George Michaelson 0:10 You're listening to ping, a podcast by APNIC, discussing all things related to measuring the Internet. I'm your host, George Michaelson, this time on ping, I've got two guests from the Thai Internet research community, Sukumar Kitisin from Kasetsart University, and Adisorn Lertsinsrubtavee from InterLabs, which operates in the Asian Institute of Technology. Or AIT. AIT is well known to the Internet measurement community. It supports the AINTEC conference, along with organizations like APAN Thailand, the WIDE project and the APNIC foundation. Is chaired by Professor Kanchana Kanchanasu from AIT and the program committee has worldwide participation. We featured AINTEC on PING before talking to students and presenters at the Vietnam meeting and recently at UNSW in Sydney, when AINTEC met in conjunction with SIGCOMM. Today, we're going to talk about two different kinds of Internet measurement activity being done in the Thai Internet community. One is about the Internet itself. The other is using Internet protocols to measure the environment. First off, let's hear from Sukumal, who's discussing a classic Internet edge measurement exercise, which is aimed at providing information to help govern the communications industry. Sukomal is doing this research for the Thai Consumer Council. It's a consumer driven Internet speed measurement study. The activity is part of the ISOC "PIMF" model, which leverages the ISOC "PULSE" community. So today we've got two guests with us from the Thai Internet community. These are long standing members of the local Thai community from the academic world, and they've got two different stories that are both about measurement in Thailand, but they're really very different. Adisorn, Skukmal, welcome to ping. Hello, George. So could you each say a little bit about yourself. Sukumal, could you go first? Sukumal Kitisin 2:23 Hi, I'm Sukumal Kitisin, I'm a faculty member at the computer science department, Kasetsart University in Bangkok, Thailand. George Michaelson 2:31 That's KU University. Is that like a state a National University? Sukumal Kitisin 2:36 It's the state regulated university, George Michaelson 2:39 Right? And Adisorn, Adisorn Lertsinsrubtavee 2:41 Hi George. My name is Adisorn Lertsinsrubtavee: it's quite a long name. So currently I'm holding the position as a director of InterLabs Research Center on working on the research on Internet and education laboratories. Yeah, we part of the Asian Institute of Technology, AIT George Michaelson 3:02 AIT which we've all heard from, and I think we've had other guests from a it appearing on the program a great, long standing institution. So we have two different stories. Sukumal, if I start with you, your story is a classic Internet measurement story, isn't it? But it's not kind of like a university research program. This is something you're doing with another body. That's right? Sukumal Kitisin 3:24 Yes, I'm doing a research study for Thailand Consumers Council in order to measure the upload, download speed, sort of like speed test for mobile operators in Thailand George Michaelson 3:39 and the Consumer Council covers all kinds of consumer service. This is not just the telecommunications consumer council. [Skumal: that's correct] So they might be interested in water quality or in the price of bandages, from Boots the chemist. They do everything, but you're helping them assess mobile and WiFi services. Is that only in the city, or is it across the entire country? Sukumal Kitisin 4:02 We project to do the whole country? It depends on the volunteers that we get on board on the project. So far, our volunteers kind of scatter throughout the country, but the numbers are not massed yet. George Michaelson 4:19 Well, let's come back and talk about that in a minute. So there are two primary different ways. I hate false dichotomies. There's probably 100 ways, but the two ways we often see there's passive measurements, where you watch things in flow, and there's active measurements, where you put traffic in the network and you measure its behavior. Are you in passive or active? Or both Sukumal Kitisin 4:41 Our measurements is on the active side. We actually have our volunteer install programs like Ookla speed tests and also opensignal. George Michaelson 4:52 So open signal is measuring things like the 4g and 5g signal strength down low. You're doing that? Sukumal Kitisin 4:58 No, it's. On the similar the same level, as Ookla speed test, right? George Michaelson 5:05 So this actually is a well understood measure. I mean, at large in the community people can just take out their phone and do an Ookla test while they're on a BTS train in suburban Bangkok, or they can do it at home. So I think there's good acceptance from people at large. Is that why you chose this? Sukumal Kitisin 5:23 Yes, yes, that's correct. It's the two of the two of the popular applications that are used for testing the speed for the mobile phones. And we are so thinking about using the M-lab MDT as well. George Michaelson 5:43 So with M-lab, there's quite stringent quality concerns around public data and access to data, but that also means independent researchers are going to be able to come and integrate into this and reflect on the data. I think it's lovely for reproducibility. So you mentioned it's volunteer driven. Yes, it's volunteer driven. How are you recruiting your volunteers? Sukumal Kitisin 6:07 Thailand Consumers Council has their network with people within different regions, different provinces in Thailand, so they helped us recruit those volunteers on board, George Michaelson 6:19 Right? But that potentially has a degree of population skew, because obviously, with a massive urban center and then a rural large economy, you risk getting a lot of dense measures in Metro Bangkok and not enough in the smaller communities. Do you have, like a statistical model in mind that you will be using to adjust for this. Sukumal Kitisin 6:41 We just started it this December. George Michaelson 6:44 Oh, this is a new program. Sukumal Kitisin 6:46 Yes, it's a new program. So we'll have to see how many volunteers we get in each part of the countries, and then we can assess George Michaelson 6:56 so also people are people, and they generally occupy the daylight hours. How are you managing to collect enough data to get a balance across the full day? Sukumal Kitisin 7:07 Currently, we divide the time to collect data into three periods during the day so they use the applications and then tested it and then send us the result three times a day. George Michaelson 7:22 Yeah? And do you have a reminder framework to try to get them Sukumal Kitisin 7:26 yes? You use like [indistinct] in helping reminding them, [George: yeah] sending the notification, hey, this is the time you have to test it. George Michaelson 7:35 There's a meta measure here. You could probably get one of your students in the maths department to do an analysis of the Gaussian distribution of their measurement time against reminder, I really wonder how they would cluster. So you're using ookla, you're gathering classic bandwidth. It's both directions. It's both upload and download, isn't it? [Sukumal: Yes] and there are congestion issues in the modern network deployment people often observe in my economy in Australia, that when 3g was turned off, things got worse, because 3g had amazing reach. Are you in a 5g enabled network now? Sukumal Kitisin 8:11 Yes, yes, we are. George Michaelson 8:13 That covers quite a broad range of speed capability. So the down grade options in 5g are quite high, aren't they? Sukumal Kitisin 8:21 Yes, we have seen some of the traffic that switch from 5g to 4g in some area. George Michaelson 8:28 Yeah, there's going to be a lot of information in this. So how long do you expect this project to run? Sukumal Kitisin 8:34 It will run until June this year. George Michaelson 8:37 Right? And you'll be delivering a report back to the consumer Council. Sukumal Kitisin 8:41 Yes, that's correct. George Michaelson 8:42 And after that, what happens next? Does that then go to the Thai regulator, or is it just part of the general community? Sukumal Kitisin 8:50 I believe Thailand Consumers Council will use that result and reflect some message to the NBCT, National Broadcasting and telecommunication. George Michaelson 9:01 But that's maybe a more social policy and governance question than a measurement. You're in data science, and you're producing the information [Sukumal: that's correct], but you're going to publish some papers out of this, right? I mean, you're an academic, Sukumal Kitisin 9:14 we hope so. Yeah, George Michaelson 9:16 I really look forward to hearing some more about that. Sukumal, that's fantastic. Thank you. Sukumal Kitisin 9:21 Thank you for having me George Michaelson 9:23 Now let's hear from Addison Lutz in thrupp TV, who is working on the measurement of air quality across the South East Asian region. This measurement activity uses small devices that measure the fundamentals of air quality and provide data over low bandwidth protocols like LoRaWAN. This is designed for Internet of Things, low power devices, and it permits Addison's project to deploy sensors deep in the forests of Southeast Asia, as well as in domestic and rural farm locations, and provide it. Time series that reflects a persisting problem of forest fires and agriculture related burn offs. These and other sources of particulate and carbon monoxide air pollution have been contributing to a long term problem of haze. The project is called SEA HAZEMON, and it aims to provide early warning signals to help guide local forest fire authorities. It uses cloud services and several generations of sensor deployed in the field. so Adisorn, you're actually doing something that's really completely different to this. It's still a measurement project, but it's not about measuring the Internet as much as using the Internet to measure a real world problem. Is that? Right? Adisorn Lertsinsrubtavee 10:47 Yeah, so now I working on a project called SEA-HAZEMON Which was actually initiated by Professor Kanchana from the AIT, who is my mentor. And this project, we aim to build a low cost sensor for the air pollution modeling across the Thailand and also. In Southeast Asia as well. George Michaelson 11:10 This measurement. Have you acquired a device? Are you using some kind of commercially available sensor? Or have you been building a system? Adisorn Lertsinsrubtavee 11:18 We are building our own system by taking the microcontrollers and then we develop it on the electrical circuit board or PCB, but for the sensor part, which is using a low cost model in market, [George: yeah]. So we have been indicators or building the hardware part and also the software parts as well. George Michaelson 11:36 And the sensor that measures this. How do you actually measure this? Adisorn Lertsinsrubtavee 11:42 The PM sensor, we are using a technique called Light scattering, so in that sensor, they have some kind of a laser beam to push in that. And then when, you have particles in the chamber. You can have some kind of reflection. So that we imagine that in the sensor box, they have small box. Here they have a, you have a small chamber, [George: yeah] inside that. And then in that chamber we have a laser generator source, [George: yeah], put in. And then we also have the some kind of a motor to suck in the air from inside. So imagine that, if you have the air has been polluted with small particles, right? [George: Yeah] the light that the lesser beam need to put in can be scattered because of a lot of so George Michaelson 12:26 you mentioned the dispersion of the beam [Adisorn: exactly], and there's a function, a well known function, that models particle size, dispersion, [Adisorn: yeah], and particle density, Adisorn Lertsinsrubtavee 12:35 yeah, density, we more consider on densities and also the particle size as well. [George: Yeah], we convert them to the units, micrograms per cubic meter. George Michaelson 12:45 yeah. So we're all used to the idea of its asthma and coughing, but it goes well beyond that, doesn't it? This actually is a significant health burden. Adisorn Lertsinsrubtavee 12:53 Yeah, it's become like critical issues in our regions, because now we have a lot of patient, like, in statistics, right? Patient have been come for to treat, looking for the treatment for the disease, like, yeah, NCD [non-communicable diseases], lung cancers, [George: yeah], stroke, so and that related to the major source it came from the PM2.5, yeah, that you can inhale this bad air into your lungs and so on. George Michaelson 13:19 So I think you can also measure other things, like CO carbon monoxide, which is more directly related to engine pollution and forest fires. You're measuring that as well using gas measurement, Adisorn Lertsinsrubtavee 13:33 yeah, because in our goals objective, we're not only measuring the air pollution in the city, but we also try to use the sensor the platform to track in the forest fire and WiFi as well. You see quite a lot of measures. George Michaelson 13:47 So this is also part of fire management and understanding exposure of risk from forest fires. Adisorn Lertsinsrubtavee 13:53 And the CO CO2, as you mentioned, is stuff, but it's kind of the signal from the forest fire that's coming in. George Michaelson 14:00 So you built a micro controller based system. Is this running a Linux, Unix operating system, or is it micro controller like ESP32 what kind of coding is this? Adisorn Lertsinsrubtavee 14:02 Yeah, we have kind of mix. Because we have been, this project is running from September 16. We've been, we have a lot of versions of this George Michaelson 14:20 So this is a long baseline project, yeah, Adisorn Lertsinsrubtavee 14:22 and they still continue, George Michaelson 14:25 yeah. So you have several versions of technology deployed. How many probes Do you have deployed in the field Adisorn Lertsinsrubtavee 14:30 right now? We have been deploying more than about 200 nodes in the regions already. George Michaelson 14:36 This is across Thailand, Adisorn Lertsinsrubtavee 14:39 Across Thailand, and also we have been expanded to Indonesia, Philippines, Laos, and recently we expand to Burma, Nepal and outside as well. George Michaelson 14:49 Wow, that's a huge network. That's really impressive. So these devices obviously can't just patch into an Ethernet cabinet in the hotel and get 10 megabit bandwidth. How are you gluing these things together? Adisorn Lertsinsrubtavee 15:03 Good question, because since we are targeting [this system] to many parts of the world, right part of the region. So we try to collect the data by using a lot of means, like we are using WiFi, LoRaWan, [George: yeah], and also 4g cellular connections as well. George Michaelson 15:21 So these protocols, I mean, 4g and cellular is just part of lived experience. But LoRaWan is a little off to one side, [Adisorn: yeah] this is actually, in some ways, designed for small devices sending small bursts of data. You're not generating a massive stream of data at a probe. Are you? Adisorn Lertsinsrubtavee 15:40 sure. Because these sensor have consumed a very little bandwidth just sending some but this kind of politically send this more frequently, but, yeah, just consume a very less bandwidth. George Michaelson 15:50 And do you have an IoT orchestration framework to coordinate this system? Adisorn Lertsinsrubtavee 15:56 Yes, we do have some kind of, let's say that controllers, and that is based on MQTT actually, George Michaelson 16:04 right. And then you have a published subscribe framework for the data feed, and it's just. Generating a time series. [Adisorn: Yeah]. So when you deploy a probe, you have a latitude, longitude, you get positional. You know, if it's indoor or outdoor, the probe goes live. Do you have to give it like a battery backup to take care of power supply, or is it just plug into the mains and it goes Adisorn Lertsinsrubtavee 16:27 depends. So because in the department areas. So some area, we place this in the building. or house? So you can use like the stable electrical source. But in some area, like we want to our work application is for this fire detection, so we need to deploy them in the forest. George Michaelson 16:43 So there isn't a convenient WiFi point in the forest, is there? Adisorn Lertsinsrubtavee 16:47 Yeah. And then we use, like the solar panels and batteries and load up and LoRaWans come to the place as well to connect those things up. George Michaelson 16:54 right Because it's able to sustain data management with unreliable underlying substrate. So you're collecting this data, what then happens? Are you doing a data processing cycle? If it's been running from 2016 you must have a significant body of data, Adisorn Lertsinsrubtavee 17:10 yeah, isn't it? It's quite huge amount of data coming to our database. George Michaelson 17:15 Have you got a predictive quality in this, for instance, could it be an early warning system that shows you the movement of things? Adisorn Lertsinsrubtavee 17:22 Yeah, we do that for because we are doing kind of machine learnings and develop some kind of data modeling as well, because we wish the compounds parameters, and we mentioned that we try to detect some kind of behaviors of the burnings like the forest by burning it to provide some early warnings to the people who live in the field. George Michaelson 17:43 So you can actually close the feedback loop. You can tell people in advance, this is not a good time to be doing stuff outside. Perhaps you can even tell people cut the burning down and allow a weather system to move through so that you don't get the problem at scale. Adisorn Lertsinsrubtavee 17:58 Yeah, this is one form that we are providing some data for the public use they know, like what to say, the status, of the air pollution in that area. But we also have other applications as well, right, analyzing those data and then provide some early warning to a forest fire rescue team, because they can know in advance, like which location that has been a fight going on. George Michaelson 18:20 Is this an AIT funded activity? I mean, this is a long commitment. Adisorn Lertsinsrubtavee 18:25 Yeah, talking about the funding is quite complicated, because have been generally support by many, many funders. Like, initially we get support from the French Government, George Michaelson 18:37 Ah, development projects, from them. Adisorn Lertsinsrubtavee 18:41 And then we also expand it. Get the funding from national fundings as well, right? One of them is like ours at Thai regulators called in NBCT, [George: yeah]. And we also get the support from the EU co-fund with the APAN community called Asia Context Programs, [George: yeah], as well. So, but so that is but they still have more behind that one, definitely. George Michaelson 19:05 But this is not a program that's looking to end. This is a continuing obligation to understand the problem. I think it's really interesting that we can now see that things that we take as part of our lived experience, building out networks, building cellular fabrics. There's this component for you Sukumal, where you need to look at the quality of them and assess them as things in their own but then there's the part you're in Adisorn, where you look at these networks. The whole point of the network is to do something useful. And here you are doing something useful using a network. So Adisorn, for your project. Do you have student directed outcomes? Are people actually managing to get higher degrees working in this project? Adisorn Lertsinsrubtavee 19:47 Yes, because yes the one of the reasons of why we try to decide everything from the hardware and software, because we want to involve with students and so on. So now, yeah, we recruit the students graduates from AIT, and also elsewhere to join these projects as well. So we have a master student, graduate from AIT joining this one. We also have the intern as an bachelor degrees, join as internships as well, to for the data analytics and so on. George Michaelson 20:15 That's great. It's fantastic. Thank you. It's really interesting to hear about this. Sukumal Kitisin 20:20 Thank you. Adisorn Lertsinsrubtavee 20:21 Yeah, thank you. It's our patient because we can share what we doing, especially the platform is live. You can visit the website called https://hazemon.in.th/ George Michaelson 20:39 if you've got a story or research to share here on ping, why not get in contact by email to ping@apnic.net or via APNIC social media channels also remember the measurement@apnic.net mailing lists on orbit is there to discuss and share relevant collaborative opportunities, grants and funding opportunities, jobs and graduate placings, or to seek feedback from the community on your own Measurement Project? Be sure to check out the APNIC website for all your resource and community needs until next time you.