This episode of Over the Edge features an interview between Matt Trifiro and Dr. Robert Blumofe is Executive Vice President and Chief Technology Officer at Akamai. In this episode, Dr. Blumofe explains how Akamai revolutionized CDNs, discusses digital transformation and internet growth, important security considerations, and how autonomous technologies will help shape the future of our cities.
This episode of Over the Edge features an interview between Matt Trifiro and Dr. Robert Blumofe is Executive Vice President and Chief Technology Officer at Akamai. As CTO, he guides Akamai’s technology strategy to assess new market opportunities and new platforms for innovation, explore adjacent segments for the business, influence the development of standards, work with Akamai’s largest customers, and convene technology leaders within the company to catalyze innovation and represent Akamai’s technology vision in the marketplace.
In this episode, Dr. Blumofe explains how Akamai revolutionized content delivery networks and continues to evolve and expand its services. He discusses internet growth, adapting to a more robust virtual world, and important considerations for security. Dr. Blumoff also talks about how AI and ML will help shape the future of our cities.
---------
“More and more traffic to and from devices, that does create some interesting problems. You know, one is just the basic problem of scale. You now have more and more traffic from more and more locations. And, so we continue to see a world in which the internet, the volume of traffic continues to grow leaps and bounds, and the importance of the edge continues to grow.”
“Security and reliability get built by layering on top of the internet protocol, not by modifying the internet protocol itself. You layer on top and the power of layering can’t be overstated.”
“While edge computing is great, when you look at all of the microservices that make up an application, I don't think there's many cases where you could say that they all belong at the edge, or that they all belong at the core.
“For a modern application, multiple microservices, you really want to be thinking about for each microservice, where does it belong? Does it belong at the edge? Does it belong at the core? Does it belong somewhere in between?”
“At this point, the cost of connectivity is so low that anything that gets any benefit from being connected well, you might as well connect it.”
“It's a whole spectrum. It's not a one size fits all. But I think increasingly applications need to have at least some component running at, or very near the edge. The core alone really just doesn't solve it.”
"The dramatic change from IoT really is as much on the security side as it is from the scale and the importance of the edge.”
---------
(02:45) Being Jack Benny’s Grandson and Getting into Technology
(05:20) First Memory of the Internet and Career Path
(10:45) Relationship Between Mathematics and Distributed Computing
(12:30) Akamai Journey
(14:45) What Akamai Does and the Start of CDNs
(21:30) How CDNs and Work
(27:30) Akamai Server and CDN Structure
(31:00) Akamai’s Expanding Services and Evolution
(35:30) Akamai’s Acquisition of Linode
(37:00) Edge vs. Core
(42:30) Akamai’s Interconnection Fabric
(44:30) Changes to Internet Traffic, Adaptation, and Security
(47:00) Concept of a Bidirectional CDN
(49:15) Network Security
(53:00) Akamai Network Capacity
(53:15) Most Exciting Innovations (remaking cities with AI and ML)
--------
Over the Edge is brought to you by Dell Technologies to unlock the potential of your infrastructure with edge solutions. From hardware and software to data and operations, across your entire multi-cloud environment, we’re here to help you simplify your edge so you can generate more value. Learn more by visiting DellTechnologies.com/SimplifyYourEdge for more information or click on the link in the show notes.
--------
Connect with Dr. Blumofe on LinkedIn
[00:00:00] Narrator 1: Hello and welcome to Over the Edge.
This episode of Over the Edge features an interview between Matt Trifiro and Dr. Robert Blumofe, Executive Vice President and Chief Technology Officer at Akamai. As CTO, he guides Akamai’s technology strategy to assess new market opportunities and new platforms for innovation, explore adjacent segments for the business, influence the development of standards, work with Akamai’s largest customers, and convene technology leaders within the company to catalyze innovation and represent Akamai’s technology vision in the marketplace.
In this episode, Dr. Blumofe explains how Akamai revolutionized content delivery networks and continues to evolve and expand its services. He discusses internet growth, adapting to a more robust virtual world, and important considerations for security. Dr. Blumoff also talks about how AI and ML will help shape the future of our cities.
But before we get into it, here’s a brief word from our sponsors…
[00:01:10] Narrator 2: Over the Edge is brought to you by Dell Technologies to unlock the potential of your infrastructure with edge solutions. From hardware and software to data and operations, across your entire multi-cloud environment, we’re here to help you simplify your edge so you can generate more value. Learn more by visiting Dell.com for more information or click on the link in the show notes.
[00:01:31] Matt Trifiro: Two years ago when I started the Over the Edge podcast, it was all about edge computing. That's all anybody could talk about. But since then I've realized the edge is part of a much larger red.
Pollution. [00:01:40] That's why I'm pretty proud to be one of the founding leaders of a non-profit organization called the Open Grid Alliance for oga. The OGA is all about incorporating the best of edge technologies across the entire spectrum of connectivity. From the centralized data center to the end use devices, the open grid will span the globe and it will prove performance and economics of new services like private, 5G and smart retail.
If you want to be part of the open grid movement, I suggest you start@opengridalliance.org where you can download the original open grid manifesto and learn about the organization's recent projects and activities, including the launch of its first innovation zone in Las Vegas, Nevada.
[00:02:15] Narrator 1: And now, please enjoy this interview between Matt Trifiro and Dr. Robert Blumofe, Executive Vice President and Chief Technology Officer at Akamai.
[00:02:25] Matt Trifiro: Bobby, how are you doing today?
[00:02:26] Dr. Robert Blumofe: Feeling good. Yeah, it's cooled down a little bit here in Cambridge, Massachusetts. We had some crazy hot days like the rest of the country, but it's cooled down a little bit, so I'm happy about that and I feel good.
[00:02:35] Matt Trifiro: How did you get into technology?
[00:02:36] Dr. Robert Blumofe: You know, I actually come from a show business family. I dunno if, I dunno if you saw that anywhere in my background, but, um, I grew up in Beverly Hills, 9,200 Oh, the whole thing. My dad was a film producer and he, he ran production at United Artists for quite a few years and was director of the American Film Institute in Beverly Hills.
And on my mom's side, my grandfather was the comedian Jack Benny. And of course he was a radio in television. Well, some, actually, probably a lot of you listening have no idea who I'm talking about. Some of you may be on the older side. Older Americans, uh, know exactly who I'm talking about and can remember some of his, his comedy routines.
And was he alive when you were alive? Did you meet him? Yeah. Um, I was 10 when he died, but I have great memories of, of being with granddad and. He was a big baseball fan and, and I was, as a little kid, we, we used to go to Dodger [00:03:20] Stadium. Dodger games. Yeah. Was those experiences, I remember that first, first really my recognition that he was famous cuz you know, we'd go to the games and people would ask for autographs and things like that.
And I did see him perform a couple of times in Vegas in front of a, in front of a live audience. And, and being at the house was always an experience cuz his friends were always there, whether it was the, the Sinatras or George Burns or Billy Wilder, or Gregory Peck or whoever might be there. But that generation of Hollywood, those, those were his friends.
And, and as a little kid I do remember and, and, uh, and even at 10, I was aware that that was special and exciting, but there was never any notion for me that that was something that I wanted to do myself for whatever reason. As a very young kid, I really recognized myself as kind of a nerd, as a math and science.
Unfortunately, I did have a role model in the family. I have an older half brother, he's almost 10 years older than me, and when I was in maybe sixth, seventh grade, he was already at Stanford as an undergraduate and he was taking computer science classes there. And I have a very clear memory of him coming home one summer and he had just gotten this programmable Hewlett Packard calculator.
And, and he showed me, he showed me how to use it. And I remember the first thing about it is you have to learn that so called rpm, and, which I thought was incredibly intuitive and cool. And, and to this day with a calculator, I still prefer to use, use rpm, the idea of like,
[00:04:39] Matt Trifiro: the
idea of a stack is like hard coded into your, your,
your models really.
[00:04:44] Dr. Robert Blumofe: I remember being really Yeah, exactly. The stack. And it's just a really cool way of, of expressing arithmetic. So I was really taken by that, but this thing was also programmable and he showed me the basics of, of programming and, and I was just, Smitten. And so I went [00:05:00] off and I taught myself the assembly cause it's an assembly language so you have to do everything.
There was no basic or, or any high level programming language. It was just assembly. And so I taught myself how to program and he was nice. He let me use it whenever I wanted to. And I would write these stupid little programs. Now was maybe in sixth, seventh grade, but that's kind of what got me started on, on technology at, at an early age.
The other thing I would just say about my childhood is, you know, as, as a kid especially then where there was no internet, there was no video games or anything like that. Kids I think at that age are, are bored a lot and you don't even have as many options on tv. Right. Cuz there's what, two or three channels?
And if your favorite show comes on, it's not on, it's
on Sunday mornings.
Were the worst . Yeah. Saturday mornings were great. Not so much Sunday mornings. And so I, I remember the hobbies and my hobbies were always building things and so I learned how to build model airplanes and, and model railroads. And, and my mom of course, she was very tolerant of all this.
And I'd spend my time in the garage and I'd build these elaborate model airplanes, radio control. Then she'd drive me out to like the, the field where you would fly the things. And I never learned to fly 'em. It was literally months and months of just build Yeah. Months and months of building and then two seconds of flying and it would just, that just, that sounds like me.
Yeah. Yeah. But I was always into, kind of, into technology in one way or another, building things. And then software really came through my, my older brother teaching me how to program.
[00:06:16] Matt Trifiro: Yeah. What, what's your first memory of the internet?
[00:06:18] Dr. Robert Blumofe: Real internet. Probably not until college. Cause I'd have to say in high school, I really didn't pursue any, any pro, I didn't take any programming classes by that point.
They were available. I went to prep school on the east coast at, uh, Phillips Academy and just didn't pursue any of that. So probably wasn't really until college. And then it would've been when I was an undergraduate at Brown. And that was when I first started taking [00:06:40] programming classes. And, and really at that point, the internet was really just email.
And then there was, uh, what, what year, roughly. ,
[00:06:47] Matt Trifiro: 90, 87.
[00:06:49] Dr. Robert Blumofe: 88. Oh, 83. Okay. Yeah, so I was an undergraduate starting, uh, I started in 82. I was class of 86 at Brown, so started in 82. I think I took the first class maybe in 83. I think I was in my second year, sort of in 83, 84. And at that point I started getting access to email through, through Brown.
And, and we, and there was also used in it. So the, the, I would say the main applic, the main internet applications we used were, were, Well, there was no web at that point. I mean, we'll go. Actually, I forget what he No, no. There was,
[00:07:17] Matt Trifiro: there was Prodigy in America Online and, Right, That's right. You had the, There wasn't really a brow browser was 95.
90 95 I think is Mosaic. I think. Yeah,
[00:07:26] Dr. Robert Blumofe: you're right. So, um, yeah, Mosaic came around the mid nineties, but Tim Burners Lee published the spec for HTTP earlier than that. Not much earlier, but earlier. And so you did have other applications like Gopher, I think, where you could use Right. The sort of a non graphical web browser.
But you're right, that came a bit later. So it would've been purely email and Usenet. And I remember back then I was, I was pretty big deadhead and there was a great Usenet forum where you could, you could get access to audio recording. So you had to download these ridiculous encoded files and then concatenate them together and then run a, a program that would ta that would turn the text into an actual audio file.
And, and, but there was, there was mechanisms for doing that even back. Yeah.
[00:08:07] Matt Trifiro: And you're, you're now the CTO of Akamai and I guess you've been there almost 25 years, right? Yeah,
[00:08:14] Dr. Robert Blumofe: yeah. I just joined in that, That's
[00:08:16] Matt Trifiro: an unheard of tech career. That's amazing.
[00:08:18] Dr. Robert Blumofe: I know. Crazy. But I, I joined the [00:08:20] company in 99. How many employees?
I think I'm employee number one. 30 something. It was actually really my second real sort of professional programming gig when I was at Brown. I took a year off and spoken a lot about one of the, one of my great mentors at Brown, sort of a legendary professor there, uh, Andy Van Dam, who taught the first programming class and was really sort of the inspiration.
And he was one of those guys who saw something in me, even though I was kind of a screw up at that time. And, and, uh, he, when I took a year off, he got me a job at this incredible company called Cadre Technologies. It was a startup in Providence, Rhode Island, and I think I was employee. 30 something there.
So it was, it was small. And that was my first professional programming experience. And it was a really great group of dedicated, really smart programmers. And they kind of, you know, I was the young kid at that point. I was only, maybe I was maybe 20. And they really sort of almost took me under their wing and, and, and sort of, they, they taught me a ton.
These were experienced professional programmers. I was more of a hobbyist at that point or coming from the academic. And it was really a, a re just a remarkable experience And I've Andy to thank for that and, and he did a lot more for me than that. But also ultimately between that experience and, and working with Andy, I managed to get myself into MIT for the graduate program.
And that really, and the reason I bring it up is cuz it's so directly connected to, to Akamai, it was really the experience at MIT and the people I met there that led to, to, to Akamai, cuz that was nine, sorry, that was 89. I, I started at, at the lab for computer science. I was in the theory group. and my PhD advisor was Charles Lyson, who maybe some of your audience might know [00:10:00] as one of the, one of the four authors of the, the famous, uh, algorithms textbook, the Corman Leers and Resin Stein algorithms textbook.
So Charles was my, my PhD advisor. And I also worked very closely with Professor Tom Layton, cuz they were really, they were sort of the two professors who were working on algorithms for distributed and parallel systems. That was my interest. I I, I wanted to be on the mathematical side of things, but I was also very interested in parallel and distributed systems.
Not for any particular application, by the way. It was really, it was really about the mathematics and I was just really interested. That's what I wanted to do. And those were the two professors in the theory group who were working. I'm gonna ask you
[00:10:37] Matt Trifiro: a really, really naive question. How, how does, how do mathematics and distributed computing relate?
[00:10:43] Dr. Robert Blumofe: Basically, I would have to say that the, the types of things we were looking at at that time were, it, it, for example, a distributed system is oftentimes represented as a network. So you can think of it as a graph. You've got nodes. That represent computers or storage locations. And then you've got links between those nodes that represent the wires, if you will, or the pipes, or tubes, whatever word you want to use, that interconnects them and.
There are basically then fundamental questions of if you are located in one place on that network and your data is somewhere else, how do you get from one place to another? And how do you move data from one place to another in an efficient way? Cuz you've typically got multiple data flows that all have to be coordinated across that network.
And there's constraints. Each node is of a particular size, the links are of particular sizes. And so the, the, the questions of where do you place data? How do you find data, how do you orchestrate the movements of data? How do you orchestrate the placements of computing on which nodes do you wanna [00:11:40] do, which computing and things like that.
All those are basically fundamental algorithmic questions. And by the way, I was mostly interested in from, from the, mostly interested in these questions from the algorithm side. There's a complexity side, so very crudely when you think of theory of computing, there's a complexity side which is about, about sort of proving lower.
And there's the algorithm side, which is coming up with solutions improving upper bounds if you will. That's way oversimplified. But I was more interested in the algorithm side, coming up with solutions to the problems and then characterizing the efficiency of those solutions, which then gives you upper bounds.
And,
[00:12:15] Matt Trifiro: And how did you go from a PhD student at, at MIT to Akamai? Now Akamai came out of mit, right? Is that correct? Right,
[00:12:22] Dr. Robert Blumofe: so that's exactly how I got there cuz my experience with Tom, Tom Layton, he then went on to become one of the co-founders of Akamai with his graduate student, Danny Lewin. They started the company in summer of 98 and I joined and I knew about it.
I, you know, I knew about Tom and I knew that a bunch of my, a bunch of my friends and colleagues from MIT had joined the company. And so in the summer of 99, that's when I joined. I've told this story fairly often cuz you know, I didn't know what the company did. Honestly, it, it was all I really knew. I knew it was Tom's company.
I knew my former advisor, Charles Leiser was taking a sabbatical there. I knew my academic brother, uh, Bruce Mags was the head of engineering. I knew a bunch of other former colleagues of mine and friends of mine were at the company. But other than that, I didn't know. . But what my logic was was, well, it doesn't really matter what the company does because Right.
I get to work with
[00:13:14] Matt Trifiro: these amazing people, ,
[00:13:15] Dr. Robert Blumofe: that's it. You know? That's it. It it, there was, it was [00:13:20] almost obvious to me that if I stayed close to these people, that something good would come out of it. It would be interesting and it would be fun cuz all these people, I'll have good sense of humor. They're fun to be with.
And, you know, motivating. It's actually great
[00:13:32] Matt Trifiro: career advice. I mean, I would say that a lot of my career has been guided by the same , you know, kind of a ad hoc principles, but I, I've never thought of quite encapsulating that way, but that's a, that's actually really, that's actually a really interesting way of, of thinking about your career.
[00:13:46] Dr. Robert Blumofe: Yeah. As I, as I've gotten older, I like to give unsolicited advice, especially to my kids, and they don't really listen much. Like I probably didn't listen to my parents, but yeah. That, that,
[00:13:55] Matt Trifiro: that's, They do, they'll remember when they make a mistake that that doesn't go along with what you told 'em to do.
They'll be, Oh, my dad told me . Yeah. That's
[00:14:01] Dr. Robert Blumofe: how I am now. Oh yeah. My dad, my dad told me, told me that years ago if only I had listened. But it doesn't really register with them. But indeed, I, I do, I, I say, you know, pay attention to, to the people. Follow the people, be around people who inspire you and motivate you, who, who support you and are positive.
And I always add in sense of humor. You gotta keep a sense of humor. You gotta stay around people who enjoy the, the humor and, and find the joy and the environment that we're in. So yeah. Follow the people. Yeah. Well,
[00:14:29] Matt Trifiro: is that so, so it's funny you say that. Back at the time you didn't even know what Akamai really did.
And I, I think a lot of people in my audience don't actually know what Akamai did. Does. And, and certainly if they have a sense of what Akamai does, they don't have a full sense of, of how. Large and, and extensive akamai's network and product solutions are So, can, can you tell me what Akamai does? Yeah,
[00:14:52] Dr. Robert Blumofe: sure.
Well, as I, as I discovered shortly after I joined, Akamai had invented and, and is still to this day, at least as part of [00:15:00] the business, what's commonly referred to now as a content delivery network or cdn, and that's now become a pretty common term. I think it's safe to say that Tom and Danny really invented content delivery, the, the content delivery network.
Now, there's, there's many out there today, but I, I think it's safe to say that they were, that they were the first with the, with the founding of Akamai and really even the, the, the concepts of what is a CDN that they might not have called at, that at the time went, went to, even before Akamai, when they were still running this as a research project at, at mit, but to what is a content delivery network.
It's basically, it's a distributed network of machines that actually act as, as proxies for an application. And the fact that they're distributed is very important cuz it basically works as. . Imagine that you've got an a website and you're hosting your website, say here in Cambridge, Massachusetts. So you've got, you've got a data center or, or it could be nowadays, it could be a cloud data center, Doesn't matter.
I don't think there's a cloud data center here in Cambridge, but somewhere nearby there is. So anyway, you're here in Cambridge. You're hosting your application here in Cambridge, but you wanna make that application available all around the world. And there might be somebody in Mumbai, India who wants to access your application.
Now it's a long. From Mumbai, India to Cambridge, Massachusetts with not only a lot of geographical distance, but a lot of network distance from say whatever network from Reliance network in, in Mumbai to some, some backbone provider across the ocean to a network here in the us, maybe at and t to whoever the hosting network is.
So there's probably 3, 4, 5 networks in between, probably [00:16:40] dozens of routers and links in between. Not to mention all the geography. So for that user in Mumbai to communicate with a website here in Cambridge, You can imagine the performance is not gonna be very good. What we do is, as Akamai, what a content delivery network does is we have servers all over the world, including servers in Mumbai, on the local network on, for example, the Reliance Network in in Mumbai.
So when that user in Mumbai goes to request some content from this web server, instead of that request going all the way to Cambridge, it instead gets redirected to an Akamai server that's nearby. For example, the Akamai server that is in Mumbai on the Reliance Network or whatever, Akamai server is nearest them, and that's a piece of, of mathematics and algorithms and secret sauce, if you will, that we call mapper.
When that's one of the early inventions, or at least an initial version of mapper, is one of the very first inventions of the company. How do you redirect that end user's? To the Akamai server that is n near to them and best able to handle that request. So anyway, so So the request, instead of having to traverse all that distance to Cambridge and instead goes to the Akamai server nearby, and that Akamai server nearby actually serves up the content to the best of its ability.
Now, it may not have all the content in its cash cashing is a big piece of that content delivery technology. Not everything, but it's an important part of it. If the server already has the content in its cash, it can serve it up right directly from the cash, and the communication happens locally within Mumbai on the Reliance Network.
If the server doesn't, well, then there's a whole hierarchy that it goes through to try and find a copy, ideally, without having to go [00:18:20] all the way back to Cambridge. If it does have to go all the way back to Cambridge, it does, but most often it doesn't have to go all the way back to Cambridge and you can find the content near the end user server from a server that is near that, that end user.
So that's the basic idea of a content delivery network is it's, it's a very, Yes. Your,
[00:18:35] Matt Trifiro: your job at Akamai is to predict what people are going to need based on the application profile and what you've determined is happening using some magic secret sauce, and then to essentially preload the content you think they're gonna need into a cash, which is these machines.
And they're all over the globe, right? How many of these machines are
[00:18:54] Dr. Robert Blumofe: there, are there Now we're up to about 350,000 machines located in thousands of locations. About a thousand cities, 135 countries. 1300 plus different networks where we're directly connected directly inside those networks. So it's, it's very large now, which of course has been a consequence of the, of the growing scale.
And, and there's a whole conversation we can have about how we've expanded from that seed of being just a basic content delivery network serving up cashed content effectively, which would be the, the images that were embedded on a webpage or the videos that are embedded on a webpage that you could cash and deliver.
That is, that's still a piece of our business, but it's nowhere near the whole business anymore. We've expanded significantly since then, but that was the seeds of the company and that really is in essence what a content delivery network is. Yeah. And, and we'll
[00:19:43] Matt Trifiro: come back to sort of the new lines of, of business that Akamai's gone into over the years.
And, and, and I just wanna talk about this sort of, this fundamental, you know, because I think when, when you, you watch these basic videos on YouTube on how the internet works, you tend to see something like there's a. And there's a server, and then, you know, it's like two cans and a string. [00:20:00] Somebody draws a line and maybe they'll, they'll draw a line with some hops that goes somewhere.
Right. And maybe they'll talk about how packets get distributed and they can take different routes and they get reassembled. But you, you tend to, to see these lines and you think, oh, everybody uses those lines and sense it there. And when we were talking earlier, you were talking about the power of T C P I P and the ability as the sort of core functionality of the internet.
But what very few people realize, and what I'm trying to draw out here is, is that like Akamai is this, it's an overlay Yeah. To what we think of as the internet. And to some extent, the internet doesn't work the way we think necessarily because like you said, when I, if I need a piece of content and that piece of content is on one of Akamai servers and there's a business relationship that makes that possible, My request isn't gonna go over the path that it thinks it's going.
Right. Because it's, It's sending something, an IP address. Yeah. And so to some extent, and I don't mean this to be a bad word, I just get to go a better word, like Akamai, like hijacks that request and process it locally to create this performance improvement. Can you talk a little bit about how it's possible to do that?
Without breaking the internet. Like how, how is that not just completely creating this, like, separate thing that's off in some other place?
[00:21:10] Dr. Robert Blumofe: The, the power of the internet and in particular the internet protocol is you couldn't overstate just how amazing of an invention it is partly due to just the, the stark simplicity of what the internet protocol is.
And there's probably a whole long thing we could talk about, but maybe it is worth, since you, since you brought it up, sort of explaining a little bit of, uh, you know, the power of the internet protocol. The way I think of the internet or definition is just basically a network of networks that all speak this common protocol, which is the internet protocol, and probably the best way, and actually did a whole series of videos [00:21:40] on this that maybe i'll, that we use internally at Akamai for training.
Maybe at some point I'll make available externally, but the, the easiest way to think about it is by analogy to actually the postal service. Because whenever you wanna send information from one place to another, you basically have to take that information and put it into packets. And packets are just like envelope.
And when you put your information into a packet or into an envelope, you then have to address the envelope. Right. And the protocol, much like with the postal service, specifies how you address the envelope. So for example, if I wanted to send a letter to my sister, I can't just write a piece of paper down and say two SISs and hand it to the, to the mailman.
It's not gonna get anywhere. What I've gotta do is I've gotta put it in an envelope and it has to, and it can't just be some piece of garbage. It's gotta be reasonable envelope. And then I've gotta address it in a certain way, right? It's, it's got her name, it's got her street address, it's got the zip code.
And if I do that and hand it to the postal service through some magic, they will get it from here to, to my sister's house. With pretty high probability by the way. It might get lost. It might get destroyed. Well, the internet protocol is exactly the same thing you take. That
[00:22:43] Matt Trifiro: was a great analogy, .
[00:22:44] Dr. Robert Blumofe: It's right, true.
You take data. True. It's true. Yeah. And you put it and you put it into a packet and you hand it to your internet service pro. You hand it to basically the network interface on your computer. And the, the internet protocol is basically what, everything in between the, if you will, the post offices, or in our case now, the.
They look at that address on your packet and they forward, forward, forward, forward, and eventually say, get it to the destination. Then at the destination, you open the packet and you can look inside and the layering works as follows. On the upper layers, it's what goes inside the packet. And you can put lots of different things in there.
I, [00:23:20] I could write in Chinese, I could write in English, I could write in pig Latin. Um, I could write in a special language that only my sister and I understand doesn't matter as long as I put it in that envelope and I address it in a way that the post office could understand. So I could invent new languages.
I can invent new ways of communicating with my sister and as, and, and only me and my sister need to understand that that language, that's, that's the layering that goes on on top. And that's exactly what the worldwide web was. The worldwide web was basically a new specification of things that you would write and put inside those packets and inside those envelopes and, and basically give us our modern, our modern web.
But when Tim Burners Lee invented the worldwide web, Nothing on the internet had to change. All he had to do was basically write the code that takes his, his protocol, which is called http, the hyper protects transfer protocol. All he has to do is basically take that, put those inside of packets, and as long as he does it in the standard way, everything just works.
So the two things on the other side know how to interpret it. And the two, exactly the two things on the, on each end, the server and the, and the client or the browser say they need to understand it, but nothing in the middle had to change. Nothing on the internet had to change because it's only the stuff that's inside the envelope.
Now, the other layering that goes on is, is underneath, which is basically how you get the envelopes from one place to another. So you could invent new postal services, but as long as that postal service,
[00:24:44] Matt Trifiro: rocket service, pneumatic tubes, ,
[00:24:47] Dr. Robert Blumofe: anything you want to, as long as it knows how to read an address on a, on an envelope and get it to the other place.
It works fine. And, and apropo to this, relative to making this point on, on the internet, somebody actually [00:25:00] wrote uh, an RFP request for proposal for what they called internet protocol over AV and carrier. And the idea was basically carrying packets with pigeons. Cause indeed, if you could train pigeons to carry internet packets from one place to another, by simply reading the addresses and sending and going from one place to another, you could literally run the worldwide web.
That'd be a very, very slow internet . It would be a very slow internet, but it would basically work, You know, your video would obviously rebuff all the time, but it would work, um, just very, very, very slow. You could actually run the internet on pigeons if you could train
[00:25:35] Matt Trifiro: your, It's the one bit turning machine equivalent for.
Yeah,
[00:25:39] Dr. Robert Blumofe: and, and so this is an extraordinary thing that internet protocol has that power of just, it, it really does nothing but solve the problem of how do you get a packet from one place to another? It doesn't try and solve reliability, It doesn't try and solve security. It doesn't try and solve anything.
So underneath, we've seen wifi, We've seen 3g, 4g, 5g. We've seen ethernet go from, from one meg to 10 meg to a hundred meg. We've seen all kinds of improvements in the layers underneath. They just keep supporting. And above. We've seen innovation after innovation from from email, and we talked about using it to then the worldwide web and all kinds of video conferencing, like what we're doing right now.
We've seen just countless things on the layers above, which also, by the way, of course include reliability and security and things like that. So those things, security and reliability get built by layering on top of the internet protocol, not by modifying internet protocol itself. You, you layer on top and the [00:26:40] power of layering couldn't be overstated.
[00:26:42] Matt Trifiro: Yeah. That's really a fascinating exercise. And so Akamai has created a layer Yes. Of many, maybe many layers, but certainly it works at a layer above that. So it's completely compatible with the existing internet, but it adds this new functionality. Exactly. Um, and it's comprised of 350 or so servers around the world, thousand plus cities.
Now, what connects those servers to each other and back to the content. Is that a private network that Akamai owns and is built? Or is it what we think of as the traditional transit networks in the internet? Or what does that backbone look like? They
[00:27:17] Dr. Robert Blumofe: all can communicate with each other and with our customers applications and with our customers, the end users, they can all do that communication just over the standard internet using, Again, we use the standard protocols.
We layer just like, just as we discussed, and it can all just run over, over internet with no dedicat. Pipes. Increasingly lately, however, we have been building out our own sort of, if you will, private interconnect as an optimization. So a lot of our service, you might look at a particular cluster of servers in some location, and you'll find that they're just directly connected to the local network provider reliance, or at and t or, or Korea Telecom or, or Telstra, whatever, just connected directly to the local, local provider with nothing else.
That's, that's very common, but increasingly you'll, you'll find that we also will connect to our own private interconnect, our own private fabric, and that's really done as an optimization for cost and for scale and things like that. Functionally, everything just works fine over, over the internet and, and much of our [00:28:20] deployment still just, just uses raw connectivity.
But yeah, as an optimization, we have been building out our own what we, what our, our fabric, our our own interconnection fabric. Yeah. And
[00:28:30] Matt Trifiro: as, as we talked about earlier, that's not uncommon in many ways. I mean, the, the large hyperscalers, the Microsofts and Amazons and Googles and Facebooks have all built out their own global private wan, essentially.
That is, it, It runs in parallel to Akamai, runs in parallel to what we think of as the traditional public internet. Although, as you point out it to me, that since the mid nineties it's been private. So it's all sort of private networks. Yeah. It's just how, how big are they? And I, I heard a statistic, I don't know if Akamai has studied this, but I heard a statistic from a naau friend of mine who shall remain nameless, but someone who would, who'd probably know, and he says that his estimate is that 70 to 80% of the traffic quote on the internet is actually traversing on private backbones.
Do you have a sense, is that a reasonable statistic or how does Akamai think about that? Well,
[00:29:16] Dr. Robert Blumofe: I don't know exactly how he, what he's calling a private backbone. So for example, he meant like on Google. Yeah, go ahead, finish. You know, so for example, if a packet goes from, say, Comcast, To say reliance and, and, and, and part of the way through it uses say a CenturyLink backbone.
Well, CenturyLink's a private company. So did that, does that, is that part of the 70% that he's counting or is he only counting sort of things like other companies that are layered above, if you will? I'm not sure how he's counting. It doesn't seem unreasonable depending on, on how that person's. .
[00:29:47] Matt Trifiro: Yeah. Yeah.
And I, and I, and I think he, he was just drawing the, the loose contrast between the, the, the historical model of the internet that you sort of taught. And again, it's not, it's not actually how it works anymore with the sort of [00:30:00] three tier where you've got a a, a local access issp and that is ISP has a transit relation with a tier one network, which is mci.
WorldCom hasn't been around for a long time, and I think the contrast, the private network is. Amazon direct connect, like right, get on the Amazon's backbone. And by, by controlling those network pieces, you actually get to deliver, and I think this was one of one, one of your points, higher quality of service, potentially control more of the route since we have resilience, perhaps add some security capabilities.
And so it, it, it that, that seems like, it seems natural that a company like Akamai would actually go into new lines of business that utilize that same network. Can you tell me more about beyond just content caching, what Akamai does today and what services it offers?
[00:30:42] Dr. Robert Blumofe: Yeah, so going beyond, Yeah, we talked about cvn and basically caching content and delivering Indeed from there, we've, we've gone far, cuz that was all the way back in 98, 99 when the company was first started.
Probably the, the, the first big change was moving from an environment where we're delivering just cashable content to actually delivering the entire application. Meaning not just the embedded images and say videos, but also being able to deliver the whole thing, the html. And once you do that, actually there's an interesting thing that happened from a technology point of view, which was that to deliver the entire website with the HTML and everything, it re you require a lot of programmability and a lot of customization.
And, and, and I know you wanna get to edge computing and computing at some point, but this is in some ways the seeds of edge computing because in order to deliver a website, we needed a lot of programmability. We it because it's not just cash and deliver, different websites work in different ways. And in order to deliver the [00:31:40] html there's a lot of customization that has to happen, customer by customer, website, by website, and therefore a lot of programmability.
So even to solve the problem of delivering a website and accelerating the delivery of a website, we, we had to develop some basic programmability and we did, and it included things like being able to modify headers and modify the body of a request or a reply. It included a thing called Edge side includes, And Edge side includes was basically the idea of assembling the HTML for multiple pieces.
On the edge, on the fly de, depending on who you're delivering it to, what geography they're in, and all kinds of things that you might use to customize. You customize the HTML by assembling it from pieces through a mechanism that we call the Edge site includes. And, and those are in many ways sort of the, the first steps toward what people now call edge computing.
The next step for us, by the way, was, um, being able to be fully programmable using Java. That time, pretty much everything was Java and the way people were delivering HTML was through things like Tomcat and IBM WebSphere. And so we actually built our first iteration of Edge computing, which is now easily 20 years old.
What we really call computing was what was Ed Java, and it was, it was actually running Java on our servers through Tomcat or through webs. IBM wants
[00:33:01] Matt Trifiro: your retirement. I mean, in many ways the first serverless, and I put that in air quotes, serverless workloads,
[00:33:06] Dr. Robert Blumofe: that's it's serverless today. It's what we would call serverless.
It's what we might call function as a service. Cuz basically the customer just gives us the Java code and we orchestrate where the Java code runs. We fire up tomcat when it's needed and [00:33:20] run the code. When it's no longer needed, we bring it back down so it it, it gets fired up in whatever location. Are needed as many copies as are needed.
So all that orchestration happens automatically. And it really is exactly what today would be called serverless or function, or function as a service. And this was easily 20 years ago that sort of first manifestation of what today would be called Edge Computing, what we called it actually we called it Edge Computing.
But yeah, the, the actual product was with Edge Java because that was the way people wanted to do the programming. Today, there there's more mechanisms available and really JavaScript is kind of the, the predominant programming model. And so if you look at what we offer today with Edge Computing, a product that we call edge workers, instead of being Java running in, say Tomcat, it's now JavaScript running on a, on an interpreter on the No Js I.
In 2018,
[00:34:12] Matt Trifiro: when I put together the first day of the Edge report, I did a bunch of historical research trying to figure out what are the origins of the term edge computing and how's it been defined over time, and can we get a universal definition of it? And one of the very first, in fact, the very first definition I found was for Akamai.
So I, I don't know if the credit is fully due to Akamai, It might have, but that is the first reference that I was able to find back in the late nineties. And. I, I think that to a very real extent, Akamai has been and continues to be a pioneer Yeah. In in edge computing. So, Yeah. Actually, we'll, we'll come back to some of this other things cause I do wanna talk about, So, so edge computing, right?
I mean, in general the, the, there's lots of disagreement on what the term is, but in general it means moving the compute closer to the device or the data or the user. Yeah. And the main reason for that is for performance. Although there are potentially some other reasons [00:35:00] and obviously tanking a piece of code that would have to run on a centralized server and moving it to an Akamai 350,000 servers around the globe is edge computing by definition.
Exactly. So one of the things that Akamai recently did was acquire le, which is a, a infrastructures a service company. You can lease servers, essentially. How does that connect to the core business of, of Akamai? What, what is Akamai seeing that's happening in the world that's moving you in that direction, which is sort of like doubling down on this idea.
Well, it's not just gonna be JavaScript's gonna be. Lots of things. How, how's a, I think about that. It, it is
[00:35:34] Dr. Robert Blumofe: in many ways an evolution and, and a, a really wonderfully synergistic add-on to where we were with edge computing. So, you know, another important point about edge computing, we talked about it being serverless.
We talked about it being function as a service. The other really important point about it is that it's not bound to a particular location. Your application, the function that you give us, the JavaScript that you give us runs everywhere as needed. It's not bound to any, any one location. As such. It's by and large, it's a function as a service operates on the traffic flow.
It's not heavily stateful. And I guess where I'm going with this is, is to say that while Edge computing is a really powerful model for doing some things, especially when you're computing on traffic, Whether that traffic flow is too from a device or a user or whatever, when you're operating on traffic flow's, very powerful model, but it's not the right answer for absolutely everything.
And when you look at a modern web application and you look at the different pieces, right? Today, modern app web applications that you think microservices, different microservices are doing [00:36:40] certain things. Some of the services are very closely connected to operating on that traffic flow. You're looking at a request come in and you're looking at query strings, and you're looking at headers, and you're looking at bodies, and you're constructing responses and things like that.
You're operating on traffic flows, perfect edge computing. Then there's other microservices that are probably more about operating on stored data. Maybe you've got a, a data lake, a data warehouse, say a database, that is the data that you wanna be computing on. Mm-hmm. , and in that case, The edge is probably not your answer.
And because you want to do your compute on that data. And by the way, just as a simple rule of thumb, I oftentimes tell people that when you're thinking about which should I use, should I use Edge, should I use core, if you will, edge versus core, which one should I use? The answer is pretty simple. The answer is follow the data.
What data or what predominantly, what is the data that you're computing on? And you're better off moving the compute to the data than moving the data to the compute. Almost always. So. Compute is much lighter weight generally, you don't wanna be back hauling large amounts of, of
[00:37:46] Matt Trifiro: yeah. A gigabyte application versus a, a terabyte of
[00:37:48] Dr. Robert Blumofe: data.
Yeah. C compute is often is very lightweight and you can move compute specifications very easily. Um, whether it's a function as in like a JavaScript function or a container, they're pretty lightweight these days. So easy to move the compute better off moving the compute to the data than the other way around.
So again, if you're computing on, on data that is in a traffic flow edge is great, but if you're computing on, on stored data, stored data, data that's not in motion, data at rest, well do your compute where that data at rest is, and that's not gonna be the edge. [00:38:20] Generally, your database is probably just in one or two locations.
Maybe you've got it replicated across a couple, couple locations, maybe three at most. But generally data at rest is pretty location bound. And in that model, you probably want your compute to also be location bound, and that's where core computing really is strong. And that's why in many ways cloud computing or computing in the core is so natural and so easy because you just fire up, you know, some virtual machines in the same place that you've got your storage, right?
All of them, whether it's Amazon or LE Node, they all offer storage connected directly to the compute. You can mount volumes or you can use other mechanisms, but your compute is tied to the data. Anyway, so my my point is simply that while edge computing is great, when you look at all of the microservices that make up an application, I don't think there's many cases where you could say that they all belong at the edge or that they all belong at the core.
[00:39:14] Matt Trifiro: Well, even when you say that, I mean there is no single core and there is no single edge. In fact, it's actually a gradient and where you. Your compute and your data can be a function of whatever the business rules are, right? Yeah. It could be a function of carbon footprint, be a function of cost. It could be a function of latency and speed.
It could be a function of reliability and optimizing what, where, what that means for any particular application can be very different depending on what the application is. Yeah,
[00:39:43] Dr. Robert Blumofe: I I think that's a great point. There is a whole spectrum. There's a whole gradient, you know, from edge to core, and as you're constructing an application, you're thinking about all the different microservices that make up your application.
There's a, you're absolutely right, there's a lot of factors that go into the placement. I gave a really [00:40:00] oversimplified answer of, just follow the data, and that's a good starting point. There's a whole bunch of other considerations. You know, local regulations and things like that, that have to also be incorporated into the thinking.
So again, for a modern application, multiple microservices, you really wanna be thinking about for each microservice, where does it belong? Does it belong at the edge? Does it belong at the core? Does it belong somewhere in between? And as you point out, there's not one core, not one edge. In the case of the core.
For example, with Lin node, today we have 11 Le node locations. Well, which of the 11 should you. The, I don't know, Or which two are you gonna pick if you're gonna replicate across a couple of them? Now, of course, we're gonna expand Akamai. Being Akamai, and given our customer demands, we want to go well beyond those 11 locations.
We're talking about having sort of a middle tier of distributed compute that will move into hundreds of locations where we can support virtual machines and containerized compute across hundreds of locations in a location unbound model, meaning you'd be running in all the locations as the demand arises.
So you know, we'll expand our LE node footprint by quite a bit in response to customer expectations. Customer demand, that's really what the market needs, what it wants. But the net result is that you've got, you've got core in, say, tens of locations. You've got kind of the middle tier in hundreds of locations.
You've got the edge in thousands of locations. They're each good for different. and they're each needed for most applications. It's not a one size fits all. It's not this application has 10 microservices, put them all at the core or put them all in the middle, put them all at the edge. No, it's, it's sort of, you need to orchestrate them across the whole spectrum.
And then ultimately they have [00:41:40] to be able to work together securely. They have build a message with each other. That's where our interconnection fabric comes in. That's where our mapping technology comes in. That's where our orchestration technology comes in. Cuz ultimately, again, multiple microservices running in multiple locations that all have to be securely interconnected and orchestrated.
That's kind of the vision for what the future of the distributed cloud is.
[00:42:01] Matt Trifiro: Now, when you say akamai's interconnection fabric, can you help my audience understand how that relates to what we think of traditional interconnection, these major ixp and the Equinix and the digital realties, and the top IXs and the diff, How?
How does Akamai. interconnection fabric relate to what I think as, as as internet interconnection through
[00:42:21] Dr. Robert Blumofe: an ixp. Yeah, it's a great question that there's multiple pieces to it. One of which we've already talked a bit about sort of these private overlay networks that in some sense optimize so that rather than using say a carrier, we actually have have our own private interconnectivity between locations.
So, so part of it is that sort of physical layer where you have this overlay, physical interconnected network. And again, that's really largely an optimization. But the other big part of it, and probably most crucially, is the software layer, which is messaging, orchestration. How do you make sure that you can communicate across all these different components Cuz your, your microservices often to talk to each other through various and, and find.
So we, we often refer to this as, as an intelligent mesh. And this intelligent mesh is mainly software services that allow the microservices, our customers microservices, to find each other, to communicate with each other and to do so in a secure fashion. It's interesting,
and
[00:43:19] Matt Trifiro: I think [00:43:20] this for explanatory power, most of the examples we've used have been of, of the type of, of a human consuming content.
And I think most of the internet has been that way, right? I mean, in fact, one, one of the things that obviously drove the need for the original CDN was humans desiring richer content, video, large images, those sorts of things that were easily casted because they're static. The video does, you know, the Game of Thrones doesn't change when you're watching it, But, but, and I'm fond of saying this, and, and you may actually have a, a, a more scientific perspective on this, given how much of.
A perspective on the entire internet. Akamai must have, but I see a trend where we're moving from an internet that's primarily humans talking to machines or maybe humans talking to humans like we are now, which is measured in ones of seconds, hundreds of milliseconds. There's a, there's a lot of tolerance for packets being dropped and that sort of thing into a world where it's primarily machines talking to machines.
And the machines may have, may produce data at a, at a much higher rate. They may need a, a much higher degree of reliability. There may be some, some some physical consequences. If you don't have a reliable network like the laser LA may, may miss its cutting angle. Right. Can you characterize the changes in internet traffic that you are seeing and that you're predicting, and then how Akamai sees itself potentially adapting to that new.
[00:44:39] Dr. Robert Blumofe: Yeah, I mean, so you're absolutely right that more and more, not surprising probably to any of your listeners is um, yeah, more and more you see the traffic to and from, uh, devices that are not necessarily with a human sitting next to them, and I think that's gonna continue. There's. At this point, the cost of connectivity is so low that anything that gets any benefit from being [00:45:00] connected, well, you might as well connect it.
Whether it's a piece of apparel or, or, or a refrigerator or an appliance or, or a thermostat or an actuator or whatever it is. It, the cost of connectivity is so low. The, the barrier is low, meaning if there's any reasonable benefit that you get from the connectivity, well connect. So indeed, more and more traffic to and from devices, that does create some interesting problems.
One is just the basic problem of scale. You now have more and more traffic from more and more locations. And so we continue to see a world in which the internet, the volume of traffic continues to grow leaps and bounds, and the importance of the edge continues to grow again, which is not to say that the core isn't also important.
I think our acquisition of Le Node is a recognition that it's a whole spectrum. It's not a one size fits all, but I think increasingly applications need to have at least some component running at or very near the edge. The core alone really just doesn't solve it, and that's becoming more and more of the case, especially with devices and especially as things scale.
But I think the other big. Factor that you have to consider with this is, and you've touched on this, is the security aspect of it. Many of these devices are consumer grade. They're very inexpensive. You don't have an IT administrator maintaining them, so they have to sort of auto upgrade. Oftentimes it's maybe not available even to do the auto upgrade because the connectivity isn't always there or even just, just it's cost.
So a lot of these devices run old software. They don't get upgrades very often. And so you now live in, in a world where potentially as a corporation, for example, you could see data being exfiltrated by a thermostat. You might have a [00:46:40] thermostat in one of your buildings, and that's exfiltrating data. We've seen stuff like that.
So it does change. I, I think in many ways, I would say that the, the, the dramatic change from, from IOT really is, is much on the security side as it is from the scale and the importance of the edge. And it, and it, and it does mean that, for example, just because something is in your building connected to your local network, you can't necessarily just let it talk.
Everything has to be controlled. And you know, people now use the term zero trust, meaning just because it's on the network doesn't mean you trust it. You still have to inspect and control all of your traffic flows, whether they're in the building, on the network, or over the internet, wherever they are.
And I think I o t really underscores the need for a model like that where every single traffic flow is controlled, is inspected, is monitored, and things like that. Is there
[00:47:28] Matt Trifiro: an emerging concept of an, and I don't know what to call it, but I I think of it as like almost like a reverse cdn, right? Cause the original problem was I have all this great content at the core that I need to get out the edge.
And now we've got potentially more over time content with data. You know, you got a 30 cent sensor that just on all day just generating data potentially that maybe somebody could find useful. And so there's a plethora of data that's being generated at the edge and increasingly be generate the edge as we connect more thermostats and sensors and cars and all these things to the internet.
Is there a conception of the Akamai network that. Goes the other direction now have, have
[00:48:02] Dr. Robert Blumofe: you thought about that? Yeah, yeah. It is actually. And, and this was definitely a, uh, sort of a bit of a mindset adjust and, and a bit of a shift, which is to think of the CDN as bidirectional. Cuz I think in the early days when you started off and really thought of our network as just purely a cdn, and by the way, the way we sort of think about ourselves is [00:48:20] we think of, we have this platform and a port and, and we have a CDN business, but it's one of three big businesses.
We've got delivery cdn, there's security and there's compute. And we think of them as distinct things, cdn, security and compute. So CDN is just a piece of what we do on our platform. So we no longer really think of our platform just as a CDN cuz it supports all those capabilities, but. The platform is definitely bidirectional.
In the, in the early days of pure cdn, it was by and large unit directional. It was really from the origin, if you will, or our customers web server out to the end users. Now the traffic goes bidirectionally. Cuz you know, even as we started taking on social media customers in the early days, it was bidirectional because there's uploading cu end users are uploading traffic as much as they're download.
Um, so the traffic going through our network was no longer unidirectional. It was bidirectional. And that was 20 years ago. Um, as things were getting started with early social media, nowadays when you start adding on compute and start adding on security, it's very, for the reasons you just described, it's very much bidirectional.
I haven't done a measure to say how much of our traffic is outbound versus inbound. I don't know what those exact fractions are, but let's just say there's a lot of inbound now. And, and every inbound request has to be inspected from a security point of view, cuz that's where the attacks come from. When, when attackers go after our customers or go after us, that's inbound traffic looking to exploit vulnerability.
So all that inbound traffic, now you gotta compute on it cuz you gotta inspect it and say, is there, is there a sequel injection attack here? Is there a local file inclusion attack here? Is it a DDoS attack? And so you gotta char, is it a boss? Some
[00:49:51] Matt Trifiro: of the. Edge workloads today are your own internal workloads that are processing data on behalf of your customers.
[00:49:58] Dr. Robert Blumofe: Absolutely. First Absolut. Yeah. Yeah, and, and [00:50:00] I think of that as a form of compute, by the way. It's, it's sort of an applica, it's sort of an edge computing application, many anyway. Security is kind of the poster child edge computing application because just as I said, you don't wanna back haul the data to your compute.
You wanna move the compute to the data. I use the same rule of thumb for security, rather than back hauling your traffic to your security stack, you're better off moving your security stack to the traffic, which means it has to be ubiquitous. You have to run your security stack at the edge where the devices are, where the users are, and also these days where the applications are.
20 years ago, you might have had all your applications in one data. Nowadays, you're probably in multiple clouds for, if you're in one cloud, you're in multiple locations. I, it's, it's rare to find a decent size enterprise that doesn't have LO applications now located in many, many locations and probably most enterprises don't even know where all their applications are.
Now,
[00:50:46] Matt Trifiro: in a, in a distributed network like AKA among 350,000 nodes and all this, is there a security advantage? You know, intuitively to me it seems like, like if you've got your security detection and mitigation software all the way out on the edge, like let's say running on all or most of these nodes, if you detect a threat, does, does because you're so far down in the network, does that give you more control over isolating and SEC and fire walling that, that bad actor away from the rest of the network?
Yeah. Dis distribution about giving advantage from a security perspective.
[00:51:18] Dr. Robert Blumofe: Right. There's a couple things that sort of advantage there. One is exactly as you said, is that we can block the attack at its source. So wherever the attack is coming from, probably coming from multiple bot nets distributed around the world, we can.
Right at its source before it gets anywhere near its intended target. And you really want to do, you don't wanna let the attack get anywhere near its intended target, cuz the, [00:51:40] the, the farther you sort of, if you will back haul that attack traffic and let, let it get near its intended target, the more harm it can do.
It can clog up pipes, it can bring down routers, it can find other sources or other vulnerable things that it can compromise and things like that. So, blocking at the source. Clearly better than waiting till it gets to its destination. You know, the other is also by, by being at the edge and by having this one platform that does all of these things, we have great visibility.
So we see so much traffic that our ability to classify traffic and identify what is harmful, what is not harmful, what is a bot, what is a good bot, bad bot. So classifying traffic is another thing that we get by being at the edge and being at scale. I mean, I also point out that it's, it's, it's all one platform.
So the same platform that's securing your application is, is also accelerating your application. You're gonna have to go to two different places, one platform to both deliver your application, accelerate your application, and secure your application. And nowadays also develop your application. Now that we've got.
So you talk
[00:52:37] Matt Trifiro: about all the traffic that you have access to. Can you size that for us? Can you gimme a sense of, of how much traffic the Akamai network handles, uh, on a
[00:52:45] Dr. Robert Blumofe: some time period? So if you think of it in terms of bandwidth, say, say it's per second, we routinely run over a couple hundred terets per second.
I think we've peaked over two 50 terets per second. And it's hard to put those kinds of numbers in perspective. I know people like to say, you know, with that you could download this many things. I don't, I don't know that I find those kinds of comparisons useful. It's a lot. I mean, I remember in the early days when we were looking at, you know, we were running the network at megabits per second and we were so excited when we hit a gigabit per second, a gigabit per second.
And then, you know, now here we are at hundreds of terabytes per second. It's rather astounding. Now, of course, a fair [00:53:20] amount of that, if you measure in traffic, a fair amount of that is coming from the delivery part of our business media. Cuz when it comes to, to bits, the way you consume bits quickly is with.
And increasingly, you know, multi, you know, augmented reality, virtual reality, things that might over time become metaverse, things like that. All these things are very data hungry and, and, and work at very high bid rates. You know, modern movies work at very high bid rates, but it's gonna just become more so when it, when you know, when you're thinking augmented reality, virtual reality and things like that.
When you look at
[00:53:53] Matt Trifiro: technology today and you know, you've been, this has been something that's captured your imagination for your entire life. Pretty much. Yeah. What, what's most exciting to you today? What are, what are the innovations that most excite you? You know, it's
[00:54:03] Dr. Robert Blumofe: interesting when, when you think these days about, you know, say the things that are on people, the tips of people's tongues, right?
Everybody nowadays is talking about AI and they're talking about the metaverse and, and web three and, and ar, vr and things like that. For me, you know, when I think about those, I try and put 'em in sort of a big perspective and, and I can't help myself, but to think about sort of the, the, the big arcs that I've seen in my life and I'm, I'm aware of in terms of technology evolution and, and, and I always go back to my first love, which is aviation.
And I try and think about how would people view the arc of technology when it comes to information technology versus say something else that I'm familiar with, like aviation. So let me, you know, run this little sort of thought experiment, which is, you know, if you think about somebody who died around the turn of the century and you could bring them back 60 years later, 1960, I imagine how astonished they would be by aviation, right?
People crossing the country in jet airplanes, just a few hours across the country, [00:55:00] just a few hours, you know, And then in the sixties, the 7 47 comes along and I mean, it's just staggering. And, and just a few years later, you're putting a human being on the moon and things like that. It would be staggering to them what happened from, you know, 19 hundreds to say 1960s.
Well, Let's do the experiment again and consider somebody who passed away in the 1960s and bring them back today. Let's, And they'll look at aviation. Well, they would recognize aviation. They wouldn't be staggered by anything. It's like, well, it looks roughly the same. The airplanes look roughly the same.
They fly it roughly the same speed. Although, although Elon
[00:55:34] Matt Trifiro: did land a rocket on a moving platform in the water, that, that would've impressed me.
[00:55:39] Dr. Robert Blumofe: Yeah. I mean, but but again, how many of us, none of us are riding. That's right.
[00:55:42] Matt Trifiro: It seems like, Yeah, somebody could figure out how to do that. It's not, Yeah.
[00:55:44] Dr. Robert Blumofe: But by and large, if you look at aviation today, it's to an out to an insider, I'm sure an engineer at at Boeing would tell you about all the amazing things that have changed and a lot of amazing things have changed.
But from an outsider's point of view, it's kind of the same thing there. Your point, the planes look roughly the same. Not a lot has changed from, from that point. Now they might wonder where, where did PanAm go? And, and why aren't we flying on a, on a, on a McDonald Douglas 15 or a Lockheed whatever. So they might, some of the details there, but recognize a.
Well sue the same thought experiment for information technology. And I think, you know, somebody who say passed away in the sixties, you bring them back today, they would be as, as unimpressed with aviation as they might be. They would be staggered by what's happened with telecommunications and . They'd
[00:56:32] Matt Trifiro: see a smart photo like somebody, somebody invented the tricorder,
[00:56:36] Dr. Robert Blumofe: right?
Yeah. I mean, can think about that, you know, walking down the [00:56:40] street and doing video with somebody on the other side of the world in real time and yeah, that just
[00:56:45] Matt Trifiro: how to mention the, Although, although at and t introduced the video phone at the 1964 World's Fair and it's,
[00:56:52] Dr. Robert Blumofe: but I'm talking about to the person on the street and their experience and what they see and, you know, completely different and it's the, it completely different.
It's, it would be staggering. So, Well, let's think, do we think what's gonna be, So let's do, let's do that experiment again. Now. Somebody passes away around. And you bring them back in 60 years, are they gonna be staggered by the changes in telecommunications and information technology, or will it look roughly familiar?
So just like aviation over the last 60 years to an outsider, looks roughly the same. The familiar. Is the next 60 years in telecommunication gonna look roughly familiar, or is it gonna be staggeringly different? And I'm not sure the answer, but it wouldn't surprise me if it looks roughly familiar, if it's roughly the same thing.
Now again, maybe some things have changed and maybe it's more ubiquitous, more available. I'm wearing a holograph
[00:57:40] Matt Trifiro: glasses instead of looking at a flat screen. But
[00:57:43] Dr. Robert Blumofe: yeah, so some dressing around the edges have changed and things like that, but it might be pretty familiar. That wouldn't surprise me, you know?
So for all the talk of AI and Metaverses and things like that, and while I'm sure those things will advance and do some wonderful things for us, it's not clear to me that they are going to rock the world over the next 60 years the way aviation did from 1900 to 1960 or the way telecommunications did from 1960 to 2020.
I don't know. But, but I'd like to think about what, what are maybe some of the possibilities and maybe one possibility might be the, the modern city. Maybe AI [00:58:20] and telecommunications can actually completely remake the modern city. I don't know. I don't think by the way, that it would be some huge leap in self-driving cars.
Just the AI side of it. I think it would be the way we think about cities, the way we construct consid cities and sort of the embedded technology in our roads and in our, and the way we use them and things like that. But I think there's a lot of room to potentially remake the city and, and I focus on this one cuz I live in a city.
I love living in the city. I love the experience of living in the city, but I hate getting around in the city and the way it, there's a lot of negatives and I think a lot of those things are solvable. With a lot of the technology that we have today, and that would completely remake what a city looks like.
I mean, it would be self-driving cars or self-driving transportation. Again, probably not because of some huge innovation in AI by itself, but because of the way we deploy AI and embedded into the cities and, and, and things like that. Well, j just
[00:59:09] Matt Trifiro: imagine what, what single family neighborhoods would look like if we no longer needed to own cars.
[00:59:16] Dr. Robert Blumofe: Yeah. Well, you might still have cars, but they would be orchestrated across much better roads in a way that is efficient and, and again,
[00:59:26] Matt Trifiro: they fold up into a little suitcase that you can like put in your front closet next to your coat.
[00:59:30] Dr. Robert Blumofe: Maybe, Yeah, I mean, , I guess maybe that's sort of wishful thinking cuz I love the experience of living in the city, but I hate these, some of these other aspects of it.
And I do believe that the technology, even AI as currently exists nearly is, is roughly able to solve that problem if you, if you approach it holistically. Again, it's not because you're gonna have a better AI in a car as it's currently, as it currently works in the way the city currently works with detours and roadblocks and the, and fog and, and the guy in the [01:00:00] street, you know, with a sign says, Hey, you better turn over here.
And it's not, not in that mixed environment, but if you remade the, the city with embedded technology and everything. Was automatic and, and self-driving, then maybe it would all would all work anyway. So maybe that's the, the optimist in me is maybe when I think about what could change so that in 60 years when somebody comes back, they're staggered by this incredible improvement.
Just like the other examples that we've talked about. Maybe that's, maybe that's a case, but if I'm guessing, I don't think it's the metaverse and AI by themselves transforming telecommunications as we know it. I think, I think a lot of that might be pretty familiar over the next 60 years. I, I don't know.
Yeah.
[01:00:42] Matt Trifiro: That, that's, that's an interesting way of thinking about it. Like what is that thing that, like, like you say, aviation, where it's just so completely transformative and new and it's real. It's a really interesting thought exercise to go through. Yeah. Yeah. Bobby, this has been a fabulous conversation. I wish we had another hour.
Maybe I'll bring you back on the show in season three, but if people wanna find out more about you or get in touch with you online or find out more about Akamai, where should they go? .
[01:01:06] Dr. Robert Blumofe: Well, you know, obviously we have the, the Akamai, you know, website, which is easy to find. And I'm on, I, I generally, um, I like to use LinkedIn and, and Twitter, and I'm just Robert Buff on both LinkedIn and Twitter, you know, and especially if you're a, if you happen to be a Jack Benny fan and like, like comedy, I like to, I like to talk about my granddad and things that I've learned from him and some of the interesting stories from, from back.
That's,
[01:01:27] Matt Trifiro: there's a great, uh, fast company article. I think we'll put a link to the show net, so it's, uh,
[01:01:31] Dr. Robert Blumofe: it's a fun. That'd be great. Yeah. I did, uh, yeah, I had fun writing about, about grandad for, for a Fast company, and, uh, wrote a little thing on LinkedIn about a wonderful, uh, letter that I have from John F.
Kennedy to [01:01:40] my grandfather, which has an amazing story behind it and is super fun. But anyway, yeah, so I'm on, on some of those social media things and, um, yeah, love to
[01:01:48] Matt Trifiro: hear. Bobby, thank, thank you for taking the time. This has been,
[01:01:51] Narrator 2: That does it for this episode of Over the Edge. If you're enjoying the show, please leave a rating and a review and tell a friend.
Over the Edge is made possible through the gens sponsorship of our partners at Dell Technologies. Simplify your edge so you can generate more value. Learn more by visiting dell.com.