⚡️ Welcome to Speed Week
Presented by: Rustam Lalkaka, Marc Lamik
Originally aired on April 2, 2023 @ 6:00 AM - 6:30 AM EDT
Join Rustam Lalkaka (Director of Product, Cloudflare) and Marc Lamik (Director of Product, Cloudflare) for an Speed Week fireside chat.
Read the blog post: Welcome to Speed Week and a Waitless Internet
Visit the Speed Week Hub for every announcement and CFTV episode — check back all week for more!"
English
Speed Week
Transcript (Beta)
Welcome to Speed Week. I'm Rustam Lalkaka and I'm joined by Marc Lamik again. So we're both here on the product team at Cloudflare and we're here today to talk to you about all the exciting things that we are doing and launching to help make the Internet a faster and better place for everyone.
Exciting week. I almost made myself another shot of espresso before the segment, but I don't need to be any more hyped and my espresso machine is not fast enough.
So, you know, need some Speed Week there.
Marc, what is the poster behind you? I know I mistook it for a broken TV at one point, but I think it's actually relevant to Speed Week.
It is, it is. It's not a broken TV, luckily, but that is the first guess of many people.
And it's not the network of our more than 250 points of presence that we've announced today, but it's actually flight routes, how they were all around the globe.
I guess most probably a third of them is not existing at the moment, but this was when, yeah, when you could get from continent to continent as a person as fast as possible.
It's actually an interesting, shows how quickly sort of the speed at which we move information across the world has progressed, right?
In late 1800s, you know, you were taking boats everywhere and then you're taking planes.
Now we have the Internet and Zoom.
Cool. So what's the right way to talk about all the cool stuff we're launching this week?
Obviously, there's a lot on the plate. Maybe since we're product managers, we should just sort of talk about our customers and what problems they have and then how we solve them sort of in the classic Cloudflare or product management methodology.
How does that sound? Sounds great. Let's start with our customers.
Who cares about speed? Yeah. Who cares about speed? Yeah. I would say generally everyone who uses the Internet, like if you start with who isn't annoyed by a website not loading, that's generally everyone.
But if you look more like into our customers, I think the first thing that comes to my mind, which also kind of is related to my personal background is e-commerce.
I think there's nothing worse than you're seeing a great product on search or on a website, you're clicking on it, you want to see the details and it's it takes forever.
And then you're very fast.
If you like open several products, you're very fast to just not select the one that loads slowest because you're just annoyed by it loading slowly.
That's actually the case with many people. And that conversion in e-commerce is very highly related to the speed of the website, especially in a time and age where everyone uses their mobile phone to load their to do their online shopping.
So, okay.
So e-commerce is a big use case. A lot of our customers have transitioned their workforces to remote life as a response to the COVID pandemic.
How does performance sort of make or break employee experiences?
I think the best example is now what we are doing.
We're like kind of you're in San Francisco, I'm in Lisbon and we're doing a joint video session together.
If we would have like a leg that would make it impossible to have a conversation, that would be pretty annoying.
On the other hand, it's also not only this, but like collaboration is super important with teams and now with like during COVID people not being in the office, sometimes there is the need to communicate in real time to have data sent over as fast as possible.
If you're working on a document at the same time, it should be real time.
It shouldn't be like delayed. And that's really important. And I think that's something we got used to quite quickly that it goes fast for most of us.
But it actually, if you're looking to the data, it's actually not fast for everyone.
I think the other thing here is that we've all seen the news. We've all talked to lots of customers who are experiencing, either experienced or are worried about sort of pretty scary cyber attacks impacting their business.
And so implementing the sort of best of breed industry standard sort of security protocols and techniques on a corporate network used to mean that you need to make your network really slow.
But that's like an unacceptable trade-off picking between a secure network and a fast network now.
And to your point, like you need to be able to collaborate and do your job effectively from home while still protecting the enterprise.
And that's with sort of legacy technologies is a hard proposition, but we're going to launch some stuff this week that hopefully makes that a little easier.
Cool. You mentioned e-commerce.
Let's walk through the sort of life cycle of an e -commerce transaction, right?
So I just bought some pants yesterday, don't have enough pants, just wearing shorts all day.
And I Googled pants, cool, comfortable pants.
What's the sort of life cycle of an e-commerce transaction from there?
I see some interesting pants on Google. I click the link and ideally I'll go all the way through to checkout, right?
What are the sort of touch points that we can optimize in that journey?
And what are we going to do here without spilling too many of the beans?
For sure. I think the first one is of course like the time to first, like when the e -commerce page loads first.
So if you click, if you search for pants, you click on the page, of course, it's super important that it's there as fast as possible.
So you actually can look at the page and look at your pants, not only on the small thumbnail that you find in search, but you can see like the real image and read the description.
AMP is the answer there, right?
Accelerated mobile pages makes everything fast and that's what we're doing here, right?
Well, if you're on mobile and if you have AMP enabled, which is also not that super easy, but yeah, I guess there's more on that front to make it faster.
Very mysterious. Okay. So I click on the search page.
I like the look of the pants. I click add to cart. How do I make that faster?
Well, generally, I think there's even one step before that, because when you open the product detail page, there's a lot of images on it.
So that's what e -commerce is all about, especially when you talk about fashion.
It's about the images.
You want to see the pictures and you want to see them in a high resolution, because if you just see like a few pixels of your pants, you cannot decide if that's what you want, if that's what you want to wear.
So you need images. They need to be loaded quickly.
They need to be, have a high quality. So I think that's actually a big part of the Internet traffic is images.
So I think that's definitely something that needs a lot of speed.
I am flattered that you think I buy fashionable pants, Mark.
Okay. So, right. So we get the product detail page, momentary images load.
Sounds like we're going to do some stuff that makes handling images on the Internet faster.
I'm excited for that. Okay. And then say we click add to the cart and go to the checkout experience.
This is a really dynamic page, right? Historically, really hard to use standard things like caching, et cetera, on a page that requires talking to the database, et cetera.
Things like Argo can help, but what are other sort of knobs we can turn there to make that experience faster?
I think there's like in the checkout, there's several things.
I think there's a lot of calls that go out to third parties where you need to optimize your traffic, your web traffic as much as possible.
You have a payment provider, you might need to check stock with your stock provider, whatever.
There's a lot of stuff you need to call, but there's also like kind of, you need to make it secure as well.
I think that's something which also makes usually checkouts slower than the rest of the webpage, because you need to make it as secure as possible.
And there's personal data involved, so you need to be quite careful there.
And yeah, you need to make sure that you actually load also, like what you said, your dynamic data as fast as possible and the dynamic content.
You can't rely on a lot of test stuff there, at least at the moment.
Yeah. One of the other things we've done is if you open the Chrome or Firefox or pick your browser of choice, network waterfall and look at what resources are loading when, feels like there's some opportunity for optimization there.
And especially with a page that might have some server think time on the origin side, is there a way we can use that time, that sort of dead time to make things faster?
Cool. What, beyond sort of, what if instead of helping sites with origins that are thinking about things generate responses faster, we actually helped customers get rid of their origins altogether?
What would that look like? I think that's actually one of the things we already talked about today.
Like if you completely move to something like Cloudflare so you do all your compute or most of your compute in the edge and you don't need your origin anymore to get to the user, I think that brings a few improvements.
And one of them is of course, you're as close as possible. I think there's a post from John, our CTO around SpeedWeek where he talks about like, even speed of light is kind of limited.
So you can't get faster than speed of light.
So you need to be as close as possible to your user. And of course, if you do like all the compute on the edge as close as possible to your user, that of course makes it a lot faster than if you need to go back to your origin every time you want to compute something.
I think that's kind of the, one of the easiest and the kind of most obvious improvements that you get from using edge computing.
It sounds like our edge computing product is pretty fast and we might even be making it faster.
Yeah, I think it's the fastest. So we're really happy about being the fastest there.
And that's also like, I think there's a lot of things that people are afraid on edge computing, which like CodeStars, which we've kind of eliminated.
So there's a lot of, I think we proved it by 60%, even like over the last month, the speed of Dell for workers, which is just great to see that even like a fast product can get faster.
And there's a lot we did to make the experience as good as possible for our customers and their users.
What I've learned over the past 12 minutes is that having a full slate of exciting blog posts and feature announcements and case studies rolling out over the next week, and then not really being able to talk about them is frustrating.
There's a lot of cool stuff coming. So you mentioned you had a particular interest in e-commerce prior to that sort of diving into some of these teasers.
Can you talk a little bit more about that? What were you doing before Kloffler?
Why the interest in Kloffler in e-commerce? Yeah, I was part of one of Europe's biggest e-commerce success companies or stories, which is Zalando, big fashion e -commerce or like the biggest fashion e -commerce in Europe.
And that was kind of an interesting journey. When I joined, mobile was not a thing.
Everything was still in desktop browsers and we needed to make pictures as small as possible because also Internet connections were slow.
And then the whole thing changed.
And like at some point in time, speed really became an issue and like really became a topic that we talked about more and more.
I think not only because we want the user to have the best experience, which is of course always the case, but also there's so much more related to speed.
I think the most obvious for a lot of people by now is Google and other search engines rank by how fast your website is.
That's a big kind of criterion for ranking websites, especially on mobile.
So if you have a slow mobile page, then you don't have a high SEO ranking, which actually is pretty bad for your website if you kind of rely on SEO.
So that's why optimizing for speed, at least in the last, let's say maybe three to five years, became really, really important for a lot of companies because they actually, and especially in e-commerce, because they actually want to get a good ranking on search engines, but they also want to get the full website to their users as fast as possible.
What have the, how have the challenges changed or are they mostly the same between 10, 15 years ago and now?
I think it changed in a way that there's not that many more like kind of obvious quick fixes.
Like if you had like at some point in time, 10 years ago, you had your one data center where you had all your servers and you didn't cache anything.
Well, then adding a CDN was a pretty good idea because that already gave you a lot of advantages.
And then like as a next step, okay, you know, cache things, you cache things locally on your origin, but you also add a CDN.
That's pretty cool. And then like as a next step, well, cloud computing became a thing.
So you kind of not only had your origins in like your own one data center, but you had them spread out in different cloud data centers, which already helped a lot in getting the compute and the data closer to your customers.
But that's where like, those were like super obvious. It's kind of, yeah, that's pretty clear that that's a big advantage.
And if you're like moving or if you can scale a lot better, I think scalability in e-commerce, like if you think about Black Friday, which is the biggest day in e-commerce, you had to kind of buy extra servers in the early years.
We had to buy extra servers to physically put them there because otherwise we couldn't have lived through Black Friday.
And then we had to actually see what we do with them because we didn't need them like for the next few months until the growth of the company picked up.
So that became a lot, that on the, as a first step became a lot more, a lot easier when you move to general cloud compute, but then it became even more easier when you're like moving to edge compute because you don't even need to think about, okay, do I need to now spin up a new instance because Black Friday is coming?
Or also do I need to reserve my instances in advance because on Black Friday, everyone wants more instances.
So that was actually also a hassle. And now with edge computing, you don't even need to think about how many instances do I need because you can just grow your traffic and it will be covered by the whole network.
And that in the end, making that step is huge, but then it's also, okay, what is the next one?
What are like, now you need to really optimize small things.
It's really about making the images 10% smaller to really get them faster.
It's really, how can we, we talk about like small cash improvements, like in 15 years ago, we talked about, hey, can we make, we start cashing.
And if we cash 50%, well, that's amazing. Now we're talking about how can we maybe, especially with like big businesses or big e -commerce, but also like other big corporations, they have like so much traffic that even like one or 2% of more cash rate is already amazing.
And already saves a lot of traffic going through the Internet also makes the Internet faster.
So I think on that side, that improved caching is really important.
And especially as the bandwidth overall grows.
Yeah. So it sounds like the problems and sort of desired end result of fast pages, fast experiences for everyone, hasn't changed that much.
And how people consume content and all that might have a little bit, but the big thing is that we're sort of farther and farther out on the tail of things that deliver improvement.
The good news for us is that makes our job really interesting, right?
The problems are quite intricate. And once we do solve them, we're able to still deliver a ton of value to customers.
It's also kind of good news because if chasing these optimizations becomes harder and harder, relying on experts like us to do it instead of doing it yourself is good, right?
And that keeps us in business.
So that sounds good. I think one of the beauty of the things that we build is that we don't build for like a single customer.
We really build for like the whole Internet, everyone can use it.
And I think that's something that makes it super impactful.
If you're thinking back at the time where everyone optimized for themselves, yes, they optimize for themselves, they make their website faster, but we are really trying to build products that are as broad as possible.
So when we like optimizes should be for all customers or for everyone on the Internet.
So I think that's the, that makes it even like smaller improvements makes them really impactful because they're spread so far.
What are you bummed that, well, maybe this is a reverse teaser.
What are you really excited about for next Speed Week and Speed Week in five years?
Like what are going to be some of the transformational things that happen to the Internet?
I think there's a few things.
One thing when we talk about speed, what we at the moment still, and we try to shift the narrative there, we try to talk about speed for few.
So yeah, my Internet is fast.
Your Internet is fast. And there's a bunch of other people that have fast Internet and that get everything super quickly, but it's not when you look at like all over the globe, there's like a lot of places on earth where Internet is not as fast and where, like where the problems are still there that a website is just not loading and where this is really a huge problem.
So I think tackling the Internet speed topic more like really globally and also try to really improve in places where at the moment, no one really cares about improving the Internet speed.
I think that's like kind of a big shift. And also like you see that there's also technologies coming out more and more that really work on this.
So we're not only optimizing for like a single group of users, but we're trying to make the Internet faster as a whole and try to make it also more robust against like, especially on mobile, the connection could like just get lost and that's happening regularly on mobile.
So how can we make also our delivery as robust as possible so we're not running into issues when there's a connection failure on mobile and you can still have a good experience on the website or like on the Internet as a whole.
That's a great point. So what are some of the things that are happening that will help what you just talked about, right?
The less connected parts of the world and not just get on the Internet, get it on the faster Internet.
Well, I think there are several things.
One is of course something that we can do and that is the easy thing for us, but it still helps us like get our presence out there more.
So I think when we talked about the new points of presence, new data centers that we've launched, there's a lot of places that are not like the most obvious ones in Europe or North America, but there's a lot of places where we see that the Internet is slow, but the traffic is really picking up.
So we should be helping to make the Internet faster there.
There's of course things if you think about Internet by satellite, that's actually the thing that makes the Internet faster for people.
So I think that's definitely something.
If it's available for an affordable rate at some point in time, that's amazing.
But that brings me back to the next point. Internet also has to be affordable.
So it's in the end, we have super fast Internet, but you cannot afford it.
That doesn't help you neither. So you need to really look into how can you make the Internet affordable as well.
But there's also like, in terms of technologies, kind of having a more flexible way in terms of delivery.
I like kind of compress more on slower connections, use lower quality images on solar connections.
So you can really make sure that you deliver like kind of the right package to the user that actually takes that connection and their location.
So you have still a great experience that may not have the super high resolution image, but on the other hand, delivers an image and makes sure that the experience still stays good for the users.
Yeah, no, it makes a lot of sense. What do you think are like the biggest trends that will come over the next five years?
I think you hit the nail on the head.
I think there will be, you know, things like satellite Internet and all that are interesting.
I don't think they're going to fundamentally change the way the Internet works, at least in the very near future.
But yeah, I mean, more and more content is going to be consumed on things that are not laptops and things that are traditionally thought of as like the ways that people connect.
And so I think a lot of the trends we've seen over the past five, 10, 15 years are going to just keep going.
And we're going to see, I think we have, you know, 4 billion people on the Internet.
We'll probably have a lot more than that very soon.
So interesting times. I was just looking at a submarine cable map for the, you know, basically the first time there are cables that now cross the Atlantic that are not between New York and France, New York and, or, you know, east coast of the US and UK and EU.
There are finally cables going from Brazil to Africa, right? So we will see more of that sort of connecting of parts of the world that are sort of not traditionally connected over time.
Switching gears a little bit, this is your first speed week at Cloudflare, right?
I'm curious from sort of external, or not external, but fresh paralyzed.
And for the purposes of everyone watching, how does a week like this come together?
Actually, the most surprising or maybe not surprising to anyone that knows Cloudflare a bit better is also speed is not only like the topic, but it's also like how we build things for the week.
It's like really to bring out innovation as fast as possible.
How does it start? Like it starts like a few months before the week where we like kind of collect what can we do for this topic?
And it's really asking all the teams, like the two of us, we work in like completely different teams.
We're focusing on completely different products, but it's really the idea to get everyone together on the product team.
And everyone is looking into, okay, what can we contribute?
What have we built? What are we going to build? What are we planning that could actually fit the topic?
So it is a lot of collaboration. It's really stretching the possible.
So because that's something you only be like really successful if you're stretching the target and looking for like real innovations and not just adding one small feature and calling it like a great success, but really changing the way of how some of the products work.
And I think that's what we try and pushing it really hard to make it happen through those weeks.
And now we're at Speed Week.
And I think we have a really cool lineup of things that we're going to release this week.
Yeah, I know. I think the fact that we're always pushing the limits a little bit is what keeps me excited and keeps me working at Cloudflare.
It also means the week before Speed Week is usually a little stressful, but it comes with the territory, I think.
Cool.
Yeah. So we've put out a couple of blog posts there. You want to talk to them quickly?
Yeah. Let's quickly talk a bit about what we've... We already teased, I guess, most of it, but we've talked a bit about the 250 cities that we've launched.
Okay. So for the purposes, who cares?
Why is 250 cities versus 200 cities a big deal? I think that's something we talked about earlier is there are cities...
We launched a lot of 10 cities in Brazil, which is a place that was undercovered by our data centers, but it's generally not covered as well for the amount of people that live there and for the amount of people that use the Internet.
So adding that makes the Internet faster for a lot of people in those places.
So that's why it actually does make a difference.
Yes, most probably adding another 30 servers in North America wouldn't have made a huge difference, but adding those servers or those points of presence in places that have not been covered yet actually does make a difference for those people to use the Internet.
Cool. Okay. So that's one post. What's next?
Next one is workers. We talked the fastest serverless platform, talked a bit about it.
It's 30% faster than three years ago. It's a lot faster than the competition, which is amazing.
And we've eliminated cold starts, which is also amazing.
So yeah, just about it. Also like on workers, it's just the beginning. There's so much that we are still working on, so much that we're adding.
And I think that's just great where we are, but it's also great where how fast this is continuously developing.
It's crazy that work has only been in the market for three years.
It feels like it's been forever. Yeah. Cool. Okay. Next post. Yeah. Next one is very firmages.
Maybe you want to say a few words around that one. This is actually, so I started at Cloudflare in 2015 as a product manager for our content delivery product.
And from the day I started, this was one of our number one feature requests, being able to serve variants based on accept headers and other things.
It's finally here. It's actually quite complicated to build. And we've done it in a way that sort of abstracts or removes a lot of the historical complexity around the feature from our users.
And we handle that for you. And so that's actually why this took a while to build.
We didn't want to just build sort of minimum, not viable, very product, and then hand it to our customers and say, hey, you got to figure out how to actually make this thing work.
We built a lot of intelligence in the product to make sure that customers can just turn very on, and then it does what you expect and what browsers expect without having to do a bunch of malarkey.
So yeah, excited about that. Yeah. Those were the three announcements for today.
But yeah, as we've said, we're just getting started. So there's many more days of SpeedWeek.
We're not saying yet how many days the SpeedWeek has, because it's always a Cloudflare.
Wow. We're really just not giving our viewers anything here.
We're not telling them what we're launching. We're not telling how many days a week there are.
There might be seven days. There might be eight days.
Who knows? Yeah. Cool. Yeah. That is worth keeping watch. We have like a daily session from tomorrow on with the product managers of each day talking about their features.
We have it actually today as well. So it will be interesting, and there's a lot more to talk about.
Interesting indeed. Thank you all for watching.