 I think we're going to get started. I think folks will still probably straggle in over the next, I would guess, about 15 minutes. We'll see a couple more bodies, but we can get started if everybody's ready. Okay, everyone has their last coffee? Okay. So for people who weren't with us yesterday, I just wanted to explain the format. So for the breakout sessions, the people who are running breakout sessions will come up and pitch their talks, and then we will tell you where they will be, and then we will break for those breakout sessions. And that is the majority of the day-to-day, our breakout sessions. After lunch, we have a future of the internet keynote, or internet, if you want to call it that, from Malka Older, who will be joining us later today, and with her keynote we'll do actually a workshop as a full group, I think, in here, which we will share more about later. Other things, any questions about logistics of the space? I think people know now the bathrooms over there, there will be food, there will be lunch later. Please eat, switch sessions as you need to, and feel free to come find me and ask questions if you have any. I'm Georgia, by the way, for those of you who don't know me. Okay, so with that then, can I ask, I think Yaa and Chris are giving their pitches. One change to this morning's program, one of the sessions, which Shazna was doing on linear Asia's research, is going to be in the second session in combination with research ICT Africa. So there's just two sessions in this first morning breakout, so, Yaa, do you want to come up? There you go. Okay. Hi, my name is Yaa Chang, I'm a software engineer with the Google New York City team. I'm here to talk about penetration of data, the history and the future, how we process the data, how we monitor the data processing, and how we will improve the data usage using new binary and new schema. So, this is a brief history of Patrick's Out that's currently running on MLM platform. The original paper was written in 2006 by a teamer's team from Paris. So the name is Paris Traceout, which sent a reverse route probing request from the server to the client and tried to get the pass as much as possible. So it resolved the several critical issues like diamond pass, how to make sure the pass is crossing the different areas and record the time and the protocol. So it's wildly recognized as an efficient transfer out algorithm. We start to collect data from 2013 till now. So the output of our data published in two kinds of format, the raw data stored on Google Cloud Storage on this address for data before 2017 June. And after 2017 June, we move the new data to a different bucket. The ETL pipeline will pass the data into schema format on BigQuery table and available for search. So this is the BigQuery location for the Patrick's Out data. Now our platform is 100% open source. You can reproduce all our data using the open source code we published. Here is a brief figure describe the amount of data we have collected for more than half a year, more than five years so far. Totally we have more than one billing test in our BigQuery table. Each hop will become a separate role. So we totally have roughly 18 billing roles in the BigQuery table. And this is the older version, but it's fine. We have roughly eight billing roles added for the first half of this year. And we expected to aid more than 17 billing roles in 2018. So the data is considered increased. We have a huge amount of useful information that will be super helpful for all kinds of research. And I will give an example of how researchers use our data in their work. And the future PT data usage in my session. If you are interested, you can come to see my talk over there. Chris, who's giving the devices one? Chris, you're right there. Can I see your hand? That's fine. Thank you, Georgia. Wow, that's exciting. I wish I could come to your session. I'm excited. I was just talking to Peter. That's a very rich unexplored data set in Paris for researchers who might be interested in that. So I am doing a talk this morning with Ross Shulman, my colleague at OTI in the back there. And Simone Basso is over here on the side. And we're going to talk about device-based measurement. What is device-based measurements? Well, you know, one of the things about I think that some people complain about NMLAB data is that the crowdsourced nature of the data is sort of not like structured research. But it's possible to do both, and I think there's value in both. And device-based measurement is sort of what we're pursuing along the lines of that in various ways. So we're automating measurement from small computers, running code, consumer devices, so allow different researchers to run test data in a structured way, collect a data set, use it for specific research using the NMLAB platform, so taking the non-crout source approach to a more research-centric approach, standardizing the hardware. So I'm going to talk about that. Sorry, I'm just trying to hit the, see the screen here without glare. You can walk closer to it if you can use this one. Oh, thank you, Jordan. All right. Yeah, and it just, there it is. Sorry, yeah. The brain cells aren't really, synapses aren't really connecting just yet. Coffee will be easy. We're going to talk about devices. I'm going to talk about running research projects with devices. Ross is going to talk about, I'm going to talk about doing that using like an IoT-style platform, so automating the deployment and management using a platform called Resin. And Ross is going to talk about Markami, which is a sort of stand-alone version of that, which we were hoping will eventually be kind of the thing you can download and use at your home. And then Sor Simone is going to talk about the software that he has written called Measurement Kit, which is the library that undergirds both of our Ross and I's work on these projects. Measurement Kit is a software library that basically abstracts a lot of measurement tests and allows embedded device developers and others to just compile that code and run the binaries and collect the data.