Google Cloud NEXT '17 - News and Updates

Building a video conferencing culture (Google Cloud Next ’17)

NEXT '17
Rate this post
(Video Transcript)
[MUSIC PLAYING] SERGE LACHAPELLE: Hello, everyone. My name is Serge Lachapelle, and for the past 10 years, I've been a product manager at Google. I'm based out of a country called Sweden, which apparently is very popular around here these days. [LAUGHS] I've had the amazing opportunity to be part of a journey, and– throughout these 10 years. And this journey, of course, wasn't done only by myself, but with many colleagues that are here in the room with me today and for the duration of those 10 years. So for everyone, please wave. There you go. And some– I'll invite them upfront, those that can help in answering your questions later. So a bunch of the people here in the room have helped build what I'm going to present here today. Let's take a look back at 2007, 10 years ago. A decade feels like a longer time than if you say 10 years. And can anyone here in the room tell me something that happened in 2007? Do you remember? The iPhone. Anything else? AUDIENCE: George Bush.

SERGE LACHAPELLE: Yeah, George Bush, another good one. My son reminded me, it was the last Harry Potter book was published then. But what I thought of as well was the iPhone. And if you look at the iPhone today, you can see that a lot of the design aspects of the iPhone in 2007 have kept, which is a true hallmark of the design, but we all know that behind the scenes, the services that power, the infrastructure that power all the software, all the hardware, has changed tremendously. And that made me think of Google in 2007, when I joined. And this is how the home page greeted you in 2007. And the font's a bit dated. The logo, we know, has changed. But if I was to switch to 2017, you'll notice that even the same words are used– Advertising, business, About, Privacy Terms, Settings. The layout– the I Feel Lucky is still there, the microphone's new, but things have– things have been fairly consistent. But as we all know, Google, behind the scenes, has changed a lot, as well, in 10 years.

And if we look at how Google was in 2007, well, Google had just gone from being 3,000 employees in 2004 to 17,000 employees by the end of 2007. So when I joined, I joined at a– just at the cusp of a period of extreme growth. And today, Google is about 71,000 employees– that was end of year 2016. And what was different about how Google grew was that Google grew globally. And we didn't just open offices in countries for having a sales presence. We opened offices to have an engineering presence. And so back in 2007, we already had 42 offices across the world, and not just sales presence in the countries to do local business, but engineering presence. So Google was one of the– is one of the few companies that really invested in a global engineering presence from very early on. Talent was global. We wanted products to be made closer to our users and have that input, so it's really helped at building a very diverse engineering workforce. So there we were in all these offices, company growing, and everyone very ambitious to build great products and to transform the world, and to change the world.

But we couldn't do it the traditional way. The traditional way of doing global engineering means very heavy processes, very long planning cycles, things that you can't change on a whim. And maybe I should have showed concrete instead of a bunch of wires tangled. The word "process," in my head, always means this. But we wanted to be able to work globally, but still be creative, be innovative, be agile. You will hear that word often in engineering. We tried these guys– we tried the phone. Every employee had a phone bridge number and a PIN. We hated it. It was bad quality. It was hard to know when it's your turn to speak. There's no queues. How do you know when it's your turn to come in? You have these long, awkward silences, and it was just too hard to understand people, and it didn't promote creativity at all. And then I took a snapshot of my inbox last night, and– maybe I don't need to explain too much here, but email is great. We overuse email. We get overloaded by email.

It's sometimes hard to get answers quickly, and it's also very hard to understand each other. And when you're a global workforce, spoken speech, written speech, tone, what you actually mean, how people perceive what you mean gets tricky in written form. And so while we got far with email, it didn't carry us to our goals. So one thing caught on at Google. And very early on, video conferencing seemed like a great way to distribute a workforce. It seemed like a great way to promote agility, creativity, and to build that team feeling. So Google's founders, really early on, decided to go for it, and they went for it in a big way. Google started to deploy Tandberg's video conferencing units across the world. Here's how a room looked like in 2007. This was a typical Google room. This one was from our Zurich office. And you can see the mess of wires on the table, the camera bolted to the wall. All of the technology was in the black cupboard there on the left side of the screen.

I don't know what was in the fridge next to it, that little gray fridge, but let's just say it was water. [LAUGHS] But a room like this would cost anywhere between $40,000 and $80,000 in equipment and installation costs. Remember, every time someone has to come up with a drill, that's when the installation costs start to climb– to climb heavily. But we had a lot of these units. By the end of 2007, we had over 2,400 rooms like this around the world. And these were powered by infrastructure– complex, on-premise infrastructure. And this was not for the faint of heart. On the left side of the screen, you'll see a MCU. This is where all these rooms would connect into. Every one of these– an MCU, by the way, is a Multipoint Control Unit, but it's really a bridge. It's something that mixes all the video and audio together. And we had a massive mix of these MCUs, these gatekeepers, these failover gatekeepers, these border controllers, to allow meetings with external, outside of Google, meetings.

And as I said, this was not for the faint of heart. Every one of these MCUs could connect 480 rooms dialing in, each room using up one port. Each port cost $4,000 if you wanted to have 720p video. We got really good at this. Imagine being able to deploy 2,400 of these rooms– the world's biggest Tandberg deployment, not only in the amount of rooms deployed, but the amount of minutes used. The systems were used around the clock. We used them to have team meetings. We used them to have our Scrums, or our agile standups. Using video conferencing informally became a part of our culture because these systems were everywhere. And yes, you even got to meet your boss, who was around the world somewhere else every week, and look him in the eyes, or look her in the eyes, every day, if you wanted. So slowly but surely, VC made Google more nimble on a global scale. And then we broke it. We had to go and ruin a great thing. We ruined it because it couldn't grow with us. We had reached the maximum capacity of what a system like this could do.

And we tried, but people wanted to dial in from home. The software that we used to dial in from home– it was written by the same people that had written those conference rooms. Those conference rooms were connected with dedicated ethernet. They had a dedicated part of the network. They had quality of service monitoring. When you're at home, on your home Wi-Fi or your DSL line, the conditions are completely different. So software couldn't just– just couldn't connect properly. The calls would get dropped. The audio, the video were horrible. Everyone dialing from home would use up one of those $4,000 ports. So then you'd have 15 people dialing in from home, and if there's no more ports available, it means your 20-person room that you just paid $80,000 to install couldn't connect to the meeting, because all the home folks would compete with them. These were real issues. I remember these bridges– I remember one of them catching fire. And we knew– everyone knew how expensive those things were.

Everyone knew that Google just had meetings, after meetings, after meetings. So I remember– this could be a legend, but I remember a Google field tech having oven mitts, pulling out a card that was on fire to save the bridge, instead of going at it with an extinguisher. That's dedication. And again, if one of those cabinets filled up, and you only had four ports left, and you had a 10-person meeting, things would fail inexplicably. I also remember having to have engineers from the various companies we were buying products from on site in California. And they were at our offices for very long periods of time, probably charging by the hour. And I'll always remember them, because they were always the ones renting the fancy rental cars at the airport– you know, the big orange– or the big orange Mustang that's always there? And it says, it could be yours for $150 more a day. These engineers that we were flying in to help us, they would always drive the fancy Mustangs. And they would work, and work, and work at trying to get our system to scale.

And we couldn't get there. So what happened? We decided that it was time to innovate. So Google– back then, the management team was called the Operating Committee. And they made an important decision. They said, we're not going to buy this stuff anymore. It's done. It's over. Google is going to go from 17,000 employees to much, much more, and we're probably going to grow quicker than what these systems can offer anyhow. We need to build at scale. And this means that the pressure was on. So what follows is a bit of storytelling about how different innovation tracks happened– some simultaneously, some serially, and some building off on top of each other. The first one was to build for scale. So here is video chat in Gmail. Let's see if we can get some audio. I did this video in November 2008– took me eight hours. [VIDEO PLAYBACK] – Ah! There you are. Hi, I'm Serge, and I help develop Gmail voice and video chat from Sweden. Our team works closely with the team in the US, so video chat comes in really handy.

Some things are just better communicated face-to-face, where you can see expressions, gestures, and I can point things out on a diagram. Plus, whenever I travel, I get to see my kids. So let me show you how this works. Gmail voice and video chat is integrated right into Gmail, so it's very easy to use. To start a voice or video chat, simply click on the "Video & more" link at the bottom of the chat window. When a person receives a call, they'll hear a ring, and they'll see a notification that they can choose to answer. You can pop out the chat window and adjust the size and positioning as you like. You can even make it go full screen through a single click. It's free and easy to use. Visit to get started. [END PLAYBACK] SERGE LACHAPELLE: If ever someone asks you to volunteer for something like that, think twice. [LAUGHS] What's important here– so I had been working with software video conferencing for 10 years. So I started around '97.

And we had a small company, and in 10 years, we were able to get 200,000 users, and we were all proud. When we launched this thing, the 200,000 user mark passed in a few hours. It is amazing, the scale at which Google operates. And having all these users, with all these different firewall settings, and all these different networks, and all these different cameras and audio card, and trying to support all of them was an amazing challenge. But we learned a ton from it. So building at that scale helped us understand how to cater to many, many, many scenarios. Let's see if I can go forward here. Oh, I'm on– sorry about that. I need to click on the white area– perfect. Oops. So we started with Gmail video chat, and then we went to– oops, the video started playing in the background. Please don't make me hear myself again. [LAUGHS] And this was all fine– one-on-one conversations– but we wanted to replace the Tandberg experience. We wanted to give group video meetings. So we have an engineer in Sweden and a UX designer who was based out of our Kirkland office, Jake Knapp– they had been thinking a lot about this.

And it was pretty obvious very quickly that we wouldn't be able to do full-blown group video conferencing inside of Gmail, so we did a little– we did a little sandbox play area where we started to play with the software in different ways. And these two individuals came to the epiphany that, hey, why don't we just tie and actual meeting, or a meeting room, to a URL? And so we built a prototype, and we called it go/meet, and we'd say, hey, let's go meet in foo– go/meet/foo. And that caught on, and people thought, hey, this is neat. Great. And this led to something called go/meet that we were using internally at Google in 2009. And this is another prototype. Arthur has not aged– lucky him. And you'll see, some other people on that screen are in the room with us today. And we launched this to Google in 2009, and it became– it picked up. It became a big success. People could work from home. The latency was a lot lower than those video bridges that had to transcode. There was no transcoding in this system.

And people could take meetings from their desk. They could take meetings from home. They could take meetings on the road. And so there was a moment where this started to grow quicker, and the Tandberg rooms were being used less and less. And this became Google Hangouts a little bit later on. One thing that was different with this back in 2009 is, video was on by default. You went into a meeting, go/meet hangout, or go/meet/foo, and video was on. And that was an important pillar for this software experience, because if all these people would join, and you'd all be sort of avatars, and be audio only, who's the first person to dare show themselves on camera in front of all these others, right? Who's that conceited person that's going to be the first one to say, hey, I'm so great, I'm going to show myself for everyone. And so those systems didn't work. The ones that didn't have video on by default, straight from the start, they don't work for building team culture.

This created a sense of team. And after a few times, you don't care about your hair or the mess the kids made in the background, because this is your team. This is your safe space. These are the people you create with. This is a people you build software with. Our second innovation track was about the stack. Building video communication software was a real pain back then. You had to license third-party software. Not just surf third-party technology, but a third-party audio stack, a third-party video stack. And then you had the fun part of taking two different technologies from two different companies, put them on top of your own network stack, and then fit this into a plug-in. Remember, the plug-in architecture is called NPAPI. The N stands for Netscape. Just, this is something people forget. And you needed the plug-in for Mac, Windows, and Linux. And you needed to have a great flow for deploying these plug-ins, but also for keeping them updated. So how do you update a plug-in? How do you download the plug-in so it doesn't affect the call?

How do you update the plug-in so you don't tear down a call that's going on? And how do you do that in a browser without the actual software page being showed up, because that's usually where these things would get triggered? Very tricky, and with one of the best install flows that you could find on the internet, we still had first attempts to join a meeting failures of about 30%. People just cannot– it's just too hard to download software on your desktop, and find that file, run the installer, complete the installer, remember to restart the browser, because this is a plug-in. People would fail at joining their meetings the first time. All this complexity of deployment, and having to integrate other people and other companies' technology, meant that innovation stagnated, and quality was a hard thing to improve, when you're not in control of your own stack. So a team at Google started a project called WebRTC. WebRTC is an open-source audio and video platform that developers can use to create video and audio communication apps.

It's available on Chrome, Android, and iOS, and there's other browsers that take the code and implement it in their own browsers. It's an open-source community effort with different standard bodies. It's highly secure. And in Chrome, this is a one-click experience. There's no more download. You click, and you're in your meeting. That helped a lot of users. And it's not a stretch of the imagination to say that WebRTC, today, powers video conferencing on the internet. If you use Google Hangouts, Google Duo, Facebook Messenger, GoToMeeting, Twilio, the list of companies is over 1,000 long, of companies that integrate WebRTC, the code that Google wrote and Google made available for creating web– for creating video conferencing experiences. So now we had started building software. Then we decided that we wanted to own the stack. We wanted to understand how audio and video works. We wanted to make that part better. But we still had Google Chrome growing like crazy in all these meeting rooms that needed videoconferencing support.

And so this is a video of two of my colleagues, Steve [INAUDIBLE] and Eric [INAUDIBLE] We bought a HP TouchSmart in 2008, and that's the one we bought. This ran Windows Vista. It ran Firefox and something called Greasemonkey, if any of you remember this. It would grab the video chat little bubble that you saw in that video, and it would make it big so that you didn't know Gmail was there in the background. And with the touch screen, it allowed you to punch in a meeting name. It cost $1,000, $1,050, instead of several thousands of dollars. There were no remote controls or devices. This was an all-in-one. And so this was our first attempt at creating a video conferencing room setup by ourselves. Oh, I will– [VIDEO PLAYBACK] – –with Eric [INAUDIBLE], who's also filming. – Hi. – Hi. We're working on the [INAUDIBLE] project to kiosk a PC to function as a Tandberg replacement. So I'm going to go ahead and power it on. As soon as you hit dial, it will go directly to this page that will show you a video conference that you will join in progress.

– This was one of the first times it all worked. – Let's see if anyone's there. Calling– – Ta-da. – Just like talk video. [END PLAYBACK] SERGE LACHAPELLE: And that was it. The actual meeting– the actual video was like 15 minutes long, because it was powered by Vista, and it took so much time to boot. But so, you know, the team did– there was so much work that needed to be done after this first proof of concept, right? We moved this to Linux, we redid the front end, we had to do fleet management on this so that we– because we wanted to deploy thousands of these. But that little video, those two individuals and the others that worked with them, led to a revolution at Google. We built a video conferencing appliance. And so this is today at Google. And all of our meeting rooms have video conferencing. And they're all built on top of the same software that I just talked about, Hangouts and WebRTC. It's powered by computers that run– that run ChromeOS today. We're going to talk about that in a little minute.

But we've got over 15,000 rooms today that have video conferencing capabilities. So this is one of the offices in Tel Aviv. Here is an office in London. This is a smaller huddle room. And this is how good we've become at choosing the cables, passing the cabling– the amount of discipline and thought that has gone through all of this is just staggering, because when you have to do something 15,000 times across 70 offices now, or 70 countries now, you can't let anything to chance. So what cables do you use? How do you pull the cables? What holes do you drill? What boxes that you choose to insulate sound from the PC? All those things are really, really important. And we put that ubiquity to use. So today, if you look at our usage at Google, we've got Googlers dialing in to a video meeting 230,000 times a day for a company that's about 71,000 employees. We spend 9.2, 9.3 years of time every day in video meetings. And this is not just meetings. This is, again, a team that is distributed, that wants to feel that feeling of togetherness.

It's people that want to have one-on-ones. It's teams that want to have their daily Scrums, or their huddles. And it's people that just want to have a virtual presence. Maybe they're working from home that week, and they want to feel close to their team, so they're just on video, and the rest of the team follows suit. And again, we've got over 15,000 of these rooms internally at Google. And then everyone has their laptops and their phones, and the same experience that they have in the room moves to the laptop and moves to the phone. That's a very important detail in this case, because you don't want to create islands of communication. If you have a different room system than what you use on your laptop and than what you use on your phone, when you try to bring all of these together, you create kind of separate islands of collaboration, and the lowest common denominator of the experience– you know, it just makes everything dumbed down a lot. And I don't know what happened, but you see, the company, since 2013, doubled in size.

But the video minutes in this system have gone up 17x. So as the company grows, you've got this kind of social graph that just explodes. And people want to share experiences, share knowledge, and just make sure that everyone is up to date. And the whole company can make decisions quicker thanks to it. And I think this is what makes Google one of the– a unique company, because we can do global engineering, and I think you can see it taking off now. There is a fourth track to all of this, and it's a little product called Chromebox for Meetings. The same team that built the beautiful deployments that you saw a few years ago understood that ubiquity was very important. They wanted to be able to scale things up to much more than 15,000 rooms. I've heard a rumor that there's expectations that we had an extra 20,000 rooms over the next two years to our system. And that team understood very quickly that having a computer, having to pass all these wires, maybe was a level of complexity that was too high to be able to scale up to those numbers.

So that team sat down and created a system that would also cater to our customers' needs, because our customers have as large deployments as we do. Some of our customers beat us in the amount of minutes they do per employee out there. It's incredible to see the power of ubiquitous video conferencing and what it brings to a company. A ChromeOS machine that is secure– there is nothing more secure than a ChromeOS computer on the market today because of the way the hardware is signed into the software and vice versa. A computer that allows you to do remote management of it very easily. And this is why Chromebooks are so popular in schools and in enterprise, because of this remote management capabilities. A computer that connects to your calendar, to your room's calendar, and all it does all day is it shows your room's calendar, and it's waiting for a click. Its just sitting there waiting for a click. And when someone clicks, boom, you're in the meeting. And that's about the amount of interaction it gets.

When you share– when you want to share a document to your meeting, you do it over Wi-Fi from your computer. You can mute yourself, you can hang up the call, and you can start the call. And everything comes from Google Calendar. So it integrates really well with G Suite. And we really believe that this is the next step to our ubiquity project. And we're working really hard at evolving this, not only for growing Google's video conferencing culture, but also bringing our customers with us on the journey. So expect a lot of great things on this front a little bit later this year. So finally, today, you could say that thanks to all this work, and all this innovation, and 10 years of sweat and tears, we've been able to support a video– a global engineering culture that fits into what Google is. It spans the globe. It saves time. It helps Google make decisions quickly. It helps people share ideas, help each other, and be able to try to reach some kind of work-life balance. If you need to work from home, you work from home.

There's no need to commute in every day. And now, this is a tool that is evolving based on your feedback and the input that you've been giving us as customers. And you might wonder what happened to our Tandberg systems. And when I was looking at this last night, I found a web page from a team in Australia that recycled them and created PlayStation 4 kiosks with them. And apparently, the shape of the PS4 fit perfectly in that cabinet, and they became very popular in certain countries as PlayStation 4 enablers. So on that, thanks for listening. [MUSIC PLAYING]


Read the video

Through the deployment of Hangouts and Chromebox for Meetings, we’ve gone from spending $150M on a proprietary video conferencing solution that still couldn’t connect all of our conference rooms to a fully connected workplace. With Hangouts and Chromebox for Meetings, virtually every room at Google that has a door is connected via in-room video conferencing, and the cost? Significantly lower than our prior solution. In this video, you’ll learn from Google’s internal IT team about how they’ve enabled and built a much more affordable and efficient VC culture within Google.

Missed the conference? Watch all the talks here:
Watch more talks about Collaboration & Productivity here:

Comments to Building a video conferencing culture (Google Cloud Next ’17)

  • Wow

    Turboleopard55 March 9, 2017 6:18 pm Reply
  • 👍

    Turboleopard55 March 9, 2017 6:18 pm Reply

Leave a Comment

Your email address will not be published. Required fields are marked *

1Code.Blog - Your #1 Code Blog