Keynote: Fastly and the future of the edge | Altitude NYC 2019

Artur Bergman (00:00):

For the people who have been here before. Welcome back. People who've never been here. Welcome. It's been a couple of years now that we've been doing Altitude and we really enjoyed the one in New York to the point that we're not doing one in San Francisco anymore, so it's because you all are great. This has been a really exciting year for us. When we started many years ago. It's not that many, actually it's eight. One of the ways that I tricked Tyler into joining, was talking to him about doing more Compute@Edge and building an environment that helps people build globally distributed applications.

Artur Bergman (00:57):

Eight years later, last week, we announced Compute@Edge, which really is the first part of that. Turns out that to get there we had to build a network that had a lot of capacity. So we have to build the control plane, logs, stats, all these things that you eventually need. We needed a lot of customers so that we could figure out what you all actually want and then the least we needed to build a large team that could help deliver this.

Artur Bergman (01:38):

And you'll hear from some of the people who've been involved in this effort, which really by now is a very large part of the company in getting the Compute@Edge out to you. And personally, I think it's really cool. It really started in earnest about three years ago when a couple of us went to Pittsburgh and locked ourselves in a hotel room for five days and kicked off what we wanted, which is a builder to use many different languages and ability for our customers to upload code.

Artur Bergman (02:18):

And I think this is a very key requirement. We wanted to run every single request in its own sandbox, which made this problem much more interesting or harder. Another big change is that we went public, so that was on May 16th which was an exciting event for us here in New York, but most importantly it really helped us reach a point where we can now keep on investing with a long horizon knowing that we have the capital to deliver on R and D and all support and also the network itself and continue building that.

Artur Bergman (03:10):

So, it really is a little bit like having someone close to you graduate and now being not completely dependent on us but actually being something larger than us, which feels great. We have a lot of new features that have been developed over the last year. We have reached new levels of traffic. I think yesterday we were somewhere around while I was looking at our front page, about 10 million requests per second and somewhere around 12 terabytes per second, which every year that number just grows dramatically.

Artur Bergman (03:55):

We have added larger servers, more storage, more networking. You will hear some of that cool stuff what we're doing on that side, on the hardware side and network side of actually delivering your websites, your applications, later. It's been extremely fun this year. We've had so many things happening. We've had new countries, we have had IPL, we've had product launches, we have the Compute@Edge and it's really exciting to look forward to what we're going to be doing next and personally I'm really excited as well in seeing what everyone in this room will figure out what they can do with Compute@Edge. It's always fun going to a customer and they're doing something with Fastly that I hadn't considered you could, and this just opens that possibility up to your creativity.

Artur Bergman (04:57):

So I want to introduce Kimmie and Rajib — Kimmie is the VP of Product Growth, Rajib is VP of Product Management — to speak more about all the things we are launching or have launched. Thanks, everyone.

Kimmie Nguyen (05:11):

Hello everyone. Artur introduced me but I'm Kimmie Nguyen, Vice President of Product Growth.

Rajib Rashid (05:20):

I'm Rajib Rashid, Vice President of Product Management. And we're very excited to be here on stage with you today. It's been a tremendous year for Fastly and we look forward to sharing some of those highlights with you today.

Kimmie Nguyen (05:33):

Our customers are at the heart of everything we do. Many of you have been with us since the very beginning. Some of you are here for the first time. Actually a lot of you. Welcome. Many of our customers have collaborated, iterated and helped us build really amazing products. We are a platform built by developers for developers. We have a philosophy of being customer first and problem-focused. So we're going to start this presentation sort of setting the stage around what we see around us and then we'll sprinkle in all the great things we did this year.

Rajib Rashid (06:09):

So as you know, the world has changed a lot since cloud computing first emerged. The rate of change has been very rapid. Consumers are different, their expectations are different and the way that you interact with them is different. Brands like Airbnb, New York Times, Warby Parker.

Kimmie Nguyen (06:27):

Great glasses.

Rajib Rashid (06:30):

That's right. Each one has helped change the way people buy, how they consume content and where they spend their time. Across industries, there are teams of digital transformation fueled by rapid growth of mobile apps and mobile devices. Companies need to be innovative in order to keep up.

Kimmie Nguyen (06:53):

Customer centricity is a thread that runs through all of us. Customer centricity, customer experience, the ability to orchestrate and personalize an entire end-to-end experience with your brand, moment to moment, at scale on any channel in real-time. I can't really snap, but if I could, I would snap for emphasis. Gartner says 2020 is a year in which customer experience will overtake price and product as the key differentiator. We think that's huge and you need to deliver that experience quickly, really fast, faster than this. And if your site fails to deliver, customers simply switch because there's no shortage of alternatives.

Rajib Rashid (07:38):

So that's where Fastly comes in. We are on a mission to squeeze every bit of performance out of the network. Now, whether that's by automatically re-establishing connections to your end user’s browser or automatically finding the best route to an origin when there's multiple providers available. We have created strategic partnerships with major cloud providers, both at product as well as service level, and we're facilitating free-flow of data between those networks.

Kimmie Nguyen (08:12):

Building for the needs of tomorrow's internet and the modern web, we've invested heavily into QUIC and it's coming soon to a transport protocol near you. As Jana Iyengar, our distinguished engineer, blogged about QUIC earlier this year. It's designed to improve performance by mitigating last mile issues of loss and latency. QUIC and HTTP/3 will provide you with a better digital experience, particularly for mobile users and users and regions where internet services body. We believe In QUIC's value. Jana is actually talking later today, so you should check it out and we're looking forward to testing and learning with all of you.

Kimmie Nguyen (08:54):

Many of you in this audience sit in media entertainment. You face a similar content challenge to consistently deliver value, to deliver fresh new content, to deliver delightful new features. The lines of publishing, streaming, and content are blurring. Your customers want to come back for more and ideally, you want them subscribing for more.

Rajib Rashid (09:18):

So over 420 billion. That's the number of images that are being transformed and delivered by Fastly every month. To help you do more, we've added APIs that will automatically improve the quality of images. You can now use advanced logic to automate and detect faces and prioritize those regions when you're cropping those images. Media Shield, something that a lot of you here have been using has reached general availability. That's an important element of our media delivery pipeline. Couple that with new features like streaming mesh with TLS and segmented caching which had been rolled out into our network. Those are designed to let you do more by removing limitations and improving efficiency.

Rajib Rashid (10:10):

So media consumption patterns are of course evolving every day. New technology is being developed to keep up with the demand and growing needs, so we are committed to making sure that we are providing the right features and products to our DevOps community. For example, the ability to build dynamic manifests. You will actually get an example, a demo of that today. I don't want to give away too much already. Soon you'll be able to convert animated GIFs to video again to improve content delivery performance and we want to continue to provide you with more insights about, for example, your origin health stats like cache myths so you can actually make better decisions.

Kimmie Nguyen (10:55):

Your success is measured by content experience but also speed and agility. Winning companies will be those who listen, react, and deliver in real-time. Agile companies are three times more likely to take something from ideation to implementation and two times more likely to take risks on customer experience. Some companies deploy over 50 times a day, pushing out new features, implementing new ideas, fixing bugs. So the question should be how are you adding value?

Rajib Rashid (11:29):

So in order to empower developers of any level of skill, we've launched a new developer portal that you can take full advantage of our platform. There are over 60 code-based solutions that you can quickly take and that allows you to do quick iteration and testing at the edge so you don't have to start from scratch. Continuous integration, continuous deployment is key to agility and speed.

Kimmie Nguyen (11:58):

We need to have a separate slide for this. This is my favorite side because we love Terraform by HashiCorp and so do you. We launched a slew of improvements including support for virtualized resources, new storage and logging endpoints, batch API support. The list goes on. Be sure to check out Davy's talk this afternoon. She can get into all the details for you. Terraform lovers rejoice and if you're not a Terraform lover it's okay. We still love you. We're not done yet. Next year we'll be adding support for Terraform WAF.

Rajib Rashid (12:33):

Speaking of WAF, we know customers want fast, reliable, and secure digital experiences. Security and trust are two key elements for any online business today. We feel that we are in a privileged position to protect our customers from all online threats. We believe in a secure and trustworthy internet by default. This year, we launched a fully self-service TLS management system, Fastly TLS and Concierge TLS. These products give you complete visibility and control with UI and API for managing your certificates and keys, and we've included managed integration with Let's Encrypt. The newest version of TLS protocol 1.3 is being rolled out that provides faster connection establishment TLS by default, security by design.

Kimmie Nguyen (13:30):

That's a good slogan. So our security position naturally supports the movement to secure DevOps processes. You need programmability, performance at scale, and real-time visibility. We are committed to expanding support for logging endpoints because it helps you clever customers do really amazing things. It also gives you a ton of visibility. This year, we added an additional six logging endpoints into general availability. That brings a total to 23 and, for people who enjoy experimenting with big data like Rajib or Sean, we are particularly excited by Apache Kafka and Elasticsearch.

Rajib Rashid (14:16):

So many of you are running global companies. Your users are coming online from pretty much every part of the world. Online transactions are growing at an exponential rate. Again, by the growth of mobile devices, and Fastly is also growing our reach in order to meet your demands. But remember, our POPs are designed to be more efficient so we don't have to be everywhere. We strategically add POPs at well-connected internet exchange points around the world. We deploy POPs in key regions to help you deliver a better experience no matter where your users are.

Kimmie Nguyen (15:00):

So you only had 15 minutes and we covered just the highlights. We know how everyone loves to see all the things. I see you taking a photo. That's great. If you missed any of these announcements, they are in our blogs, in our release notes. We also have a ton of Fastlyans, as Hooman says, wearing black because it's New York. So go find them. Go find your account teams. Rajib and I are available. We'd love to talk to you. We've done so much this year that it's not the end. It's been a really exciting journey and we're going to take you through some moments of how we got here.

Rajib Rashid (15:36):

So here's our timeline. In the beginning, there was the big bang, although we didn't really have anything to do with that.

Kimmie Nguyen (15:42):

Oh, we needed it though. In 2011, Fastly was born to deliver fast, secure online experiences.

Rajib Rashid (15:49):

Streaming media services were launched in 2013.

Kimmie Nguyen (15:52):

Our commitment to rea ltime visibility expanded with additional logging providers in 2014. 2014 was also the year that we opened our London office.

Rajib Rashid (16:02):

In 2015, we released Soft Purge. That's a feature loved by most of you.

Kimmie Nguyen (16:11):

2016 was a year of our new user interface. New UI's, always a fun migration. That year, we served a record five million requests per second as the world went online for the elections

Rajib Rashid (16:22):

In 2017 we extended our edge security capabilities with our WAF offering.

Kimmie Nguyen (16:29):

Last year we reached new bandwidth and request milestones with a major unnamed sporting event that was very big. We also launched platform TLS at last Altitude.

Rajib Rashid (16:42):

And this year was huge. Artur already kind of mentioned it, but we became a publicly-traded company in May. It was an incredible experience for everybody. Kimmie cried.

Kimmie Nguyen (16:55):

I cry, but I cry a lot so don't worry.

Rajib Rashid (16:55):

Which brings us to today and our next part of our keynote. Last week, we made a big announcement about Compute@Edge, our new serverless computing platform. We have been building towards this moment since the very beginning. To tell us more about it, please welcome our CTO and Co-Founder, Tyler McMullen.

Tyler McMullen (17:17):

High five. All right, high fives. Great. Hi everybody. So if you've been to an Altitude before, you've undoubtedly seen me on stage talking about edge computing in one form or another. So today I'm very excited that I get to stop just talking about theories and algorithms and prototypes and the weird ideas that we have built out over the years. So instead we get to show you something concrete and real, but I'm not going to do that. That's actually coming next from Sean. So you just get one more barrier between you and the actual thing you want to see. So what I wanted to do was take a minute to tell you a little bit about what went into building this thing and why it's different than what you might be used to and why I think it's significantly better.

Tyler McMullen (18:19):

This project has been one hell of a journey. It started with an idea, the core of which was this: we wanted to provide a language-agnostic, developer-friendly edge computing environment without sacrificing speed or security, which doesn't seem like it should be that crazy. But, as it turns out, it's a little bit harder than one might imagine. So speed, as it turns out, isn't really just about being close to your users. It's also about how fast we process requests, how quickly we can get the data actually out onto the wire for your users. For instance, the concept of cold start time is annoying enough in the central cloud. I'm sure a lot of you have been annoyed by this problem over the last several years. If you've gotten into the serverless thing. So it's annoying enough there, but when you talk about edge computing, we're talking about taking this program and running it across thousands of nodes spread around the world.

Tyler McMullen (19:13):

So essentially cold start time becomes multiplied by thousands, right? So this is a deeply problematic thing. So in classic Fastly fashion, rather than just going, "Oh well we'll inch it down a little bit, we'll shave off some milliseconds here and there," we decided that we wanted to just eliminate it entirely, so that's what we tried to do. But if you layer on top of that security. So security isn't really just about multi-tenancy either. Right? Too often, there's this focus on let's protect our customers from each other or protect us from you. The way I look at it is that security is part of it, yeah. But security is really about protecting your users from attackers and from accidents. So what we wanted to build was something that seemed impossible at the start, instant startup, but also extremely granular sandboxing. And when we started the research and development process for this, we quickly discovered that this was in fact, a completely unsolved problem. So obviously we went into research mode.

Tyler McMullen (20:26):

So we read tons of papers, we filled whiteboards, we filled notebooks, and then we filled more whiteboards. Seriously, a lot of whiteboards. So we went from research to prototype development. Then we started building out this multi-talented, diverse team of low-level engineers and compiler engineers and security engineers. And then what was interesting about this is that team then didn't just have to go off and write software. They actually had to go off and interact with the community, build relationships with other companies and the standards community as well. A computing platform like this can't actually be successful by itself. It requires community buy-in and standards buy-in and language support and so on. So we contributed to open-source software and we open-sourced our own software. You may have seen Lucet, which is our WebAssembly compiler and runtime, which has become essentially like the best way to run WebAssembly on the server. Very fast. Very cool. Very proud of this one actually.

Tyler McMullen (21:40):

We engaged with the standards community as well, so we've been contributing to numerous different WebAssembly related specs, pushing some of our own as well as contributing to others. And likewise, you may have seen this yesterday, but we actually developed a pretty strong partnership with Mozilla, Intel, and Red Hat to work toward this, to a solution to this problem as well. It was known as the Bytecode Alliance. This was announced just yesterday. So the whole idea with this is that we're working together to continue to openly develop the foundational pieces that this product is built on.

Tyler McMullen (22:14):

So, this really has been a truly tremendous journey of a project that started years ago at this point. For me, honestly, if you've been here before, you've seen me talking about this repeatedly, so I really cannot even truly describe the feelings involved in launching this thing. This thing has been my baby for the last several years. It really is incredible to approach an idea that seems impossible from the start and watch it go through all the stages, all the twisting and turning and dead ends, and eventually get to the point where not only can we launch a product, not only is it real, but we can also launch a real product around it and be able to show it to you. So thank you very much. And without further ado, I'd like to bring up Mr. Sean Leach to actually show it to you.