transform I think a lot of us work to really reinvent our teams A lot of us work, even us consultants, to reinvent and be relevant and ultimately embrace things like design thinking and storytelling and a lot of the things that we saw here earlier today And again, how did that really translate into you don’t want to come out of that and then go into the standard waterfall development lifecycle that takes six months to do the integration that you’re trying to do You want it realtime You want to drag and drop You really want to continue and extend the culture that you’re fostering through your design studios, through your innovation centers, all the way through the entire lifecycle and really create those experiences for your customers and the ways that you do your business So it’s really an important part of what we’re trying to get right >> CHET KAPOOR: Awesome Thank you very much, Gene I look forward to working with you more >> GENE REZNIK: Absolutely Thank you, Chet Thank you >> CHET KAPOOR: Next I would like to invite three of our customers These are customers that are actually already doing ecosystems, a company that has actually been doing ecosystems for quite some time They now are thinking about it differently And a couple others that are actually new to this and are changing the way their companies work So please welcome our customer panel >> LYNN LUCAS: Hello >> CHET KAPOOR: You’re ready to extend this for another two hours? >> LYNN LUCAS: I’m getting hungry >> CHET KAPOOR: All right So why don’t each of you start? And maybe I’ll start with you, Lynn Introduce yourself Veritas and Google had an announcement yesterday Tell us a little bit about that And then tell us a little bit about how Veritas thinks about ecosystems because you’re not new to this I mean, you’ve been a tech company for a long time How are you thinking about ecosystems differently now? >> LYNN LUCAS: Great So thank you very much Lynn Lucas I’m CMO and I lead marketing at Veritas We made a major announcement and a strategic partnership with Google yesterday It’s really all about what’s been talked about here, which is the importance of data and how it’s changing businesses How do you have visibility into your data? How are you moving the right data to the Google Cloud? How do you protect it there? And then for you G Suite users, how are you ensuring that you continue to have regulatory compliance to increasing regulation? Now, Chet, what you said is that Veritas has long understood the

importance of ecosystems We’ve invested heavily in it for years I mean, IT is built on the fact that you have to have ecosystems to make it easier for you This couldn’t be more important in the era of the Cloud When your data is spread amongst Google Cloud, but as Eric said, probably more than just that Many of you are probably having your CRM, your HR applications and other workloads in Clouds We’re building ecosystems with Google Cloud and with many others to ensure that it’s easier for you to manage and protect that data >> CHET KAPOOR: Awesome Thanks, Lynn Roger, same for you Give us an introduction And tell us, what is this Pitney Bowes, you know, been around for 100 years What are you guys doing with ecosystems? >> ROGER PILC: Yeah, absolutely So, Chief Innovation Officer of Pitney Bowes We are a global technology company that powers commerce We have been around for 100 years and we’ve been undergoing a very, very exciting transformation That transformation has been powered by what we call the Pitney Bowes Commerce Cloud Apigee As Chet knows, they’ve been an absolutely outstanding partner in helping us digitize everything we do for clients as we discovery, identify, locate, communicate, ship, and pay So the Commerce Cloud in Apigee and APIs has allowed us to power a very exciting commerce ecosystem with many millions of sellers, many millions of buyers, multiple e-commerce marketplaces, tens of sellers, financial services companies, and it’s really helped transform our company More recently, we also engaged with Google around the Android operating system for S&B appliance and our S&B ecosystem and recently chose Orbitera as well as the apps to our technology and our Pitney Bowes Commerce Cloud for S&B So it’s been a great journey >> CHET KAPOOR: Awesome Fatala, I was really moved by Fred, your CEO’s letter in the annual report where he talked about we’re going to stop being a retailer with a digital area We’re going to become a digital company with some physical space and a human touch >> ANDRE FATALA: Right >> CHET KAPOOR: You’ve been doing this for a while There were two people and a dog in a small room when you first started out Can you tell us – introduce yourself and tell us a little bit about your journey as you’ve expanded Magazine Louiza’s perspective on how to think about ecosystems >> ANDRE FATALA: Okay, right So my name is Andre Fatala I am from Brazil So I am glad to be here talking about technology and not carnival We started our journey in 2012 We were a really small R&D team that tried cracking the systems of the corporate enterprises company After two or three years, we just get all the points of sales development inside the company using a lot of lean methodologies And we moved everything to the Cloud since then We are starting for the APIs So when we get together and talk about how to leverage the innovation and then decouple the big legacy systems Since then, we got this idea of our CEO to try to be a digital platform with physical stores Since then, we just turned out entire company to do this thing We are seeing really good results Last year our digital just goes up 35% I don’t know if you guys know, but in Brazil we are facing a really bad economic crisis >> CHET KAPOOR: For sure That was great Fatala, one follow-up question for you I’m sure this was not easy, right? I mean, yeah, you’ve started out in a small office and a dog and everything and you’ve transformed quite a bit What was the hardest part of the journey? >> ANDRE FATALA: I think the hardest part of the journey was to get everybody understanding this changing of the mindset of the company After that, we have the challenge to change the entire technology to provide the flexibility and agility to to the business But I think that we are doing a great job on that >> CHET KAPOOR: And so – go ahead >> LYNN LUCAS: If I could add into that, I think that the transformations – and Veritas is undergoing that transformation as well – the hardest part is the people and the culture It seems like that’s what you’ve experienced as well The ability to move forward with the people and culture makes all the technology come to life >> ANDRE FATALA: That’s right Technology is there

It can be used for everything We need to get the right people to do this >> CHET KAPOOR: So, Roger, how do you – I will ask a very direct question – how do you deal with antibodies? Right? I mean, it’s a cultural shift It’s understood in the G Suite, the board, everybody, we’re going to make the transformation, but not everybody gets it How do you deal with those cultural aspects of this transformation? >> ROGER PILC: Yeah It’s similar to the other panelists For us, culture has been the biggest positive in our transformation We’ve had a company that for 100 years has valued our employees, has had great employee engagement, has had a culture of innovation, and most importantly collaboration For us, we were able to harness that That was a great surprise being there are only three companies So the key to us has been have a very clear vision of where we’re going, be very consistent with it, have a focus on operational execution and the discipline of getting things done, and just maintain a belief There are pockets of concerns where maybe there was past failings, so maintaining that vision and that constant belief and sharing that sense of belief has been critical >> CHET KAPOOR: And the sense that it is okay to move the goalpost a bit >> ROGER PILC: Right Absolutely We move them very far >> CHET KAPOOR: Final question for all three of you We’ll start with you, Fatala If you were in the audience a year ago, what advice would you have liked to have given yourself? >> ANDRE FATALA: Try to focus on our customers, be creative to solve their problems and not ours, and I think when you decide to do something, execute like crazy You need to try a lot Get data, get some insight from this data, and then try to evolve our product >> CHET KAPOOR: Awesome How about you, Roger? >> ROGER PILC: The key is to believe that great things and great transformation is possible With the right vision and recognizing the availability of great technologies today, picking the absolute right technology partners, we couldn’t misstep on that So choosing the right technology partners and executing against a vision has been critical >> CHET KAPOOR: I love that You’re starting with believing >> ROGER PILC: Right >> CHET KAPOOR: And Lynn, how about you? >> LYNN LUCAS: If it’s your digital transformation, you absolutely have to focus on people and culture first When it comes to data, which is the heart of how we build our businesses, get visibility into what you have, 30% of it is junk or rot, and then move it quickly to the Google Cloud, rethink your storage, reduce waste, make a more sustainable environment >> CHET KAPOOR: Awesome Thank you very much I would like to now invite Urs back on stage >> URS HOLZLE: All right So you see a lot of innovation at this conference behind me You see actually just a subset of what we announced today because we don’t really – we can’t really fit anything on one slide We showed how Google Cloud has the best infrastructure, the best security, the best productivity suite, and we showed a host of new collaboration features, including the awesome Jamboard So I’m really looking forward to that And then, of course, there is a lot of help to get you started where you are, and we also showed you yesterday how we’re building our ecosystem with global partnerships with companies like Pivotal, SAP, Rackspace And we understand it’s not just about the technology, it’s about helping everyone be successful in using it So that’s what this conference is all about So thank you very much for coming, and thank you for the trust that you’ve placed in us Have a great day