Latest / Technology Now / The new frontier is supercomputing in space: The History of HPE in Ten Objects
Transcript
- MICHAEL BIRD Aubrey Aubrey Aubrey Happy Thursday, good morning. AUBREY LOVELL Good afternoon, Michael. What have you got for us today? What's on the kitchen counter? MICHAEL BIRD Well, let me tell you what's on the kitchen counter today, Aubrey. It's the final episode of our mini series. So I thought we'd go big with this one. So my question for you is, if you could send anything to space, what would it be? AUBREY LOVELL Okay, that's interesting. Well, I don't know. I mean, maybe a piece of technology like a phone or something just to, you know, I know that we've put out random things in the past in our history. what about you? MICHAEL BIRD I think I'm quite a sentimental person so I think something like I like the idea of putting like a family photo like something that will go on to like a spaceship that will just sail into the distance forever. I quite like the sentimentality of that you know like the Voyager probes have got those gold discs in them AUBREY LOVELL Well, I assume, yeah, and I assume obviously we're going to be talking about space today, right? MICHAEL BIRD Well, not us personally Aubrey, unfortunately, that would be quite cool. But we are going to be looking at computing in space on the moon and beyond! I’m Michael Bird. AUBREY LOVELL I’m Aubrey Lovell And welcome to Technology Now from HPE. MICHAEL BIRD So Aubrey, as we said in the introduction, this is our final instalment in our history of HPE in 10 Objects mini-series and this time we are exploring new frontiers. AUBREY LOVELL Ooh, very exciting. And like the other episodes, this is of course a video episode in case you haven't noticed. So make sure to check us out, not just where you listen to podcasts, but also where you watch them too so you can see the objects we are actually talking about today. So Michael, with that said, I have a sneaking suspicion about what today's topic might be. MICHAEL BIRD Yeah, go on then, what's your suspicion? AUBREY LOVELL Okay, it's not far of a stretch here. We talked about sending things to space, and I know that we've done episodes in the past about sending supercomputers to space, which I would say is a pretty big deal, right? So can we possibly be talking about supercomputers in space again? MICHAEL BIRD Yeah, now you're absolutely right Aubrey. Because in the final episode of the mini-series we're going to be exploring new frontiers and the Spaceborne Computer Program which has already sent a few supercomputers into orbit already and we'll be sending them even further in the near future. AUBREY LOVELL Very exciting. I think this is one of our most fascinating topics. So what are the final three objects we'll be looking at today? MICHAEL BIRD Alright, so the final three objects are the Spaceborne Computer 1, the Spaceborne Computer 2, and... drumroll please... I think we should work on our drumroll. AUBREY LOVELL That was my poor version of a drum roll. MICHAEL BIRD Anyway, the final object is Spaceborne Luna, the third object, Spaceborne Luna. Yeah, so I think we should just go straight into my chat with Norm Follett, the Senior Director... HPE Global Technical Marketing, Space Technologies and Solutions, who told me all about the why, the what and the how of the Spaceborne computer program. AUBREY LOVELL Let’s do it! MICHAEL BIRD So Norm, thank you so much for joining us on technology now. You know we have such a cool selection of objects for this episode. NORM FOLLETT Yeah, it’s fun. MICHAEL BIRD but let's start with this one. That where we stood in between. NORM FOLLETT Well, this is spaceborne one. This is the chassis for spacebourne. One. This is the device that started it all. I would say in HP history, this is perhaps as interesting or as impactful as the oscillator or the HP scientific calculator in the impact. And, with respect to the innovation and breakthroughs that it facilitated, MICHAEL BIRD why is this object in particular so important to HPE’s story? NORM FOLLETT think it's beyond the HPE story. I think it's a story of, you know, humanity exploring into space. So NASA over the years, as they started to develop spacecraft and started to develop compute systems, they would build their own computers, proprietary computers, and they would spend tremendous amounts of money on it, tremendous amounts of time. And what really happened as part of the commercialization of space, NASA had that initiative in mind to use, you know, use more commercial products, get private companies involved, get companies that had expertise in areas such as computer science, directly involved in what they were doing. MICHAEL BIRD Can you sort of tell us story of how this got to being up in space? NORM FOLLETT its initial mission was one year in space, and it was only a test. You know, this is only a test. And what they did is they had a duplicate system in Chippewa Falls, and they were running high performance compute benchmarking tests. And NASA wanted to see if there was any delta, any performance change between that. And there wasn't. It did so well that after a year, NASA asked us to stay on station. And so we spent an additional six months in orbit, picking up operational tasks. So they immediately the kind of the light went off and they said, well, you know why you guys are up there? Can you actually do some work for us? So we went from a test system to a production system. MICHAEL BIRD So this chassis sort of standard size and shape for the International Space Station. But it's not a standard size and shape for traditional computing. NORM FOLLETT No it's not. So whereas like in your data center, it slides out. This is slated this, this is slid in sideways so that was kind of the first bit was, you know, really getting in there. And when we talk about hardening, physical hardening. Right. Survive the shake, rattle and roll. But it's also a software hardening. Now that I mentioned that philosophy that NASA ahead of it can't fail, can't fail, can't fail. One of the philosophies that, Mark and team brought forward was, well, it can fail because things do fail, but it recovers gracefully with no data loss. And so it's a software hardening process. So physical hardware hardening and then a software discipline and an approach applied to the systems inside. And that's what made this system successful. And also, I think, you know, as most space enthusiasts realize, the problem across the board al ways, I mean, always, if any mission is power and cooling, right. But so the first, first job from a power perspective was to take 28 volt DC that the ISS generates solar power and then convert it And then it goes in and it went into some traditional power supplies that you would like you would find in your data center. MICHAEL BIRD the servers inside of here, they're off the shelf. NORM FOLLETT Now this was meant for no maintenance, no astronaut intervention whatsoever. It either worked or it didn't. And when it went down, it was going to be down. It never went down year and a half. It never went down. So this is a, you know, a definition of a black box, although it's white. MICHAEL BIRD so this was in space. Presumably there wasn't any very, very long cat5 cable going from the space station down to Earth. Are we sending tons of data down to Earth. NORM FOLLETT Well, I mean, well, I mean, this is the ROI of what we're really doing here. So we actually, we operate, you know, there's a small command window, we operate, you know, once a week. Candidly, we get a command window where we can talk and interact with it. We often stage up software and execute it and kind of batch processing and do all that type of stuff. But what this addressed and this is where the light went off for NASA. This is the the classic, classic edge compute solution. And what I mean by that, do the computational work at the point of data acquisition and send back the answer, don't send back kilobytes rather than gigabytes of data. MICHAEL BIRD So, so rather than, you know, if you had a much and it was taking photos rather than sent every single photo down, maybe analyze those photos and then just say yes or no to something, that sort of thing. NORM FOLLETT Yeah. That's a that's actually a good prompt. For one of the solutions that we actually developed, But I will say that this that spaceborne one also won the NASA Very Distinguished NASA Technology Achievement Award. So and that's, you know, very few of those are handed out. That's how excited they were about what we were able to prove and do. This sort of drive style edge computing, not the necessity of competing just due to that, well reinforced it. I mean, reinforce that this is a really good thing. And again, originally it was the task was to, you know, be up there for a year and why they were preparing for deeper space missions. Right? Mars. And they figured, well, we need a system that's going to last a year. Can a commercial off the shelf system do that? They did not have the intention of having an edge system, high performance compute on the space station until they realized its value once it was in place. MICHAEL BIRD So this is this is spacebourne 2 and we’ll come onto this gold thing thing in a second. So just describe what we're looking at here and maybe compare it to spaceborne one. NORM FOLLETT Okay. So a couple of things. So this chassis is designed to be serviceable right. So that's the biggest thing. You can't get those doors open. But you know here you can yeah you can. Yeah yeah you can actually get that open. Right. And so again it's also turned to where the pizza box when you servers are turned sideways. So kind of a nontraditional installation. But again it's designed for astronauts to interact with. And I'll talk about that in a moment. But a couple of a couple of deltas you'll notice right away. So these are fuzes. Very traditional fuzes. Input output the liquid cooling And, then these are the power connectors. Now see all that? Remember all those plugs on the other unit. These are the two, power connectors. Now what's what's interesting here is that we actually have I described we had a power converter outside space where the power converter is inside this chassis. So we have a power converter in here. We found a 48 volt DC power supply, but the station produces 28V. So the power converter in here jacks the power up to be compliant with these systems. So the delta here, as we moved it inside, we're dealing with it ourselves. Less set up for the astronauts. That's another part of the mix, right. They don't have to deal with any of that as well. So that's one of the things that changed. And I just want to say both of both spaceborne one, spaceborne two, we're not involved in the avionics of the ship. What we are is a scientific platform. We're a scientific platform for the international scientific community to do work on the station, to do high performance edge computing on the station. And we've enabled a wonderful series of experiments. You mentioned, just that simple thing about photos, right. And so that's an inference engine, if you will. Astronauts on the space station, they actually have what they call an Eva season, and that's spacewalks. And that has our friend who major Tom, our friend right here, major Tom. That means that there's certain times of the year where there's less radiation outside ship. So they schedule their tasks and they do all that type of thing a little less, a little less risky. But the things that wear out are gloves, because that's the touch point. Just like you working in your yard, your gloves wear out. So what they would do is they can't come in from the spacewalk and they take literally hundreds of photos of an individual glove, and then they stream or send those photos down to, a shop down here at Johnson, not far from where we are today. and they look at those pictures and they go, oh, this one looks a little worn. And they circle it and they communicate back to the astronaut. Look at left hand glove number 14. And we think there's a wear on index finger like. And what they have to do is they go through a certification process to certify that those that glove can go back out. It takes a lot of effort. MICHAEL BIRD We talked about how slow the connection was. NORM FOLLETT Exactly. So as such, they carry all kinds of gloves. They have a, you know, extra gloves, extra gloves. What we ended up doing was building an inference engine, a machine learning engine, and I'm running it on. Spacebourne took all those photos in seconds. We go look at glove 13 index finger right. And so it just eliminated all that processing. Now today NASA does both because they're still proofing that system. They came back and they said they gave us the, NASA Team Flight award. So another award just say, that's amazing because they're thinking, what's what's what are we going to do on the moon? You know, we can scan our equipment and we can say, you're good to go or you're not changing. one of the things that we like to say, you know, proven in space available on Earth. So meaning that if we have the ingenuity and the wherewithal to solve some of the toughest challenges, you know, in space, in some of the harshest environments, we can help you in your data center, Secondarily, moving along the human condition, right. Because we're able to do science and help the scientific community learn things that they were not able to do and are not able to do on earth. there's a new market opportunity.i Right. Edge computing on the moon, in orbit on Mars. It's going to be a necessity for human survival. And there is a legitimate business case out there. MICHAEL BIRD Like the key thing, I guess, is that you're not having to send data back down to Earth or every single piece of data the or capturing, you can plug it into this computer this, that next year. And so. Run the experiment on this I mean send it back down. So that's what saving time right NORM FOLLETT Yeah. Absolutely. And having a high performance compute platform that can actually handle some of that experimentation and that also included something called a federated learning model and the federated learning model for, people that aren’t familiar with it. It's really about having large data sets distributed and everybody doing work in that distribution, and then sharing the result and then sharing the result with each other, and it becomes a model that shared. So you're not having to move large data sets to synchronize with everybody, but you have the opportunity to work with those large data sets because you're sharing the resulting model. And again, NASA very, very excited about that, because what that means is that as they explore, you know, other planetary bodies further out in our solar system, that we can learn as we go and continuously share information and again, create self-reliance. MICHAEL BIRD spaceborne, one. Spaceborne two will come on in a second. Is that has that will there be a spaceborne three. NORM FOLLETT Well, actually there really is a spaceborne three. we did a reef flight. We upgraded, the internals to one of the lockers, that's really the third iteration. So, on January 30th, 2024, we launch from Kennedy Space Center. MICHAEL BIRD Do we have a fourth one then? NORM FOLLETT We do indeed. Would you like to hear about it? We have, an opportunity working with a partner to go with them on their first rover, that they're bringing to the moon and that mission is called Griffin one. The rover climbs off that rovers called flip climbs off the back of that lander, starts to explore the moon, uses our spaceborne Blade computer to actually map, monitor and gather intelligence on the surface of the moon and conduct the first AI experimentation on the surface. MICHAEL BIRD So this is, spaceborne. This is spaceborne for spaceborne for for flip. So it's, it's about what? About seismic graphics code? I think it's pretty. Yeah. Yeah, yeah, it is, it is, it's, you know, less than less than a foot long. how is this able to do what this is. NORM FOLLETT Well it, it it doesn't really. Right. So and what I mean by that is that we have a different challenge, with spaceborne two, we flew the largest amount of memory to ever go to space, That was 120TB of space. This little fella, also we're working with Kyoka, has one terabyte. Okay, so there's only so much you can do Now what we did differently because this is a fundamental difference. These guys, as I described earlier, we are existing in an atmosphere now. Atmosphere is the internal spaceship. The international Space Station and same thing the astronauts breathe. So we exist in the atmosphere here. We are operating in the vacuum of space. very different. Right. MICHAEL BIRD Lots of different challenges. So what this chassis that we've designed is actually it is the cooling system, right? NORM FOLLETT So we actually have a series of techniques inside that actually allow the this, chassis to heat up. Right. And then it heats up and it's, it's bolted to the convection plate on the rover, and it's designed to transmit or let the heat pass from the chassis to that cooling plate and then let the flip radiator system take away the heat. this will be the highest performing computer to ever be on the surface of the moon when it's there. But we're going to quickly surpass that because we plan to bring our EL product line up to the surface of the moon on the larger rover, where we have more space to work with. And we also have the notion of building ideally what the what Astro lab calls its power towers. So think of distributed edge computers creating a mesh network that will also be delivering what we're doing with 5G networking as well. The ISS the is a was a preparation for us to go further in space. And that that next further destination is the moon. The moon is preparation for us to go further in space. And that next destination is Mars. MICHAEL BIRD Okay, before we go on to Mars, What’s it’s purpose on the moon? NORM FOLLETT Well, we're going to do a couple of different things. So I mentioned the Federated learning model. We want to show that you can have a multi node federated learning model. Now what's really cool is the three nodes will be Earth probably Chippewa Falls from our lab there. We'll have the ISS and we'll have the moon by the south pole of the moon. And they will be sharing data sets and information as this, rover drives on the surface of the moon. And we're also going to be doing a light arm mapping exercise, because despite, you know, we have great photos of the moon. It's been, you know, we've had photography for decades now over the top of it. But it's very different when you get on the surface. There's a lot of information to learn. we're going to be driving around mapping the surface of the moon and this computer will facilitate that. MICHAEL BIRD what sort of connectivity does does this, the spaceborne lunar have? I feel like the ISS is slightly more obvious. You know, it's of going around. Right? Right. It's probably got direct contact at various points right on Earth. Does this speak to us? Well, how on moon does it speak to planet Earth? NORM FOLLETT You know, so a couple of different things. Again, it's a symbiotic relationship with the flip rover. So we're dependent upon the flip rover subsystems. They have two different satellites that they're going to be working with. One in Australia, one in Europe. So we're going to try to keep kind of that continuous dish of communication. The pipe is very narrow as far as the connectivity back and forth, what we want to do, and it's a shared experience, you know. So in fact, some of the things that we're going to be doing, just as we do with the ES, is that we queue up the resulting data sets, and then we get our command window to communicate that information back and forth. The number one export is going to be data. And we're in a position to help, you know, mine that data and make that happen. And just to put things in context, when you think about the volume of the moon, the moon is actually the size of Africa, right? So you need some wheels to scoot around and explore that. And you need the ability to analyze, measure, and, you know, learn from that exploration. And that is what we hope to do with our continuous spaceborne program, bringing edge computing to some of the most, diverse and challenging environments. MICHAEL BIRD What's the next step? The spaceborne lunar. NORM FOLLETT we have the notion of a spaceborne five. And that is going back to the ISS. in the industry, you'll hear a lot of conversations and talks about orbital data centers. You know, people, you know, saying, hey, you get power for free, sort of that you get cooling for free will, sort of, or it's going to be less expensive. I would contend that we actually probably do already have the first data center in orbit with what we've been doing with spaceborne, to in the spaceborne program in general, but we plan to expand that. We plan to bring, Gen 12 servers, up there and give even more compute capability. We may go larger than two cabinets. We may build you know, four cabinets, six cabinets, whatever. We'll just see, you know, where that leads us and then of course, our partnership with Astro Lab, it has the, you know, they are not just focused on the surface of the moon. They're building rovers and equipment, targeting other planetary bodies, aka mars. They're they're also part of the Artemis program. And we have we have a relationship with them to be part of the Artemis program. That's a return to the moon. And that's where we plan to bring, you know, probably some more, you know, expanded system bigger than this. And one of our new L 8000 product line, potentially. We're working through that with the product teams now and the teams and so that's where a part of our relationship extends. And then also, Astro Lab is competing for the lunar terrain vehicle contract. to be announced in November of this year. It's been a long process, but that will result in a fleet of rovers. So we have the potential to have a fleet of spaceborne computers roaming, exploring, and learning about the surface of the moon. AUBREY LOVELL I mean, I think that was a wonderful conversation. really love just learning about all the latest and greatest of what we're doing in the space because it is so, like, to your point, frontier. It's so extraordinary to be able to, like, blend the lines between, you know, technology and outer space and areas where humans normally don't go. Like, you think about when we did human exploration on the moon, that started in the late 60s, right, at least for the United States. And then, you know... almost 60 odd years later, we're looking at going back to this place and going back to, yeah, I there's some missions that apparently are going to be planned next year for humans to return to the moon. And then you also have this whole technology story with HPE and our spaceborne computers going up and seeing what the possibilities are in today's era versus like years and years ago when we've done it. Yeah, it's pretty exciting. MICHAEL BIRD I think I'm such a fan of sci-fi, I think, and space. Some of my favorite films are films about space going to Mars. And so, I don't know, it just sort of feels like this sci-fi stuff is starting to become a bit of reality. We've heard it on the podcast before, but the story about the gloves, I think it's so cool. It's such a clever use of technology. It seems so obvious when you think about it, but it's such a clever use of technology and clever use of the AI tools that we have available to us today AUBREY LOVELL So Michael, when you were there, and this is for our audience that's listening traditionally through like podcast audio, right? What, like how big was it? Was it like, you know, what was the size of it? How much did it weigh? Like what, what are we talking about in terms of like the overall essence of it? MICHAEL BIRD I think they were the size of a suitcase that you could fit into a cabin hold, that sort of size. They were more than 23 kilograms, I reckon. It took a couple of us to have to sort of lift it onto the table. And I don't know, it's probably about the same size as maybe a... bigger than a 4u rack but just sort of on its side because it was sort of perpendicular to the way that you'd usually slide it into a rack. So it's quite tall, quite wide, quite deep. AUBREY LOVELL Yeah. So 23 kilograms is about 50 pounds. So it's basically when you check your luggage to go on a flight, that's what the size of it is. MICHAEL BIRD Yeah, I think it was probably slightly more than that. I didn't get my of my luggage scales to check. But yeah, but the Spaceborne Lunar was about the size of a graphics card. I could hold it in my hand and it was completely gold colored. It looked like the most blingy graphics card. No one's going to tell me off because it's so much more than that. But it was sort of like, it was quite cool AUBREY LOVELL I think it's gonna be really exciting to see like this resurgence of space exploration. I feel like we had a little bit of a lull. Like you think about the space shuttle program and things like that that have gone away and we've kind of had this like gap, but at the same time building the technology in the backend to now where we're starting to see this come to life. I think it's gonna be really exciting like next 10 years or so to see what comes to light MICHAEL BIRD Yeah, I 100 % agree with you. Now, Albrey, throughout the series we have asked each of our guests why the objects we've discussed are so important to the story of HPE. And I've got one I want to share with you because I think Norm's answer is a brilliant summary of everything we've done and everything we sample. So just have a listen to this. MICHAEL BIRD why is this object so important to HP story? NORM FOLLETT I think it's important for our future as a company, but it's also a terrific symbol of who we are as a company. Our capabilities,our innovation, and our passion to make a difference and do something that hasn't been done before and do it well. MICHAEL BIRD There go, thank you Norm. Now, before we finish, we do actually have another bit of news to finish off this episode with. Aubrey, this is going to be your last episode of Technology Now for a little while. AUBREY LOVELL It is. I have some pretty big news to tell everyone. So my spouse and I are expecting. So our baby is due at the end of January and I will be going on parental leave, which means you probably won't hear from me for a little bit on the show. But we're super excited. MICHAEL BIRD I am really really excited for you and obviously from the whole team we wish you the best over the next few months. AUBREY LOVELL Okay that brings us to the end of Technology Now for this week. Thank you to our guest, Norm Follett And of course, to our listeners. Thank you so much for joining us. If you’ve enjoyed this episode, please do let us know – rate and review us wherever you listen to episodes and if you want to get in contact with us, send us an email to technology now AT hpe.com and don’t forget to subscribe so you can listen first every week. MICHAEL BIRD Technology Now is hosted by Aubrey Lovell and myself, Michael Bird. This episode was produced by Harry Lampert, Izzie Clarke and Spencer Trinwith with production support from Alysha Kempson-Taylor, Beckie Bird, Alissa Mitry, Rahil Ali-Mohammad, and Renee Edwards. Our video editor was Mikey Nissembaum and as always, our theme music was composed by Greg Hooper. AUBREY LOVELL Our social editorial team is Rebecca Wissinger, Judy-Anne Goldman and Jacqueline Green and our social media designers are Alejandra Garcia, and Ambar Maldonado. MICHAEL BIRD Technology Now is a Fresh Air Production for Hewlett Packard Enterprise. (and) we’ll see you next week. Cheers! AUBREY LOVELL Goodbye for now!