 Live from the Mission Bay Conference Center in San Francisco, California, it's The Cube at Google Cloud Platform Live. Here are your hosts, John Furrier and Jeff Frick. Hey, welcome back everyone. We are here live in San Francisco for Google's Cloud Platform Live. This is The Cube, our flagship program. We go out to the events and extract the signal from the noise. I'm John Furrier, the founder of Silicon Amgen. I'm John Frick, my co-host. Our next guest is Lee Chen, head of product at Fastly. Just tweeted your news on our crowd chat app. Welcome to The Cube. Thanks very much. Thanks for having me. So, I got to say, I'm very impressed with Google's announcements today. Obviously, very developer-focused, you know, fresh, clean sheet to go after Amazon, win some share, developer goodness, but a lot of nuances on the back end. Very large scale. Interconnect, a lot of the peering stuff kind of comes out of the woodwork. You can almost connect the dots and say, you know, really win the connectivity back end and then connect on the front end. You squeeze the middle. That could be a great opportunity. You're playing on that important side infrastructure. Talk about Fastly and what you guys announced. I just tweeted your cloud accelerator product. Give us a quick chat. Take on what you guys announced. Yeah. So, the snapshot is that we're the first partner on the Google Cloud Internet Platform, right? So, we have a direct connection from our origin shield pops, which are in San Jose and Ashmo, Virginia, directly to Google's backbone. So, that allows us to really accelerate what we call Fastly Cloud Accelerator, any content that's hosted on Google Cloud Storage, Google Compute Engine, or App Engine, right? So, the tests that we have so far, and the customers that are using it so far, are showing up to four times the speed increase in terms of performance and delivery of large-objective video. So, that goes- And you just pop in Virginia and send us a score at San Jose? San Jose, okay. So, okay, on the backbone. So, you have to be right on the naps and then you connect into their backbone and both spots are out here? In both spots. Okay. So, they get the national backbone? Correct. Yep. How big is their backbone in terms of, like, speed, kind of- Goobles? Yeah. I mean- I don't know that anybody knows that. I'm not sure, right? I wish I knew for sure. Right? We're trying to figure it out. Yeah, I know. I think everybody- They've been investing, honestly, net neutrality is something that they're watching on a whole other front, but that, you know, what this interconnect brings up when we come in the intro is this Netflix problem that most people are familiar with when they're watching a great movie and all of a sudden, yes, I was going to say something, but it crashes on you. Shut something to bed. It crashes to bed. That's really about being throttled down. You've got feedback and inspection. You've got a middleman with service providers. This is a challenge for streaming and this new frontier and everyone's base timing and with mobile. Tell us what the challenges are and how does this new interconnect? Is this more the norm? Are you going to see more of this from other vendors or is this a unique situation to Google? It's a great question. I think the question for a lot of content providers is taking advantage of Google's interconnect program in order to actually be able to deliver content in a really quick and seamless, meaningful way off of their origin structure. Whatever their origin infrastructure is, their web host, their application stack, their API input, whatever it might be. I think that's really where the Google interconnect program is targeted. For us as a content delivery network, for Fastly, it's very different. We're really aimed at delivering this, taking advantage of the interconnect program in order to be able to deliver our customers and join customers with Google, their content faster all the way to the edge, whether that's the last mile and the cell tower or into a content end user last mile network. So what's the reality for the cloud guys? So now I'm a developer, the challenge is okay, I get all this infrastructure developers don't get along and DevOps shows that infrastructure as code. So if you're going to go to the next step and say infrastructure as code, you're going to have a programmable infrastructure. So Google is very API centric. So how are they connecting all that? Can you share some of your take on that, observations, what they're doing right, good, what needs to happen? And the API economy is clearly happening. It definitely is. I think every app that's built today needs to be that API built first and then an interface that's built on top of that. And that's definitely how we approach it when we started building CDN. We started from a DevOps perspective of building first and foremost an API to control your content delivery network the way that you want to as a developer. So a lot of our argument when we go into pitches is talking about, hey listen, you have an app or you have a site. How much of your application logic do you actually push to the edge and how much of the application logic are you actually caching right at the edge with the CDN? And you've got a shirt on the back that says you can't read it. Cache is caching, caches and caching your data. Is caching more than norms? Some people who look at caching as, hmm, I don't really want caches, you know, potential security risks. It's the latest and greatest cache coherency. These are all the known problems people kind of kick around. With virtualization, this seems to be an opportunity. What's going on with the whole caching argument? So caching is something that's been around for a very long time, right? And in the traditional caching world, you had to wait for a couple of different problems to get resolved. One of those is caching validation, getting rid of your assets, right? So Fastly allows you to instantly purge. We call this a purge. Industry typically refers to as caching validation. You might wait five, 10, 15 minutes, sometimes as long as hours for that to happen on another content delivery network. On Fastly, it's 150 milliseconds or less for that to happen globally, right? Another one is transparency and being able to see what's actually happening on your network with your data, whether that's your API or your web content, whatever it might be. So the way that we end up solving that problem is we stream blogs back to you directly in real time. There might be a couple seconds to date, but that's versus looking at a batch log file 24 hours later to see what's happening with your traffic and the actual characteristics of everything that's happening across your technology stack, right? So, and then the third problem is giving you the ability to make changes to your configuration at the edge and really push the application logic and make those changes in real time, right? So configuration changes on us take about two to 15 seconds depending on where you insert that change in the network, which is tremendously faster than I think a lot of how your app stack might perform in native by itself. One of the things we're hearing in the Google conference is, and there are no stranger to open source, obviously, you know, the following these guys that they built on, they donate a lot, Kubernetes is also open source. You guys are very open source driven company. Absolutely. How does that give you an edge? I mean, obviously, people think of Akamai and Limelight's had some issues, struggles and whatever you want to call it. Are they still around? I mean, are they still in business? They are. They are still kind of humping around. So, now the world is changing, right? So the game is still the same, but a new world. You got to pack it fast and now mobile devices. So, certainly a disruptive force to have right now. Are you guys feeling good that open source is going to be a lever for you guys? Is that a part of the competitive strategy? Absolutely. Yeah, so we're actually based on an open source project called Varnish, which is a reverse process. So we've taken that and added our secret sauce to it and done a ton of development work on top of it. But we continue to contribute back to that. And we end up getting really great ideas about how to handle specific problems, right? Steve Souters, who is formerly of Google, is now our chief performance officer. He spends a ton of time working on open source projects and contributing back to the community as well. So, we are very heavily invested with those tools. Is he part of the deal you guys had in there, a little Google alumni? Google, well, it never hurts to have Steve on your team, right? Give some take, yeah, it's always good. And the only insight, and by the way, Google is such a large scale between YouTube and Google, I mean, massive cloud. Absolutely. So, certainly they have a lot of chops there. Talk about, give the folks out there who don't understand the size and magnitude of Google just how freaking massive it is that they bring to the table in terms of current operational scale that they operate in, and then how that would translate to potentially cloud as they spoon feed the market in a way, because they have to kind of give out the baby food to the people trying to be scaled. Well, so I think that's something that Google's really, really good at. It's actually spoon feeding the developer, right? So, they're great about giving you iterative steps in which to progress. And if you look at the applications that they built and the products that they're releasing, there's a very clear path from A to B to C. And they've got a massive network and a massive experience-based drawing, and that is, like, you can't really put a value on that. You can't really put a product price line on that or a product roadmap on that. You just kind of have to watch it happen and see it evolve as it delivers, too. So, I think Google's got some, they're huge. Like, their network inside, you know, there are rumors about how large it is, but. Has the service been good? I mean, Amazon gets complained about being lumpy, obviously, with SLAs and that's something that they're just trying to shed that reputation. You guys are a little bit different. You need to have latency issues. I mean, you're running a CDN. You can't really have lumpy service. So, we try and make things as fast as we possibly can. One of the things that we've seen so far is that for customers that are using Google Cloud Storage versus, say, Amazon or some other cloud storage providers, Google is actually about four times faster when paired with fastest cloud accelerator, right? So, that's a tremendous performance increase and a tremendous benefit to the end user, right? So, when you guys are playing your VODs, you're probably looking at a two-second video start time, right, that's what you're aiming for. That doesn't really happen a lot of times in the video space because you've got all this other stuff that's happening to the player, right? That's a lot harder when it takes about a second and a half to 1.8 seconds for the actual content to be served back to the CDN for us to send it on to your player. So, I gotta ask you as a product, you have to think about the future, right? Ship today, plan for tomorrow, and it's a tough job. I mean, you don't have to look as far as down the street here at Twitter where they're going through product heads like you read about. Now they have a decent one with Kevin, big data guy at Stanford, I think we'll do terrific. But you're in a dynamic environment, so how do you look at the mobile infrastructure? Because this is the new web now, right? Connected devices, Internet of Things, certainly changes the game in terms of notion of routing packets and or device connectivity. Really, it's complicated. So what's your vision and how are you thinking about that at Fastly? So, there's a couple of challenges with mobile, right? First of all, once you get to the base of the antenna, you have 75 milliseconds no matter what before you get to the device. So, there's only so much optimization you can do before you get to that, right? That's just from the transceiver to the tower. That's just from the tower to the transceiver, right? To your phone, right? So, once you get past that, I think you still have a ton of work that you can do in terms of optimizing and getting into the network, right? Because what you have to remember is that the cell networks are operated by a provider who has a network who does think it's too bad it's a transceiver. A middle man. Yeah, oddly enough. It's cool we're going to have a cell tower soon. I mean Google Cloud, you know, LTE network. I mean, you could almost go there. Listen, they've got gigabit network. Next thing you know. They've got the back halls, easy to pop up some towers. I will see how it goes. Continue, I want to get this as good. So, I think there's a lot of optimizations that you can do at a very low level on the protocol stack for networking. And when we look at iterating performance as an objective of ours, I mean, fastest in the network, right, so we've got to be pretty quick about how we're going to run content. We do spend a ton of time thinking about how can we make mobile networks perform better in content that's being delivered to mobile networks. Do you see any kind of download code, I mean, I hate to bring up the whole flash download kind of made PCs better in the old, in the old archaic days of five years ago. That I get what you're talking about is stating both of us by the way. For all you youngsters out there who don't remember the dot com bubble, this is what we had to deal with. Totally, totally. I mean, literally 10 years ago, video was very challenging. Yeah, absolutely. Now today, so, but you know, the iPhone is seeing more and more devices. I mean, even the iPhone 6 bigger, the battery life's better, so you see in smaller, faster, cheaper in terms of at the edge. Yep. Is there promise there, any kind of innovation? So every time that we think it's gonna get better, it gets worse, right? Because listen, the problem was delivering 720p video, or 480p video, like five years ago, right? Then it was 720, now it's 1080, and now that we think we've got 1080 tackle, we got 4K on the horizon. And then somebody's talking about 8K, like, what's going on? Like that's maybe people's point. Can anybody see that? It keeps you employed. And listen, it keeps me in a job, and it also makes for some really challenging things for us to tackle from a network perspective and how we build things up. And then now the emphasis on web scales, the buzzword of the day. I mean, this is like the San Francisco, Silicon Valley area. I mean, they invented web scales. So you've got the Yahoo Google, these guys are. Web scale natives. How is that translating the mainstream? And obviously the enterprise in the cloud, people are like, oh, the enterprise cloud. Still, I mean, people in the middle of the country are like, ah, I'm happy with my T1s, or I got email moving around. There's a tsunami coming. What's going on in the mainstream? What do you see there? So I think one of the great things that, I can't remember if it was Brian Stevens or one of the other keynote speakers said, was that cloud is on the verge of making things change, and he had this great point about things become, when something truly disruptive comes along and you sort of take it for granted before it actually becomes disruptive, right? Cloud is one of the several things that are online where you're moving the application logic which you traditionally used to do on one server with one database server and a T1 connection or something. Moving that closer and closer to the edge and closer and closer to the end user is really what that's about, right? So when you can distribute that application infrastructure all over the place, that becomes a really interesting problem. You've got so much horsepower now, right? So you don't, you can add intelligence where you have to dump cash before all over the place. So, yep, and we actually try and build that into our cash servers in the way that we design our network. So if you go to our website, you can see on the networking page just the server stats that we have, and it's all SSD, there's times of RAM in it, there's times of CPU, and we let you actually execute configuration code right on the cash engines. So that allows you to change the way that you handle your content. So talk about your, talk about the service. Talk about Fastly, what do you guys offer? You guys charge us, streaming, we're getting, you know, a news stream here. Obviously we're doing a three camera shoot. That's the trend. More people are going to be going on doing live events. You know, the backbone, ultimately, or the cell tower will be the criteria, kind of the pacing item if you will, the bottleneck. But what are you guys doing for business, are you guys targeting with some of your customers? Share with the folks out there about the company. I appreciate that. So we're content to live in there, right? So what we try and do is move bits and bytes across the internet, faster and faster. So whether you've got an API, you've got a live stream, you've got a BOD content, whatever it might be that you're trying to move, we try to move that across the internet faster. So Twitter's a customer base, right? New Relic, GitHub, Firebase is a new Google acquisition as a customer of ours. We've got a ton of really interesting technology stories and a bunch of big media companies as well. So, heavy media, basically. The Guardian, UK newspaper, United Kingdom as a customer of ours and we've been working with them really closely. You charge for the video streaming or more? So we charge on a pretty standard model for CDNs, which is you charge per gigabyte served and by 10,000 requests. You would be a big customer. You might be. We might be. I guess we're eight hours a day. Live streaming is a big, big deal. Well, so you stream over provision, so how do they make money? We always scratch our heads and you stream in Twitch, I mean, they're broadcasting a lot of... They're pushing a lot of bits and bytes across the network. They really are. So they're building their own CDN or how do you see that? So a lot of the live stream providers out there leverage multiple CDNs because the reality is, particularly for live video delivery, if you get a million people tuning in, I think you stream have close to eight million peak concurrent viewers or something like that during the PS4 announcement last year, that there's no one CDN or no one network on the planet that can handle that well by themselves. So it needs to be distributed across. I think folks like you stream and Twitch and others who are in the live stream space are building really intelligent ways of switching between multiple CDNs. You guys scratch your head and apply the big data analytics to some of the, everyone wants metrics, right? Hey, how do we do, was SLA, was it up? Did one deliver the stream and also kind of what happened? What's the metrics update? So we actually give you the ability, our customers the ability to pick and choose what metrics they want for themselves, right? So we can do live stream log delivery. So we will stream log delivery of every single HTTP request that we see, every single request that we see on our network, right back to a big data instance or drop it onto a storage bucket somewhere for a customer to analyze any way that they want. So anything that we can see on our network, the customer can see as well. So that gives them the ability to get really granular about what the traffic is doing. And I was going to say too, can you give the audience just a feel for really what's happening with mobile? I mean, you've been involved in gaming and obviously gamers are kind of out in front of the platforms using a lot of CPUs and they're younger kids a lot of times and you know, they're out on the edge a little bit and really kind of the teotonic shift in the consumption of media on these little things instead of these big things. So I'm a little bit biased when it comes to gaming. I think gaming has been pushing the edge of technology for a very long time, right? NVIDIA made better and better and faster and faster video cards because gamers needed it, not because we needed it for a better Excel spreadsheet. So I think mobile networks are no different. I think if you look at the app store, 90% of the apps that I've seen there are new games. A lot of them are free and most of them are really bad but somehow they end up sucking up eight hours a day when you're bored and you're screwing around. Mobile networks are hugely important to the way that we interact with the world today and they're a critical background. So for gaming, whether that's gaming, whether you're getting your news feed, you're checking Twitter, you're watching a live stream and you're watching this show, whatever it is, mobile networks are really key important. So figuring out how to get data effectively across a mobile network is a pretty key challenge for anybody that's building a new laptop. It is the gaming piece of it, it's still growing at a faster clip than more of the traditional media. If you look at all the reports, that's what it says. I definitely believe it. I think there's more engagement and there's more interactivity and there's more fun to be had in gaming than in any other kind of business center that you might see. It's entertainment, right? Right, right. But I guess the counter to the statement would be I want more horsepower than I can get on my little guy. I'd rather have an i7 with all kinds of RAM in this and that. I mean, how's that kind of get shake out? So I think there's a, you got a segment when you look at video gaming, you got to think about whether you're talking about a casual audience, sort of a mid-tier audience and then the hardcore audience. The hardcore audience wants a keyboard, a mouse, a headset, they want a fully immersive experience. They're probably playing 18 to 20 hours a day with a WorldCraft, like it's hardcore, right? They're really super engaged in whatever it is that they're building. I used to take a train two and a half hours a day each way going in and out of New York City. I played my iPad, I was playing the Jewel. I mean, there was all kinds of games that I was just screwing around with on my phone. That's a very different kind of engagement, right? So it's a different kind of activity. It just depends on what you're doing. You might sit down for two and a half hours and watch a movie. You might just sit down for 15 minutes and play a game for the company. Lee, thanks for coming on theCUBE. Fascinating conversation. Kind of gotten the weeds on the packets, but this is really important. Obviously the backhaul of the data, the content or the application drives everything. The last mile, as they call it, is always to the end device or network or computer. Really important. So thanks for sharing. Fascinating conversation. You guys are very compatible with the API economy, which is fantastic. I think it's going to do well for you. This is theCUBE, but we are broadcasting packets live from San Francisco Google Cloud Platform Live. This is John Furrier with Jeff Brick. We'll be right back after this short break.