Network flow monitoring, Birthday Week update, and a bit of “evenly distributed future”

Hello everyone and welcome to This Week in Net. It's the October 20th, 2023 edition. This week we're going to mix it up a bit with a few different topics, but also a few different guests. At the end, our CDO appears in the segment A Bit of History. I'm João Tomé, based in Lisbon, Portugal. We start with a birthday week recap. Our birthday week was already three weeks ago, but this is a good chance to touch base with some of the highlights and also give some updates. Dina Kozlov is a product manager and she worked in our main innovation week of the year. So here's Dina with a recap. Hi everyone. My name is Dina Kozlov. I'm a product manager here at Cloudflare. I'm based in Lisbon and this year I got the honor to help out with birthday week, which is what we're here to recap. But essentially this birthday week was really exciting. Birthday week is one of our innovation weeks where the whole theme and mission around the week is to give little gifts back to the Internet, which is the best way that we like to celebrate our birthday. And so we had a whole week of exciting announcements starting with on Monday, we told the story around our global footprint. And so we have spent years investing in our network. And so one of the really exciting things that we did was we launched a technical deep dive about how our network works and how we use machine learning to intelligently route traffic across our network. The other thing that we talked about was about how if you're moving from on-prem to Cloudflare, you can reduce your emissions by up to 96%. And so that allows companies to get even closer to their green goals, especially since climate change is top of mind for a lot of companies right now. And they're putting an effort to reduce their carbon footprint. On Tuesday, we called out AWS for the egregious taxes that they put on their customers. So there was one that was recently announced, which is a $43 tax on IPv4 IP addresses. And so what we decided to do was for any customer that is using IPv6 from Cloudflare to their Amazon origin, and therefore bypassing the tax, we are going to give those $43 back to our customers in Cloudflare credits while allowing them to continue supporting both IPv4 and IPv6 through Cloudflare. The other tax that we called them out on, which I think most people are familiar with, is the tax that they put on egress. Their egress fees are incredibly high. And so a lot of developers that end up storing their data on Amazon get surprised with a really large bill at the end of the month. But we have our own alternative to S3, which is called R2, where customers can store their data and we do not charge for egress fees. And so during developer week a few months ago, we announced a product called Super Slurper that allows you to take all of your data from S3 and migrate it to R2. But if you're not ready to make that big of a transition, this birthday week, we announced a new tool called Sippy, which allows you to incrementally move that data at a pace that works for you. And so that way, once your data is on R2, you can enjoy the zero egress fees. On Wednesday, Wednesday was actually our birthday. It was our 13th birthday. And also, in my opinion, the most exciting day because we had so many great announcements around AI and around giving developers and companies the tools to be able to build the next generation of AI applications. So we started out the day with a number of partnerships. So we partnered up with Nvidia to be able to deploy GPUs on our network to allow customers to run complex, complex inference on our network. We partnered up with Hugging Face to be able to give our customers a catalog of the most optimized models. We also partnered up with Meta to get Llamatu to our workers, customers, and finally with Microsoft so that companies will be able to run inference not just on the edge, but also on the device where they see fit. We also launched Workers AI, which is our first inference cloud. And like I said, it's running on GPUs. We launched Vectorize, which is our own vector database, which essentially allows developers to tweak their AI. It allows them to build in customization and essentially give hints to their AI models so that it can get data faster and remember new information that it gets. We launched AI Gateway. So one of the things with AI is that it can get expensive very quickly. And so that is why you want to, one, have visibility into what your cost is. And so that's one of the things that we're giving to customers. We're giving them analytics and observability. But the other thing is that the same with just a regular application, you need to be able to protect it. And by protecting it, you are also keeping unnecessary requests from hitting your server and therefore spending that cost. So we're also allowing caching and rate limiting through our eight-way AI Gateway product. Then Thursday was all about empowering developers. The announcement that I was most excited about was a brand new product called Hybrid Drive. So essentially today if you're using a Cloudflare Worker, you probably have a database that's set up externally and so you need to make requests to it. But your database might be set up, let's say, in North America, but your customers may be located anywhere in the world. And so you want your regional database to be as fast as a global one. And so what we deployed is Hybrid Drive, which is essentially super caching for queries from your worker to your database. And it boosts your performance. And what it does is it does essentially turn any regional database into a globally distributed one and significantly reduces your latency. So definitely recommend playing around with it and trying it out. And then Friday was all about giving all of our customers the level of security that is going to keep them, I guess, future-proof against attacks that are going to come. So one of the things that we did is we made all of our customers' systems on Cloudflare post -quantum secured. All customers automatically get this. We don't think that post-quantum protection should only be available to a subset. It should be available to all. We also launched Turnstile GA. So Turnstile is our reCAPTCHA replacement. It essentially allows you to significantly reduce the number of CAPTCHAs that are served, but also in a privacy-preserving manner. And so this is something that is now available GA, and it is also available for free for anyone to use. But there are many more exciting highlights. And so I recommend to everyone to go back and read through the blog posts and let us know what you think. And that's all. Thank you so much. And that was amazing. You did it in a few minutes. Thank you, Dina. It was a bit too long, but feel free to cut me off. No, no, no, no. No, it was good. It was good because it was so many things. Of course, we presented so many things. You made a few good highlights. And there's more. And we have that reCAP blog post that has all of the announcements and also actually an addition in terms of AI, our launchpad, in a sense, funding, right? Exactly. Those are startups that can participate in that program. We are saving a number of spots, specifically to startups that are building their platform using some of the tools that we announced on Wednesday, the AI Day. It's an addition to that AI Day, in a sense. But just now that it's almost three weeks after birthday week, any feedback that surprised you? Any update that we should make in terms of what was birthday week 2023? I think what we really saw is that, first of all, the technical deep dives really resonate with our readers. The blog post about how we intelligently route traffic across our was actually our most read one. And so, I like that we continue to encourage that at Cloudflare to be able to write about how we solve these technically challenging problems. Another thing that we launched during birthday week that I didn't mention is a simplified pricing, both for workers and for Cloudflare images. And so, I think that's another thing that really resonated with our readers, where they're a lot more encouraged to build and use Cloudflare when they have one simple pricing model that's predictable. And so, really excited to have that be something that helps our customers start using our products even faster and start adopting them and do it in a way that's cost efficient for them. And then I think what we saw from Tuesday's announcements is that customers, and from just overall from the week, is that customers are really ready to make the move from AWS to Cloudflare. What we're hearing is we offer all of the tools that they need to officially make that migration. And so, we're excited to help customers do that to, again, also save up on all of their costs. And then for Wednesday, we officially surpassed 1 million inferences. So, we really are seeing that customers are adopting our tools and are building out the next set of complex applications. Those who don't know what inference is, AI inference in that case. Can you explain a bit? Yeah. So, essentially, when you have a machine learning model, it has to take in some information. It takes context clues to come up with an answer. And so, every time that it runs through that, it's a computationally complex and expensive process. But essentially, that is what inferences are. And at the time, and it's still ongoing, we had people putting GPUs all over the world to make that happen, that AI process happen. Exactly. Last but not least, I was surprised, to be honest. Usually, Hacker News is one of those publications that is very popular in terms of technical, creative. Those who are technical usually love Hacker News. And it's kind of amazing to see this week, one of our birthday week announcements was trending on Hacker News. So, it was on the front page of Hacker News. It was Koffler CP, Incrementally Migrate Data from AWS S3 to Reduce Egress Fees. That was trending there. And you mentioned CP before. It's kind of amazing to see, even a few weeks after, people are still going back and see some of the things and using those new features, right? Exactly. Any last thoughts in terms of birthday week? For you, it was like a big adventure in this case. It was, I bet, an exhilarating and intense experience, right? Yeah. No, it was great. It's such a cross-functional effort. I think almost a hundred people helped contribute to it. But what I love the most is, honestly, learning about all of these different technologies and learning about different protocols and different things that we're doing. And so, that's why I highly recommend, if you're interested, go find a blog post that trickles your interest and read it and try out the product. And almost all of them are available in beta, so you can get started at no cost. Last but not least, things that you're working on right now, and you can disclose, that excites you the most? What am I working right now on? Not related to birthday week, but I get to work on Cloudflare Secrets Store, which we actually announced during Developer Week. It's going to be our secrets management platform that customers will be able to use with workers, firewall rules, access, and eventually just with their own applications that may be hosted on Kubernetes, Google Cloud, and so on. And so, that has been a really exciting project to essentially build from the ground up. And yeah, I'm excited to get it into customers' hands. Good luck. Thank you so much, Dina. Thank you. Bye. From Dina in Lisbon, we go to Atlanta in the US. Chris Draper is a product manager that wrote, with the help of an amazing team, how network flow monitoring is now generally available. So, what is network flow monitoring? Let's hear from Chris. Yeah, so the name of the product that just became generally available for enterprise customers is Magic Network Monitoring. And the whole idea of Magic Network Monitoring is that enterprise customers can take network flow logs, which are generated by their routers, and they can send those logs to Cloudflare for analysis. Cloudflare will look at a customer's traffic logs, parse it, understand it, and then it can alert customers if any DDoS attacks are protected or any other traffic volume anomalies. I'm really excited about this product. It'll give our customers 24 -7 monitoring, DDoS protection, traffic volume monitoring. And we think at the end of the day, it'll help make our customers' networks more secure. Absolutely. And it has different use cases, right? Yeah, it does. So, there are two primary use cases for Magic Network Monitoring. The first use case that we focus a lot on is allowing Magic Transit on-demand customers to use the product. And so, for anyone that needs a little bit of background information, the idea of Magic Transit on -demand is that customers can purchase Magic Transit and run their own network traffic across their own infrastructure. But if a customer detects a problem with their network traffic, then they can reroute their network traffic through Cloudflare for protection. And they can do that in an on-demand nature, hence the name Magic Transit on-demand. One of the things we noticed is that Magic Transit customers often have trouble identifying if there are any DDoS attacks across their entire network. So, the idea of Magic Network Monitoring is that it will monitor a customer's network, look at their network flow logs, and then tell a Magic Transit on-demand customer, hey, a DDoS attack is detected. There are some traffic volume anomalies going on in your network. You should activate Magic Transit on -demand for protection. So, that's kind of our primary use case. There's also a secondary use case. Customers can use Magic Network Monitoring just to get better visibility into their internal traffic. Oftentimes, customers will send a lot of their traffic through Cloudflare's network, but there will be portions of their traffic that they don't send through Cloudflare's network. It's really important to get visibility into that traffic for troubleshooting reasons, as you're trying to find a particular problem within your network, or to look back at your network's traffic and do some analysis to understand why a particular problem occurred and when that problem occurred. Magic Network Monitoring can give customers that end-to-end traffic visibility for traffic on their internal network, so they can make troubleshooting, debugging, forensics just a little bit easier. Last but not least, more generally, outside the blog, what are you working on that excites you the most at this time? One of the things that I'm thinking about right now is Cloudflare 1 traffic visibility. Products like Magic Network Monitoring and Magic WAN are a big part of that. Customers can use lots of different Cloudflare products together, whether it be Magic WAN or Gateway or Cloudflare Tunnel, to manage traffic across their internal enterprise networks. And one of the things that's going to be really important going forward is making sure that customers can actually see traffic flows in between these products as they use them all together. I think it's really going to differentiate Cloudflare's Zero Trust network offering, and I'm really excited about this feature coming out, to be able to show traffic flows in between products in a dynamic, live-in-the -moment type of way. It's something that puts the ecosystem more together, in terms of having visibility to all those products, to different experiences, but they will be more similar in a sense, because you will have visibility to all of them. Yeah, exactly. And I think it'll really simplify customers' troubleshooting experience, once they have a single pane of glass that can show what their network traffic looks like across every Cloudflare product. That's great. Thank you, Chris. Thank you so much for having me. I appreciate it. See you next time. Now, let's go for a few highlights of the week. So, I'm going to share my screen. Here it is. We had a few blog posts this week. One of those is related to hardware. We're going to talk a little bit about hardware at the end, with Rebecca Weekly, but this is about introducing the Project Argus Data Center Ready, Secure, Control Module Design specification. So, this is all about how this decouples server management from the server motherboard. It provides flexibility to implement multiple server management and security solutions with the same server motherboard design. So, this is a little bit on the technical side. It explains a bit how it works. It has a lot of images, CPUs-related things, cost reduction, flexibility. It's all about that, if you want to go over that. Now, let's go to Prisma. So, how Prisma saved 98% on distribution costs with Colfer R2. So, for those who don't know, Prisma is all about giving developers data access layer for modern applications. So, that's their goal. They use Colfer products, and Colfer products provide much of the underlying infrastructure for Prisma Accelerate and Prisma Pulse. So, two products from them. So, this is all about empowering user-focused product development. This is an ongoing collaboration that extends to enhancing Prisma ORM. So, if you want to read all about it, there's this blog post too. It's all about R2 storage, the cost-saving aspect of it, as we discussed previously with Dina. Before we go, let's talk about malicious Red Alert Rocket Alert application targets Israel phone calls, SMS, and user information. So, this is a blog post from our Colfer's One Threat Operations team that became aware of a website hosting a Google Android application, APK, impersonating something that is legitimate. So, it's the Red Alert Rocket Alerts application that is being used in Israel to alert the population when there are rockets being launched. So, to protect themselves, in a sense, and see where something was alerted, in a sense. So, this is all about what we saw there and capabilities of that attack. This is an attacker trying to impersonate a real service. There's a lot here in terms of tests and how to detect malware on your device. So, some advices there too, how to protect yourself with Colfer, Colfer Teams, and also Quad One for Families. It's something that you can try to use to be more on the safe side for these types of malicious situations where someone is impersonating a real-world application that is being used a lot in Israel because of the conflict. We also have another announcement. It's how Colfer expanded operations in Mexico amid a growth across its infrastructure and customer base. So, we now have a local Mexico team. So, that's the sum up of the week. Next, we're going to San Jose, California to the OPC Conference, a summit for the IT ecosystem. Our VP of Infrastructure, Rebecca Wheatley, is there. Hi, my name is Rebecca Wheatley, and I run Infrastructure Engineering at Cloudflare. And I am here calling in to Cloudflare TV and Net to talk about what we're doing here at OCP's Global Summit. So, this is the second year Cloudflare has come in force to the OCP Global Summit. What we are doing is we've made a contribution called Project Argus. Project Argus is an implementation of the data center secure control module, which is basically an opportunity for us to disaggregate our root of trust in our BMC so that we can use more consistent solutions and maybe more integrated solutions over time other than just an ACM stack. And that allows us to have longer amortization and reduce the cost of our main board as we go into the Gen 12 time from an assistance perspective. We're not just here talking about our contributions. We're here learning and working with others who need the system. So, this organization, OCP's Global Summit, has over 4,500 people at the event. There are more than 200 companies. We've had meetings with Mitac, with CPE, with Avis, with obviously Lenovo, with all of our different partners and vendors. They all come. And it's a great opportunity to connect. It's a great opportunity to hear their roadmap updates. It's a wonderful opportunity for us to share what we think about the market and connect. And probably the most interesting thing is that there's over 12 different technology tracks on server design, management, automation, serviceability, security, composable memory solutions, chiplets, a million other topics. And certainly the topic that was most active here was about AI. And every element of AI from training and inference and sessions into systems and server design, immersive cooling solutions, how we're moving towards optics and where packaged optics can help us. It was a really fascinating conversation from folks like Andy Bechtolshein, but over 16 different speakers from NVIDIA, speakers from AMD, Google, Microsoft, lots of different places contributing back for domains from XLA and sort of hardware instruction layers that are in the open source ecosystem to systems specs and rack cooling designs. So, it's one of the only places where you can do hardware, software, the software that matters to hardware people, but still it's software and really get a sense of how the ecosystem is coming together. So, it's been a great week. Cloudflare has shown up in style and just wanted to let you all know. Last but not least, it's time for a bit of history with our CDO, John Graham Cumming, and we're going to see you next week. So, over time, there's been different moments at the company's history, of course, of different products. Back in 2016, we were discussing, you wrote a blog post discussing how Cloudflare, what does Cloudflare do? It was all about evenly distributed feature, the process there. We just announced this week where we were recording during birthday week, our connectivity cloud, in a sense. Back in 2016, there wasn't workers around for developers. Zero Trust was not a particular thing at the time for the company. So, a different company, in a sense, at the time. What was an evenly distributed feature idea back in 2016? Well, I mean, it's based on this quotation from William Gibson, the author of Neuromancer and other things, where he said that the future is already here, it's just not evenly distributed yet. And the idea being that some people are already living in the future versus the rest of things. And I think this applies to us very, very well, because when we got going at the beginning of Cloudflare, if you think about the companies that had really good performance on the Internet and really good security and the latest technologies were a very small number that they were essentially living in the future. Google and Yahoo at the time, Amazon, these were companies that had the best of everything, the best security and the best performance because they had the engineers to do it and the scale. And I think one of the things from Cloudflare's perspective was to say, that's a bit crazy because anyone who's lived through technology over time, things get pushed down into everybody having them over time. I mean, if you think about something like Google Maps, before Google Maps and before some of those mapping programs, having that level of zoomable geographical maps of the entire world was something that even the world's intelligence agencies wanted. It was the best technology and the people who worked on it were very highly paid and working on this crazy geographical modeling stuff and geographical databases. And then all of a sudden, we're just completely used to, I'll go on my phone, just zoom in on an arbitrary city somewhere and have a look at it. And so technology tends to get distributed out to everybody. And what Cloudflare said was, it is crazy that good performance, good security isn't in everybody's hands. It should be a service. And that was really the idea of Evening Distributed Future, was to say, look, what we're trying to do is we're trying to flatten out the future, bring it into the present for everybody. Exactly. And fast forward a few years after, we just announced Connectivity Cloud, in a sense. Well, I mean, Connectivity Cloud is really like, how do you talk about what Cloudflare does? If you think about the different cloud providers, you go back in time. I mean, at some point, you and I are old enough to remember Hotmail being created. And Hotmail was amazing. Oh, you can have your email on the web. It's crazy. Because before that, you had to have an email program and you had to download email onto your computer and all that kind of stuff. And Salesforce said, forget about software. Remember, that was their big thing. No software, right? It's a service. So you had this cloudification of applications. And then 2006 or whatever it was, Amazon comes along. It's like, hey, just get servers. Just get them in the cloud. We'll scale them and all that kind of stuff. And the thing that had yet to be really cloudified was the network. How do you connect all this stuff together? And those other clouds really assumed the network just kind of existed. And the truth is, we're all connecting all the time to everything, right? Your smart meter at home is connecting to something. You're connecting to work. You and I are on a Zoom call right now connecting to each other. Your phone is doing stuff. And this connectivity and the fungibility and the changeability of software needs to be brought to the network. And that's what the connectivity cloud is. So hopefully this allows people to build really, really rich experiences, connecting people and machines together. In a sense, historically, of course, a lot of things were in a roadmap. But were you surprised in terms of evolution of how it was when it was evenly distributed as the future and what it is now in terms of what connectivity cloud is all about? I'm not really surprised in the sense that I don't think Cloudflare ever had to pivot and turn to something else. We had a pretty good idea of what direction we were going from very early on. It was just a question of what do you do stuff in? What does the market want? Do you have time to build it? All this kind of stuff. So we had a pretty long list of stuff. I think perhaps the thing that took off faster than I realized it was going to take off was the death of the corporate VPN and the uptake of Zero Trust or BeyondCorp as Google called it. That's the most obvious way to operate is to use the Internet as your native network for the office. And that came about quite quickly. But beyond that, I don't think there's been big surprises at Cloudflare. I'm just glad it has grown as it has. And there was a pandemic. There was a pandemic in 2020. So that helped remote work. The use of VPNs was more. People could see that they were not very good because they need to use them. People that never used those before. So that probably helped. What the pandemic did was it gave a boost to everything that was working from home, obviously. And that gave a boost to the technologies that made working from home easy. And certainly showed that the VPNs were just very hard to scale and very hard to manage. And a bad experience, frankly, to use VPNs. But we had actually already launched our Zero Trust solution well before the pandemic started. And it was clear that that was the direction things were going in. I think the pandemic just gave it a bit of a boost. Makes sense.