Microsoft Research Podcast - 037 - A Brief History of Networking (and a Bit About the Future Too) with Dr. Victor Bahl

Episode Date: August 15, 2018

If your idea of a great job includes pursuing untethered research, shepherding brilliant researchers and helping shape the long-term strategy of one of the largest tech companies in the world… oh, and also publishing prolifically, authoring patents, winning awards and speaking around the world… you are in good company. That’s what Dr. Victor Bahl, Distinguished Scientist and Director of Mobility and Networking at Microsoft Research, does for a living. And he loves it! Today, in our first live podcast, recorded at MSR’s 2018 Faculty Summit, Dr. Bahl shares some fascinating stories from his long and illustrious career, gives us an inside look at what’s new in networking, and, explains why, in an industry where it pays to be the smartest person in the room, it’s important to be a world-class listener.

Craig Bundy, who was the CTO of this organization, took me out for dinner. I was again a young researcher and I didn't know what it was about. And he started talking to me about this cable DSL duopoly. So he started to explain to me how in rural America people had no access in the Internet. And the duopoly concept was about the fact that everybody was relying on these cable and telcos to provide them connectivity and they would provide connectivity where the business model made sense. So as he was explaining this to me he started to say well Victor we must be able to do something in wireless to prevent that

and then he said so are you gonna do this? This is exactly how you see the are you gonna do this and so here's a CTO of Microsoft Research, you know, that time asking me. So that got me going on this stuff. You're listening to the Microsoft Research Podcast, a show that brings you closer to the cutting edge of technology research and the scientists behind it. I'm your host, Gretchen Huizenga. If your idea of a great job includes pursuing untethered research, shepherding brilliant researchers, and helping shape the long-term strategy of one of the largest tech companies in the world, oh, and also publishing prolifically, authoring patents, winning awards, and speaking around the world, you're in good company. That's what Dr. Victor Ball, Distinguished Scientist and Director of Mobility and Networking at Microsoft

Research, does for a living. And he loves it. Today, in our first live podcast recorded at MSR's 2018 Faculty Summit, Dr. Ball shares some fascinating stories from his long and illustrious career, gives us an inside look at what's new in networking, and explains why, in an industry where it pays to be the smartest person in the room, it's important to be a world-class listener. That and much more on this you be the guest, Victor. You're a distinguished scientist and the director of mobility and networking at MSR, and you say you have one

of the best jobs in the industry. How would you describe what you do? What inspires you? What gets you up in the morning? Well, first, thank you for having me here. Yeah, I do have the best job in the world. I think if I reflect on what I have, you know, I'm around some of the smartest people in the world. They are on the top of their game. Talking to them ups my game. And every day is a learning experience for me. I also feel that this organization is set up amazingly well, providing complete flexibility to the kind of things we want to pursue.

So we can really go after our own passion. I also feel that our CEO has empathy and has shown that. Then that allows us to not only just go after tech problems, but also things that affect the societal benefits. And hey, the pay is not bad either. So, you know, if you take all that into account, it's one of the best jobs I could have had. How would you describe or define the big goal of mobility and networking researchers at MSR? What big issues are you looking at, the big questions you're trying to solve? I think we are trying to build the most reliable, the most accessible, the most secure, fastest network on the planet. And so if you think about it from that perspective,

there are many, many problems that you need to solve, including things like broadband access to the billions of people who don't have it. When I talk about accessibility, that's what I'm talking about. When I talk about reliability, I'm talking about a network that never goes down. And then similarly, there are other problems. Same thing with security. You don't want your When I talk about reliability, I'm talking about a network that never goes down. And then similarly, there are other problems. Same thing with security. You don't want your packets or your data to be compromised in any way. So yeah, we work on some of these things.

And the scale is so large, it's just fabulous for us. Let's talk a little bit about philosophies of research. I kind of ask myself, you know, do you ask different kinds of questions as a researcher in technology? Or do you just ask the same questions with different deadlines? Where do you fall on that question? And where do you think MSR falls on that? Yeah, that's a good question, I think. And you bring up an interesting point about timelines. And I think that's kind of the key here, that I believe in Microsoft research and we have time on our side.

Whereas engineers and folks who work in the product, they have to deliver pretty quickly. Now, with time on our side, that luxury allows us to now try things that may not get there very quickly. And it also allows us to take a lot of risks. And so in research, I believe we deal with ambiguity very well. You know, you give us a whole set of data. You give us a generic problem. And we have trained ourselves to try to identify with great crispness what the hard problem there

is. And then we go after that. So yes, there is a difference between how, you know, engineers, very smart engineers will deal with the problem in the product space and researchers will deal with it in the MSR space. That's an interesting comparison with the engineer and the researcher in the sense that if you're an engineer, you have that deadline. And sometimes that can be a benefit, right? I have to solve this quickly to deliver on a deadline. So it's not like we don't give ourselves deadline, too. So we impose these deadlines on us to sort of put a stake in the ground in some sense of saying, you know, we are on top of this stuff.

So, yeah, deadlines do help. Now, there are different kinds. If you have very short deadlines, then you might compromise just here and there to get to that deadline. Well, and I would say that from a researcher perspective, yeah, deadlines can help incentivize you, but also they don't allow you to take the kinds of risks that may come up with a serendipitous discovery along the way. Yeah. But I mean, to be very fair and honest about it, there's a culture of trying to produce a lot of papers, right?

So they put these self-imposed deadlines and try to publish. That's their product, as I say. In reality, we should be able to take time. I don't think research can be done without leisure, to be honest. You really need time to sit back and think and reflect. Not everybody understands that. I mean, people might sort of say that you're not working, but that's actually not true. This is a completely necessary ingredient to doing this. And that is why, if you think about MSR,

we don't tell our researchers what to work on. But we hire the kind of individuals who are so self-driven that they won't just not do anything. You know what I mean? So yeah, I think that there is a... It's a balance and an ebb and flow probably a little bit off the topic, but it's like slow food and fast food. Sometimes you have better benefits if you take time to reduce the sauce. Yeah, but the hamburger stays very good though. When you're super hungry.

Absolutely. So, Victor, collaboration is a big deal in systems and networking, which totally makes sense if you think about it, networking. Yes. What individuals, organizations, industries do you collaborate with? And why do you think it's so important to work together on the kinds of problems you're addressing? Yeah, I actually believe in this very deeply. So there are multiple types of collaboration, of course. I mean, the most obvious one is where you're collaborating with engineers

and where you can sort of think about they have some very hard problems to work on and they're not sure exactly how to go about it. And so you understand what they're trying to do, and then once you've understood that, you try to help, and you're sort of collaborating with them. And so this is sort of the typical research engineering collaboration. But if I step back, I think in research, we have to collaborate with more than just the product group guys. Collaborating with, believe it or not, the government is really important.

Government officials don't necessarily know what is doable, what is not doable. So you have to sort of work with them for them to understand the aspects of technology that can compromise people and the ones that don't. We collaborate with academia. I mean, this is the source of talent that we are to hire. And they have to train these people to work with us. So then, you know, these brilliant people that we work with, once they understand the problems we are working with and their students come and work with us as interns, that's another form of collaboration. And then a fourth form, I would say, is just the businesses. I mean, a lot of the stuff in technology is actually creating new business models, right? I mean, if you think about Google, I mean, one of the big success was not necessarily

the search. They had a breakthrough in the business model. So yeah, I mean, if you really want to be successful, you've got to work with all these different individuals, I think, or individual players. Let's talk briefly about network systems writ large. What are the historic challenges of network systems? And how has the landscape changed with the advent of the cloud? So I've been doing this for a long time. So I guess I should be able to say the world that I walked into after getting my graduate

degree to the world that exists today is different. But it's way more exciting now than it was then. So when I graduated, there was no cloud, right? There was this, there was internet. And even the internet had not really picked up. there was no cloud, right? There was internet. And even the internet had not really picked up. There was no web, for that matter. The web happened after I started working. And I was working in wireless.

The telco industry was taking off. Cellular networks were big. And we were starting to connect to one another. Our generation connected the world. And the next generation will figure out what. So in terms of connecting the world, we've gone into many dimensions here. But you asked specifically about the cloud. In the cloud, it's all about scale.

I mean, we have such massive scales today. You know, we have something of the order of 54 regions. And these regions have multiple data centers. We've got over 150 data centers around the world, and they are getting further built. We have tens of thousands of miles of fiber in the wide area. We have hundreds of thousands of miles of optics inside the data center. So now think about these millions of components that are connected to one another, and you rely on them so much, you don't want them to fail ever. That's the problem, right? We have now created these complex systems and we rely so heavily on them. And we have to constantly figure out what is the discipline

with which we design, we build, we add more features so we can do well. So I think the world earlier was a little bit simpler, but it was beautiful because we were discovering new things. Now it feels that we have discovered a bunch of things, but now everybody wants them. And so now it gets very, very complicated to try to make sure that everybody gets it. I love how you describe my generation connected the world. Now you guys go figure out how you're going to keep that going. It's very true. In this giant scale. Every generation, this is our generation's gift to the next generation. It is. Victor, you have some great stories from your career here. There's one I'd really like you to tell our audience today, and it's the one about

how the first wireless hotspot came about. Right, right. Oh, how many years ago was that? That is 1999. So I had graduated. I got my PhD while I was working, and then I came to Microsoft Research. I was this newly minted PhD, and I wanted to do something great with my life.

And so we were in this building, and I said, well, we'll create a Wi-Fi network. At that time, there was no Wi-Fi, by the way. There was a standard. This was called 802.11. And I wanted to create a lab out of this building that I was in. So one of my managers was very supportive of that. And so he gave me the green light to do it. So I created that lab. And I put 802.11 in that building. So now this was also the time when 3G networks were starting to come. And in the media, everywhere you read, it all talked about 3G. They talked about that's what was going to create ubiquitous coverage. Bill Gates was the CEO at that time, and we showed him what we had. And we convinced him, and he made the decision to expand this network that I've created to all

the buildings in Microsoft. It was about 35 buildings at the time. And that was a big deal. There's nobody in the world at that time who had this, right? So it was Bill's forward-looking vision that he made that decision and made that happen. But then that lighted up everybody's eyes. And then other companies started to look at, oh, maybe we should do it. So as I was thinking about all this stuff, I thought, hey, I got this going here. I can probably do it in public spaces too. I can go to malls and I can go to shopping centers and I can go to airports and I can be connected. I don't need the 3G network. I mean, this is all indoor technology. All I need is a way to authenticate people. And whoever's put the network there, they can figure out how they want to charge. Either they can give it free or not.

So I wanted to do this for real. So I went to Crossroads and said, your shopping center would be the first Wi-Fi hotspot in the world. They had this bragging rights that they were convinced of. So then over a period of a year and a half with another student of mine at the time, he and I, we built this up. And this actually ended up becoming the first Wi-Fi hotspot. Well, before we get into two big topics that I want to cover with you, let's talk for a minute about the breadth of work that's coming out of your group. And there's some really exciting stuff going on.

Fantastic projects, for example, in live video analytics. So give us some highlights of the research that we're seeing in this area. Yeah. Just kind of high level. And what research from other groups has really enabled those products to come to life? Yeah. So I like this project for a number of reasons. You know, you asked me when we began why I like this job or why I think it's one of the best jobs.

So this is a great example of tech that can be good for business and it's good for society, right? So the scenario is straightforward. You and a lot of people get, all of us get stuck in traffic. And we want to not be stuck in traffic. That is just wasted time. So we started looking at the cameras, right? And then we thought, well, we've got all these cameras that are taking this video of traffic flowing and things of that nature, but they're not doing anything active with it. They sort of like take the video and then later on you analyze it. So that got us going in that space. Now, there are many technical problems here, but I think since you asked me to stay high level, you should think of it this way. In Beijing,

there are 20 million cameras. And in London, there was a camera for every seven or eight people there in the cities. If you go to New York, you see cameras. And the question that you ask is, okay, who's looking at the video? I mean, when something bad happens, you go back and look at the video, right? And so in this day and age, why can't I build technology where I can do vision research? There's been a lot of great work that has gone on in computer vision. And so we decided that we could take some of that work, we could build these systems and roll them out to try to take advantage of these streams and create workflows that would have an impact. I would love to spend an entire hour with you just on live video analytics, but we've got to move on.

Right. Because I can't let an interview go by with you without talking about the dynamic spectrum, the spectrum opportunity, TV white space. You were there pretty well from the beginning. I was, yes. So in broad strokes again, tell us how the TV white-fi came about. Sure. And how it's developed over the past couple years. Yeah, so you're right.

I was there from the very beginning, actually. My story on that starts the year 2003, when Craig Mundy, who was the CTO of this organization, took me out for dinner. I was again a young researcher and I didn't know what it was about. He started talking to me about this cable DSL duopoly. He started to explain to me how in rural America people had no access in the internet. The duopoly concept was about the fact that everybody was relying on these cable and telcos to provide them connectivity. And they would provide connectivity where the business model made sense. So as he was explaining this to me, he started to say, well, Richard, we must be able to do something in wireless to prevent that. And then he said, so are you going to do this? This is exactly how he said, are you going to do something in wireless to prevent that. And then he said, so are you going to do this?

This is exactly how he said, are you going to do this? And so here's the CTO of Microsoft Research at that time asking me. So that got me going on this stuff. And then this amazing thing happened that we discovered that this new spectrum was going to become available. And this was a spectrum that was reserved for digital TV. It was being used for analog TV, and it was going to be available. And this was spectrum that was reserved for digital TV. It was being used for analog TV, and it was going to be converted to digital TV. And the rule of thumb there is that digital TV will use one third the spectrum of the

analog TV. So that means you've got two thirds spectrum left over, right? This is the frequencies that are left over. Of course, the telco companies came and said, we want that. We'll pay you for that. And then we went and said to them, if you give it to the telcos, you'll have the same problem. All these millions of people in the US, and even not US, just billions around the world,

which will again be left out. You would not have solved the problems. So what they did is, even though everybody was broadcasting digital, they started putting TV stations in analog space to not to lose that spectrum because they know they couldn't sell it, but they didn't want to give it up too. They knew this was a major resource. So this is a serious problem now for us. So the technical problem then became, can you now sense the frequencies?

If nobody is using it, just use it. So that is where we created this notion of opportunistic networking. And we went after that. A lot of papers written, a lot of press, nobody was really doing the work. But we took on the work. We actually built a system like that. This was cutting edge. Nobody had ever built anything like that.

We brought the FCC chairman. I gave him a two and a half hour sort of a lecture on how these things work. I showed him the network on Redmond campus and showed him the demo. He and his colleague asked us all kinds of questions, but we sent them convinced. Five days later, they voted to say, yes, unlicensed Spectrum is allowed. Nice. And then so this was a big, big win for us. And the final thing I would say is that Microsoft now has an airband initiative where we have promised to connect 2 million people in the United States by 2020 based on this technology.

And in Africa, we've done the same thing. Hundreds of thousands of students have come online and they've written. I've gotten mail from them saying that this has changed their life. They actually said God should bless us because how much we had changed their life. Wow. So yeah, this is a big story for us. Well, and also, if listeners want to drill in a little bit more, I did a podcast

with Ranveer Chandra, who explained in more detail the dynamic spectrum, TV white spaces. All right, a lot of talk lately about edge computing or the intelligent edge. Can you give us an overview of what it is and why it's important and what are the big research challenges for edge computing? Today, cloud vendors sell you storage and they sell you compute. I want them to sell you latency as well. So that's how we conceived of edge computing.

It was actually conceived in 2009. So the idea of edge is that you have many applications which are very latency sensitive. So edge is very, very important when computing has to be very close to the data, but then the devices are not, you don't want the devices to be very big, they're just very small. So when I thought about edge and conceived it, I thought about edge in conjunction with cloud. And what has happened in the last two years is the marriage of the edge with IoT. Any sector you look at it, it's all about IT. It's all about automation.

And so it's all about sensors. And so now they have to connect to some computing device, and that's your edge. That's your intelligent edge. So I think in research, there are many problems that still need to be solved. All the problems we solved in the cloud actually apply to the edge too. Things like availability, reliability, security, all that has to be solved in now this domain. And the assumptions that were made for the cloud are a little bit different from the edge. So where does that live in the research organization? What kinds of things are you working on to make that a reality? Yeah. So the biggest thing is about reliability. So,

you know, if you're going to create a dependency of your business on the cloud, you need to make sure that the edge is always available. So I think one of the problem is that in the cloud, you've got hundreds of thousands of machines and we know how to make availability. If something goes down, we'll just switch you over. You won't even know. Now in edge, we don't control the system because it's sitting on your premise. It's some heterogeneous piece of hardware. So we've been thinking about how do we give you the same sense of security that you have

in terms of the availability of the system as we have in cloud. So that's one problem we're working on. Other problems have to do with security. I've been talking to some researchers yesterday about there's this problem with physical security. In cloud, we have everything contained. It's locked and key. People are watching. You know, you put the edge here. Now you can have physical security problem where people can actually turn some things off, take the disk out, read it, put it back in. How would you ever know this has happened? And one of the things that always came up was that we're building these data centers so quickly. Why do we need these edges? So one of the things

was to just build applications that convinces people that you can do this without cloud. So we've been doing that as well. You know, I talked to Ben Zorn a while ago, and we talked about the Internet of Things and how, not unregulated necessarily, but I could make a device and sell it to you. And I don't have all of the same fallbacks that I put into it. So you've got a smart doorbell, you have a smart fork connected to everything. And it's like a host of problems. Yes, it's very serious if you think of it. So we all like this stuff.

But when things go down, you need to fix this. What are you going to do, right? Who are you going to call? Who are you going to call? Who are you going to call? Ghostbusters, right? Anyway, but seriously, though, so I think the challenge for all of us is to actually build self-sustaining systems, self-managing systems. But I don't say that too much because when I say that, it always tends to take people in the direction of, oh my God, we are going to the Matrix or we're going to something else. Yeah, see, there's a whole question I ask all of the researchers is what keeps you up at night, but we've kind of covered that on that. So let's talk about a big thing and kind of live there for

a second. The big announcement this week of the Open Network Emulator, which seems to be predicated on the underlying assumption that the cloud is too big to fail, to use a phrase. Tell us about Open Network Emulator. What is it? Why do we need it? How did you build it? Yeah. The Open Network Emulator is, once again, now it's going to become a theme in this podcast about reliability and availability. Let me put it in perspective as to what it is. As the name suggests, it's an emulator. It emulates networks. And we've known that for a long time, how to do that. But it's been very hard to emulate cloud-scale networks, right?

These are very, very large, as I've been saying again and again. The larger story here is verification, network verification. And the way network verification ties into availability is as follows. So let's say everything is working perfectly well. Barring hardware failure, everything should be fine. But then somebody who is part of your team goes and changes something somewhere. And I have horror stories about that that I can tell you, but I will not. But goes and changes something at some point. And that thing, as has happened in the past, can bring down an entire region. Because if you break the network, your packets are going nowhere and not going to the right places.

Let's put it that way. And that's a complete no-no for businesses. So this thing, talking about what keeps me up at night, if something happens somewhere, things will go bad millions of people will get impacted and i don't want to be the source of that right so the emulator was built to prevent that and the way the emulator works is that now network engineers and operators when they do changes they make the change but they actually don't even know if they're making the changes in the network they actually are changing the emulator

because it mimics the the network underneath so amazingly that you can't tell the difference. So once you make the changes, the emulator will then try them out and ensure everything is good. Once everything is good, it's going to go and put it in the network below. And voila, if we did the work right, it should all work. Now this is how we achieve availability, and it's been phenomenal. I'm not going too deep into technical details here, but there are some major technical problems that we needed to solve to make it that way.

And so now we have decided that this is such an important resource for everybody that just hoarding it ourselves is not the right thing to do. So we are making it available to the entire community so that they can now. And it's not just for production systems, but also for students that are now graduating. Because it emulates the network so well, they can actually do some amazing research without having major data centers available to them. So we just think that this will be very valuable to the entire community, both researchers and industry and governments also, you know. And so that's what we are doing with the Open Hub together. As a director of a large group at Microsoft, you have to have some skill in dealing with people. And when we talked before, you told me a little about your own management philosophy, particularly

as it relates to how you inspire, shepherd, and even protect the people that you're working with. And you said that it was sort of anchored in the way that you were treated when you were a young researcher. So tell us about that, what you think. Yeah, so you have to grow as a manager. And what I mean is, let me start by saying that what worked then doesn't seem to work now. So you have to continuously evolve. There is no fixed style. I mean, there are certain core things here. That means respect for the individuals, making sure that you don't smother them too much in the sense of like they can't even grow. But as to how you deal with people,

I have realized it's generational. So when I started doing this, certain set of things work really well. As new set of people are coming in with a different sort of outlook towards life, those old things don't work anymore. So what the people who are in my group tell me a lot is, Victor, listen. So I try to listen a lot.

But generally, I think that when somebody comes into my room with an idea, they should leave feeling great about the idea. Because I have enough humility to say that I don't know it all. I can express opinion, I can give advice, but I don't squish any idea because I feel like again and again people have surprised me massively about what started off and what it became because they are just great. And I think that part of your job as a manager is to provide enough air cover. So the other individual actually believes that you support them, you will go to bat for them.

So you can have a little bit of discussion here, but once I have said I would go to bat for any of my guys at any point. Just briefly, give us the story of Victor Ball and Microsoft Research. How did you get here? What was your path to Microsoft Research? Yeah. So I was like any other immigrant students at that time. I'd come from India.

I came here when I was 18. I did my bachelor's and master's, and then I got a job at Digital Equipment Corporation. And I was fortunate enough to land in a group where there were very smart people there too. Most of them had PhDs. Most of them had graduated from MIT at the time. So anyway, being around PhDs sort of convinced me that I had to do a PhD myself. And so DEC had an amazing program at that time. They had a program called JEEP, which was for Graduate Engineering Education Program. It was really for educating people who had come with bachelor's to master's. But every so often, they would take

somebody who had master's and fund them for a PhD. I was actually selected for that. And so that allowed me to go get my PhD while I was working. When I came back, they gave me a directorate role and wanted me to lead that team. But I had been doing networking and that's where I wanted to be. So then my brother was in Microsoft. And so he said, come here and at least you'll get to see your niece. So then I came here for an interview and I met Rick Rashid, who was the head of Microsoft Research at the time. And they said all the right things. And it sort of like felt to me that I had this amazing opportunity. And, yeah, so then I ended up here. And then I was, there was a funny story there, too. I ended up in a group which was called Operating Systems.

So when I ended up, I was doing networking. So I said, hey, we got to change the name of the group because I do networking. I mean, you know. So they were thinking about it. And I was in a room with like so many people, I guess. I was the only one guy. So I said, networking and systems. And then I was voted down seven to one, systems and networking.

Are you kidding? Because they were all from systems, and they didn't want networking to be first. That's how it could have been networking and systems everywhere instead of people saying systems and networking. Well, as we close, what do you see on the horizon for the next generation of networking and systems researchers? What big, exciting unsolved problems might inspire the next group to be the next Victor Paul?

Well, I don't know about the next Victor Paul. I'm sure they'll be there on South. But I think we live in great times. It's so amazing. So I already mentioned to you things around cloud and things. But what is exciting beyond all this cloud stuff is think about what other variable devices that are coming along the line. I'm wearing this watch here, which tells me a lot of stuff. This is just the first version of what we're doing. So I am not able to predict what precise world we will go towards. I have some broad strokes there of that, but

what I'm convinced about is that the set of problems that are coming our way and the kinds of things we're going to do are just going to be phenomenal. I mean, if I was to be born again now, I wouldn't mind getting born right now. So it's just a great time to be alive, yeah. Victor Ball, thank you so much for being our first live podcast and being such a good first live podcast. It's great. Thank you very much. Thank you for doing this.

Okay. To learn more about Dr. Victor Ball and Microsoft's vision for the future of our networked world, visit Microsoft.com slash research.