 So first of all, thanks for coming. The idea of the talk is basically to share some experiences. The other day we were talking with some people from the Debian publicity. And we came to the same conclusion that basically even though the topic is related to Debian, like could be related to other Linux's, the idea is basically to get a perception of how big the project is or what the impact of your work is. So basically we work in a company that builds a satellite. My name is Erg, he's a team. When you think of a satellite you really don't think about stuff like this. Which in fact those are satellites. They are classified as nanosatellites, which means in comparison to something that weighs between 1,400 kilograms. It's really ridiculous. Those weigh 2 kilograms. So basically a company we work for is called Satelloshik. It started building satellites like three years ago. We already launched it prior to the satellite we're going to dedicate the whole talk about that we're called Capitan Beto and Manolito. Those are them. Those satellites orbit at 600 more or less, 600 kilometers. So they are cataloged as low-airbit satellites. That's because basically commercial satellites or commercial-grade satellites usually are 2,000 kilometers. They are 20,000 kilometers, a little bit more than that. So basically the idea of these two satellites which are, as I mentioned, 2 kilograms and 10 and 20 centimeters for what's average consumption, power consumption. They are really very small. There's something like a little bigger than this, but square. They have those form factors because they are prepared basically to be launched from... We are going to show the images later from satellites that deliver like 50 or 70 of them in a single flight. Basically the idea to test stuff. For example, that's the first one we shipped in July 2015. And this one was a little bit more advanced in terms of the parts that were... Sorry, not advanced. They were basically spin-off on the original design. So the idea behind the launching was to see whether a company has the capacity, the technology, the means, and basically to see whether it was really feasible to do technology. The company is eradicated in Argentina, but we have quite a couple of people from different parts of the world. So in this satellite, which is Manolito, some of the components were replaced with things built in-house. So the idea is that 40-year-old space industry works like world-proof technology. So to be world-proof, you have to first have an experience or spend some time in the space, which means 10-year iterations because satellites really tend to take from 4 to 6 years, 4, 5, 6 to 10 years to be designed, built, and shipped. The idea basically with the company is, well, in this case was to replace some components. Between others, there was a camera, some magnetorcas that you see to basically use the magnetic field over there to grab from it and to... To control the attitude. Yeah, so basically there are several components that control the attitude of the satellite. This particular one is like one of the forces that tends to turn the satellite in different directions. They are component. Between the camera, magnetorcas, reaction wheels, solar panels, and some other stuff will basically replace it in-house. But yet the view was to come. I mean, this is 2 kilograms, maybe it would be... They are catalogue as nanosatellites, but they are basically experiments to play with. If you were in the talk with... I remember it was Vdale. He mentioned he was basically going to deliver. He was building one of these. He was referring to literally one of these. So Tita was a satellite that jumps a little bit farther. It's considered a small satellite. It's no longer like a nanosatellite. So as I mentioned, the idea of the company differing from different companies doing space observation of Earth is that it's based on relief, release often, release early. So the regular space industry works on like 40 years cycles. The technology improves little by little. It's almost the dimensions, as I mentioned, the weight of a regular satellite, which is not this one. It's like something like this weighs a ton. It's basically between 50 and 100 million dollars to get built. Some of them are even like near the two or 500 million dollars. They're prepared to have a life expectancy of 10 years. So basically they are built to last for longer times than these kind of approaches. But the idea of the company is basically to build something that you know about, you learn, and to improve and release faster, way faster in terms of time. So this is Tita. This is an engineering model that is digitalized. It took between... So it was launched literally one year ago at the previous day. We were just releasing it. It costed like 250 in terms of building. So those are like to have an idea, comparison to 100 million. It's significantly different. 250,000 dollars compared to 100 or... Sorry, 100 or even more to be one of those. So as I mentioned, the designing, the building, the engineering took eight months approximately to build two of them because basically on the industry you cannot build the software and, well, generally speaking, build the software or build the hardware and later improve some stuff without testing it. It's like regular software in terms of basically pushing into a testing environment or pre-production or QA, whatever. So basically we built two satellites in eight months, the engineering model, which is called Rhodesia, which is called... I don't hear my finger. No? There. Which is called Rhodesia. The first time I saw the shipping box, which is that one, I thought of this image. I don't know if anyone recognizes it. The image is not really very good in the projection. Anyway, so meanwhile in Russia, the satellite basically... So you can launch satellites from several parts of world. Basically, the difference is the price. Nowadays, commercially speaking, you can launch from China, Russia, France, I mean, the European Union and the US. Well, the difference being like 10 times between them. So we launched from Russia. This is part of our crew. There you have HEDA, which is our CTO. You have VILA, which is our payload engineer. And Garza here. He's a mission and power engineer. They were preparing their... They were unboxing and loading the satellites from the shipping box. I'm going to show you just more images to have an idea. These are power cells. More power cells. Well, we're going to see more images. So this is basically unloading the satellite on the space head model. So these satellites are basically loaded on these things. That's the head of an average rocket from Russia. It was actually used as a... How do they call it? Never mind. It was basically a missile a couple of years before that. Yeah, intercontinental missile. So this is basically how the satellite is positioned. We can see very here. So basically this is... You can see there at the top corner, different boxes. Those are the launchers. You have the satellites in there. And you see there's a small raven, red raven there. Those are basically the remote before launch ravens. So they are checks. Those boxes are literally ejected by a powder charge. So you don't really want to... You don't really want to buy missile checkers before getting launched. These safety things are called remote before flight tax. So that's the distribution of the satellite internally. That's the rocket head. So you have a perception. This one was... I don't really know how to pronounce it. It's deeper than whatever. Space head model. So those are payloads. So to have an idea, that's basically a large footprint satellite. Those are regular satellites that are involved in communications and traditional Earth imaging. So as I mentioned, the launch was a little bit before the conflict last year. We have that crazy idea which ended up... Yeah, which didn't work because we were just after the launch and we were very excited with the telemetry and whatever. We were thinking of, yeah, let's go like a photo group and let's do the... How do you say it's here? Deviant swirl in the floor and let's wait for a satellite pass so we take a picture from there. It was a little bit too... How can I say it? Yeah, well, it was a little bit crazy. Anyway, I'm going to show you why later. So the satellite process of getting lunch is very stressful. Basically, the components that are built into the satellites have scales of micrometers, predictions, like the telescope. So a lot of testing is required prior to get lunched. There's also a large budget of tests scheduled at the release of the satellite from the launching platform. Basically, once the satellite gets lunched, it has to go through a process of getting sync into the desired trajectory. It's a very large amount of calculation. So the idea is once you get lunch, you get like a black curtain. It's something that you have to wait in the corner of the table waiting for someone or some entities, governments or whatever to tell you the lunch has been really successfully performed. Otherwise, you're not going to see it. It's too small to be self-former. It's not like the International Space Center station. So basically, you will release and just wait. So luckily enough, we have amateur hand radars all around the world. One of them, 30 minutes after deploying the satellite from the launcher, saw the telemetry beacon. That's basically it. So it's basically like a string notation of I'm satellite. If you see me, drop me an email in this direction. This is literally the stream you see on the floor. So by the time it got released, we were like... So as I said, after 30 minutes of getting lunch, this guy at Japan saw the telemetry. He saw it like with 24 minutes of time. That's UPT, it's uptime. So basically the whole sequence of launching, deploying the un-finance for communication and everything was in those four minutes. Just after it was released, this person was like, you know, able to hear it. Those are the musicians. You can see Teenie here. That's in. That's the lunch party. We also have a sales too. I don't know if you know what a sales is. They are great. No, those are not barbecue. Never mind. So basically, what's about it? Hello. And what's Debian about Tita? So what about Tita and Debian? This is the heart, the Tita's heart. Tita has two sets of computers. One in charge of navigation and communication protocol. And the other in charge of mission and payload. Well, in fact, has two hearts. From the heart perspective, every Torino. That's why it's called Torino. Torino 2K, the first one was called Torino. I don't know why. Every Torino has Cortex R4 with lockstep and a set of six overall come sticks that you can see there. Vertical orientation. They have Cortex A8. Another switch and the rest of the satellite is compounded by a high resolution Earth observation telescope camera and a couple of macro black and white and color cameras. And the attitude instrumental, like a star tracker. It's not there. No. Reaction wheel, gyroscopes and magnetic torque. As we, as he said, Tita was launched on 19 July last year. While the system D discussion was getting warm, we put system D on Debian. We put it on six overall, six come sticks and launched them. To make a note, we basically put it and port it. It wasn't like safe enough. So we basically put it there and release it on the space. System D was a key at many core services. Because it allowed us to handle recovery automatically on service expected to be eventually running. We shortened that the voting time from 30 seconds to less than 10 seconds. Which is really critical on the satellite operation. What we use from the Debian infrastructure is a chain building cross compiling, GBP, it will package workflow and pay bill there, a software deployment tool. We are resting on DPKG, Robo's platform and state machine for deploying software. We needed to minimize new upstream releases, releases footprint, because we have a very, very low bandwidth. So on top of the Debian format, we build something similar to delta dev. Which is present on the archive. Except that DD patch is optimized for non-binary stuff. This is the workflow. As you can see the operation there, the operation guys upload the diff. That's a satellite link. The green arrow is the satellite link. They put the diff in there, then it triggers an attended upgrade which calls the patch with the previous dev in the git diff. Actually, the diff is used, we use git in there. And then generate the new dev and deploy of six of a gamma six. So I don't know if you can understand this. So basically the idea is we just push diff. That DD package which is something we built on house and we will in a couple of weeks put in the archive is really simple, but the consequences for us are huge. So basically we patch a dev from the original, sorry, the previous version with the git diff. Actually to build the diff, you have the old dev and the new dev. The previous dev, the new dev and you unpack them and then do the git diff. You apply the patch to the source and they compress it. But this allows basically to lower down the footprint of a package from dev delta, sorry, delta dev from 10K or 15K to even less than 1K. Which was great because one of the things with the link is that the satellite spins at 7,000 kilometers per hour. It's basically like 24, 7 kilometers per second or something like that. It's very quick, so it's very fast. Once it goes through the ground stations, we have one in the north pole that belongs to Norway at Svalbard. Which is great because since we have polar orbits, every 19, 7 minutes we get telemetry and communications from there. But also we have communication ground stations on Buenos Aires, on Tortuguitas. Once you get a session for transmission, you really want to get the most out of it. Because that's basically one of the parts of the orbit in which you upload commands and you then get the whole task processes and the images. You have 10 minutes. Yeah, so basically you have to really, really go for it. One of the funny things we came through, the release process as I mentioned before, sorry, the release cycle of TETA was eight months. Those eight months include basically the building, literally building pieces of the satellite engineering mechanics, getting the telescope, the thermal protections, joining all the pieces together, the integration process, the shaking. As I mentioned, the process of launching the satellite is very stressful in terms of the mechanic parts. So you have to go through a process called shaker which basically shakes the satellite at three, four, five Gs, something like that. It's crazy, you see it moving. Remember, you have nanometer precision on the telescope. You see that being shaken and it's very scary. As I mentioned, the whole process is very rushy. I mean, even though it's eight months, those eight months includes everything in the process. Even the design of the software is going to run there. One of the funny experiences we work through is that you may know it as Cluster SSH. For us, it's internally called as Parallel Breaker. Parallel Breaker. Well, this is actually very funny in terms of the experience we went through, but it's extremely similar to Apollo 13 Odyssey, I mean, making a comparison. So you can see here, those are basically the six of Earth. Overs are these things, those things, which are literally the pieces that have deviant on it. So those things have, as he said, those are, so you can see the switching area there. Those are two different boards that communicate between each other through the backplane, and you have a switching for Ethernet. So basically they have IP stack on top of them. So what happened is that, what were you doing that day? You were uploading software for the camera or something? No, I just cleaned it up. So he was basically doing some, yeah, the last clean, literally. He was basically performing some cleaning operations. We were to literally ship the satellite to the shaking facility the next day. So that's SSH. It basically opens a console there at the right top. You type commands there and it gets split on all your terminals. So you connect from the computer to a satellite through something that is called an umbilical cable, which is umbilical. It basically is a cable that you unplug once the satellite is shipped, but it emulates communication if you were with an Athena. The idea was that, and this link is really fast in comparison to the real link that happens on Earth orbitation, but it even emulates latency. Because, I mean, you have to have behaviors that are reproducible on Earth. So the latency in this situation was something like a, you might see there at the top corner that there's just very small amount of lines. That's basically, um, that's basically RM slash RMFF from the route. I didn't know safe RM. Yeah, definitely you have to install safe RM. It's less than 1K or something. I'm going to definitely save, for example, a satellite. So basically the thing was that a little delay on the screen made him to, you know, push enter. The other didn't refresh and he went through something like that. We basically started to delete everything from the hard drives. So you have to take into account that, as I mentioned, these satellites, you put it into something that shakes, and those are not hard drives. You can, you know, unplug and replace and put something in. Literally they're glued together with a high density epoxy. Epoxy, yeah. So as you can see there, once he breaks, he conserved there at the first line. There was no LS. So there were just a little bit amount of directories. Some of them were like documentations, which were useful. Sorry, useless. The opposite. Yeah, the opposite. But, and this is the part that correlates with Apollo thing. There was a QAMU statically compiled on the server, which, For boost trapping. Yeah, for boost trapping, which had activated and compiled the remote functions. This crazy guy called Phil, that works for us, he's French. I don't want to make any relationship between French and anything, but this guy basically built, well, he found the part where he could drop at AirSync from a GDV. He basically opened GDV remotely. That's it. From a different, which was the one that really went through the common successfully. And basically he recovered the whole thing from a remote GDV platform. Basically he loaded the, yeah, he loaded the AirSync so he could remotely load files from there to, through GDV on, on Divagini was crazy. So basically he said, Jim, that was funny. Actually he's, this experience took, I summarized it, but they took like 16 hours. They eventually went through all the different possibilities of, you know, breaking the thing and replacing, luckily enough, we have QAMU there. Anyway, that was a very funny experience once you go through it. Not that. Yeah, so he, so one of the crazy things that, he was the guy that did that, he run that common, but, well, yeah. So another crazy experience was that, well, we have operations, right? So we have people that have a list of tasks to go through and to check, like a regular operation guy in an IT company. So this guy, one day, let's call it Facundo. So Facundo was having his nice tea at the office and he was reviewing, you know, the last activity from Tita, the satellite. And he wasn't really even close to, you know, imagine what he was going to be put through. So that's basically one of the images from, downloaded from the satellite. So at the third glance, it appears that, you know, it's a regular image, it's a black and white image, but what's that? I mean, that's not really, so from the orbiting perspective, that's not actually a satellite. So we basically were through three hours or maybe more, putting names to that thing. So eventually, we realized that we were going to put watermarks on that image, so we don't get that image split all over the internet. We wanted to know what the hell that was. So what happened is that the CEO and this guy basically uploaded in a very hidden way, like uploading EXA, this image, like three months before this guy downloaded it, and eventually this guy got this image. And everyone in the company started to think this image was real, but it was in fact, you know, pushed it from Earth and eventually downloaded from the satellite. Everyone was thinking, you know, we are going to invade or something that's the beginning of the end and whatever. So we do have a lot of metrics. So we basically have, like, I don't remember exactly the number. We have literally tons of metrics per minute. We don't even store them because otherwise we're going to destroy the internet storage. Some of the information is later queued from the download. We use tags also. If you know Grafana, you can see those red lines that are tags. That's basically telemetry. You can see temperature there. You can see uptime. So these things have something that are RTC. So they are devices that have their own power to basically get a clock running, even though the satellite is off, for example, because of power reasons. Telescopes tend to consume a lot of power, so they might eventually take your satellite down. So that's the battery there. That's the current. That's the mission operation. So basically you have, like, when you go through cruise mode, so there are different modes of operations. Like, you have the satellite going in... Regularly it goes in tumbling. This is called tumbling, right? So basically the satellite spins one degree per second. So it doesn't receive... So the satellite, as you saw in the images, is basically a large CPU box, something like this, a very old CPU box. And it has some thermal protection, but if you basically rest on a single position through the whole orbit and you get the sun of the light directly to the telescope sensor, you can basically consume a lot of power into a single point. So basically the satellite is running on a tumbling mode, like it spins very slowly. Thank you. So that's basically disinformation. The screen is not really projecting very well. So basically this is a satellite going like this. It's like spinning. You can see the information. Oh, yeah, that's all, also. So the satellite has some hardware on it that helps change the attitude. So basically it grabs from the Earth field and some reaction wheels, which are basically like wheels with very, very high precision. So basically they grab... Well, basically that's information you get from the sensors. That's part of the information. All of them are attitude components. So those are reaction wheels, gyroscopes, gyroscopes refined, mining torquers. That's more information. That's actually a TCP dam running from the devices that run Debian. Those are literally ARM devices stock. I think we just compiled the kernel because of some drivers of the ARM chipsets. But it's basically stock. So that's telemetry. Yeah, I cannot even see it from here. Well, that's more information. This is one actual image taken from the satellite. I think it was the first month the satellite was released, which was like... You can see and hear the projection. This is like a model. Baxat is the name that is distributed using the satellite community. It's theta slash Baxat. Baxat. Yeah, that was basically a Baxat satellite, which was kind of fun. So this is the image that is projected in here. This is like the computerized projection, like the expectation, which is literally matching, as you can see. So fun enough, this was... As you can see and a stripe in here, this was the Florida peninsule. You can see Canaveral there, Disney, Miami. It's like the continents in this direction because the peninsule like this. This is another image. Well, the screen is not really helping us here. This is an image. This is the atmosphere scattering, removing the atmosphere scattering. So we are just finishing. We realize this is kind of fun once you are talking about it, but it's not fun when you are getting all the images there. We found that indeed the world does really have clouds. Every fucking time. Yeah, every single time. We were getting images. There were a lot of them, like, no, really. Not about women, really. Every single time it goes... It went through a very interesting position for us. We were like, you know, half or more part of the time getting images from clouds. So that's basically it. As we mentioned, this was basically something we found fun doing and we wanted to share with you so you have an idea of what a truly universal operation system is. Basically, those are deviant running on what they are called single-core computers, but they are literally computers like this desktop or wherever, or a server. They have very, very, very slight cosimizations on the kernel, mostly, well, on the system D by that time. But they are basically stock deviant there. So, you know, at least for me, it gives me a perspective of what I do and deviant really helps building this universal operating system. Okay, so we are out for questions. If you have any, just raise your hand. Any questions? No questions. There we are. It was too good or too bad? Very interesting talk. Are you putting updates regularly, for example, when new point release goes, are you on some different mode of updating software there? So, you want to talk about that? Yeah, we upload weekly updates to the satellites, but the satellites live for three years orbiting, so we just launch with Weezy and it goes down with Weezy. Yeah, so basically, that's actually true. So one of the things of having an orbit satellite is that it will eventually get drowned by the Earth field and get literally burned, so maybe it will happen three years, maybe more, maybe less. But as you might guess, security updates are not really that mandatory. They are really, really welcome. I mean, we do operates in terms of... So far, for example, the other day, we uploaded for some of the metric collective systems a Python LXCM compressor system, a compression algorithm, which basically for us is great because we have an attendant upgrade there, we just put the file in some place as that diagram. We just put a file there, it gets written on the NV RAM, then it gets pushed through the deviance, if they have to build, they build the package, otherwise if they recognize there's a dev, they just distribute it, they deploy them through the whole devices. So we do perform updates, yeah. Questions? No questions. Two questions. Awesome. What are roughly the specifications of one of those OBEROS RAM and disk and so on? It has one gigahertz of the clock, one gigabyte of RAM, 5012 of flash, built-in on the single chip. We have a micro SD card. Basically for bootloaders, right? Yeah, yeah, for... Small storage, yeah. Micro SD card with the 8 gigabyte for... Yeah, those aren't the regular SD cards. They are like... Is that five? Okay. Those are like... They are very similar, SD. Instead they are like enterprise, micro SD, something like that. So they have these correction algorithms. So as summary, it's not really the mission of the company to build huge and very, very HA-tolerant systems because you are going to deliver one every couple of months. So we do have redundancy in very specific areas that are critical and that they tend to fail. For example, these are... Those are basically where all the computing happens. This chip... The one there has... Which is called TMS. The TMS is a computer that... Very small computer. It runs free RTOs. It's basically just for switching on the communications and switching... Well, it controls thermal and power. Yeah, it controls thermal. It turns on heaters. The device goes from minus 20 or minus 30 degrees to 80. So some parts of the satellites are very critical in terms of temperature, like the batteries. If you expose the batteries to... Let's say... I think Karsa told me like... 10 degrees. Yeah, so if you expose the batteries, which are LiPo... LiPo? Is that... If you expose the batteries... Sorry? If you expose the batteries to less than 10 degrees, the life expectancy shortens down like half in a month. So they have like, you know, heaters and stuff that basically makes the whole temperature of the satellite go nice and easy. And those are like part of the... Well, actually we made this on radiation of every single piece of there, including the microSD card. Right. So one of the things we do on the satellites that are part of the, you know, regular space industry tests are going through, as he mentioned, radiation tests, because one of the things you cannot avoid is physics there. So radiation happens, and once you are, you know, at 600 kilometers, you get really radiated, irradiated. So we do test them for radiation. It's one of the things we do from the standard industry as shielding stuff, like, you know, really you cannot avoid. But those things aren't prepared for space. We build them, we design them, we test them before in the other months, and they keep running. And they just got one year running, which for us was like, you know, a celebrating thing. But yeah, we test them and they are running. Yeah. We have one minute, one question. Do we have one question, Ma? Yes. Yes. I just want to ask, can we write emails to the email address? Oh, yeah. Of course. Actually, if you, there's something, our CEO, Herat, tends to say that if you can break into the satellite, you are definitely directly hired. So it's like, yeah. Maybe we have time for one more question. Maybe one more question. There we go. Your system is quite limited in the space. So the question is, is the minimal Debian installation small enough, or would you like to see a more compressed, small Debian that you can use? No, no. We build an image from Bootstrap, and then put all the packages that we need. We check the bootstrap. I mean, literally, the bootstrap run, the output of that we put it on the SD cards. The tool of Debian. Yeah. The one you regularly use for building. We have eight gigabytes for each of our, so the root file system is two gigabytes. Yeah. And with mostly everything like from the platform. So it's like, those are usually used for a platform mission. So basically they do stuff around the telescope or they perform tasks. So the raw images are huge. I mean, huge. It's certainly impossible to download them. Some operations they perform are built, are post-produced there. Like once they got the image, they are post-produced. And eight gigabytes turn it to be like, you know, more than enough in terms of handling those images compressing, and then do some stuff and push it to F. Yeah. So they are actually, I think a little bit more than 50% used. Which is crazy, I mean. Okay. Thanks. Yeah. Actually the question they mentioned was that we can afford to have QEMO running there. Yeah. Well, not running. Well, it can run, but it's just for recording file systems. Not the Setsu for recording file systems. Okay. Thank you all for coming.