Data Centers in Space Pt. 1
Space has long fascinated humankind, and we’re more prepared than ever to commercialize it –but we need to bring the edge to orbit.
Announcer: Welcome to Not Your Father’s Data Center podcast brought to you by Compass Datacenters, we build for what’s next. Now, here’s your host, Raymond Hawkins.
Raymond Hawkins: Welcome to another episode of Not Your Father’s Data Center. I’m your host, Raymond Hawkins. As a reminder, each episode features four trivia questions related to our guests. Please email your answers to me firstname.lastname@example.org. We’ll draw a winner from all the correct responses. The winner will receive a $500 Amazon gift card. Again, email addresses are email@example.com. You can find all our past episodes on our website at compassdatacenters.com and feel free to email me questions or contact our office via our Twitter handle at Compass DCS. I am recording today with Rick Ward. It is Thursday March. Let’s see, as it eight? Eight, I think it is. Rick Ward CTO and founder of OrbitsEdge. Rick, how are you this afternoon?
Rick Ward: I am doing great. It’s 11:30 where I am, so it’s not quite afternoon, but we had a rough storm last night, this morning, and now it’s beautiful and clear.
Raymond Hawkins: Let’s say I know you’re in the Southeast. Everyone’s safe with the storms that rolled through here in the last day or so.
Rick Ward: Everyone that I know is.
Raymond Hawkins: All good?
Rick Ward: Yes.
Raymond Hawkins: All right. Well good. Well, Rick, if you don’t mind, we’ll kick off if you’ll give us a little bit of your history. So I’m going to lead off with one point because it’s near and dear to my heart, Rick and I are both Marines. So appreciate your service. Thank you for serving. If you tell us Rick, a little bit about your background and then we’ll get into OrbitsEdge, and how in the world does a data in space have anything to do with the data center business for Rick, if you mind giving us a little bit of your history, where’s home, where’d you grow up, school, how’d you get in the business and we’ll sort of drive from there once we hear a little bit about you.
Rick Ward: Yeah, sure. As you would never have guessed I’m from Alabama. And I grew up here, went to the Marine Corps after high school, carried around a rocket launcher for four years. And then ultimately I ended up working for Deep Space Industries, which was a space mining company. And they were focused on asteroid mining and worked there for two years. And our contract with NASA ended. And then it was time to do other things. I was interested in the space industry and staying a part of that because I really felt that that’s a place where I could make an impactful contribution with my life. There’s a lot of things you can do that can make you a leader.
Raymond Hawkins: So the movie Armageddon is real. We can actually mine on asteroids?
Rick Ward: Absolutely, one for one, everything there is perfectly accurate.
Raymond Hawkins: Yeah. That movie was a 100%. I figured as much. I mean, I can’t imagine Bruce Willis telling us a lie.
Rick Ward: No.
Raymond Hawkins: Walk me through how that got you to OrbitsEdge. Although there was another step between mining… Was there another step between mining and OrbitsEdge?
Rick Ward: Actually not really. Kind of checked out the start-up area in, in Orlando after OrbitsEdge, because that’s where I’d moved to kind of skip that part. The Deep Space Industry’s job was in Orlando. And I was looking at what’s involved in conducting space mining. And I saw that obviously it had to be autonomous. It couldn’t be a person out there for year three, out there with a pickax. So it had to be robots. And that led me to, well, what kind of computers exist in space right now? And that was eye-opening for a lot of people who don’t have extensive decades of space background. That’s when I found out that the RAD750 is one of the most popular processors. It’s actually on the Mars perseverance Rover today. And it is a single core 200 megahertz processor approximately equal to the iPhone one. So, if you think about that and this thing debuted in the mid 90s, so you’re not talking about a powerful compute and in terms of what kind of AI it can push, the answer is not very much.
Raymond Hawkins: And that’s what currently deployed in space?
Rick Ward: That is current. I mean-
Raymond Hawkins: Right now.
Rick Ward: Those are common. You see Raspberry Pis basically arm processors that are going into cube sets today. But those things have a life cycle with like six months if you’re lucky. Those missions are intended to last a very short duration and do a very limited scope of operations. So they’re A not powerful enough to run good AIs and B, they don’t last long enough to even get you to the location. So that’s kind of a non-starter. That led me to, how do you make a computer space hardened? And the process involves basically a ground up redesign. And I did not see where that was tenable. As you get greater density, greater complexity, yes, you have better tools for doing it, but it still becomes harder and harder to make each new generation of compute hardware RAD hardened. So we looked at it from the other direction, or I was actually in conversations with a friend of mine who has done IT for 20 plus years.
Rick Ward: He became an advisor. And my thought was, what if we build the box that the computer can live inside? What if we build something that has radiation shielding, at least enough radiation shielding, has thermal management, has power management and you can just put the computer inside that box and it will keep it alive for long enough to do some useful things. And by doing that, you kind of sidestep that whole, how do you harden the computer?
Rick Ward: You can have current year model hardware as new equipment comes online, it’s relatively easy to add it into the system and you get the side benefit of using cots commercial off the shelf, compute and hardware means that you can also use cots software. So whatever you are used to using here on the ground, you can use in space. And that’s actually a big deal because there’s a serious bottleneck of people, of programmers for space compute for space software, because they have to work with relatively old languages. They have to be ultra efficient in processing power and power management on their coding. So it’s not a trivial thing to write software for space applications. So, where it might take, I’m just going to make up some numbers here, it might take 40 hours of coding and debugging to make an application to do X task here on the ground. To do a very similar task in space might take 400 hours.
Raymond Hawkins: Right. Because we’re running against older processors. Processors that have shorter life cycles. And so we’ve got to be incredibly efficient in the code writing to be able to handle the tasks. So it’s not just a compute pure hardware constraint, it’s also finding software optimized for space is also a challenge. This is what I hear you saying.
Rick Ward: Right, exactly.
Raymond Hawkins: I got you.
Rick Ward: So by allowing this modern compute, then you can actually take that application and say, “Well, this is similar to what we need to do in space. Let’s take it and modify it to make it work there to do this somewhat different task.” And you’ve just flipped that on its head.
Raymond Hawkins: I got you. All right, Rick, we’re working in trivia question number one, Rick, you are not eligible even though you have Google. Yes everyone can Google the questions but the first question is in what year was the US NASA’s landing on the moon? What year bonus question number two, who were the astronauts who landed on the moon? So that’s our first two questions while we talk about space and compute in space and yes, data centers in space with OrbitsEdge founder, Rick Ward. All right, Rick. So we got really efficient code and we got processors in space. I liked your term cots. That’s got to be a space industry term. In other words, we didn’t have to go special, make a hammer. We could go down to a Home Depot and buy a hammer because it’s now applicable in space that that’s where you guys got that term. I’m assuming, right? Off the shelf I like that. Right?
Rick Ward: The term actually comes from the IT world. One of the funny things is we’re kind of a mish-mash of IT people and space people. We’ve had to kludge together our own little language because in some cases you have the same word meaning two different things.
Raymond Hawkins: I got you, all right.
Rick Ward: So that’s been fun.
Raymond Hawkins: So OrbitsEdge, you guys are trying to solve the problem of how do I get robust compute in space and robust computing space that can actually survive the journey and survive long enough to produce meaningful compute power and meaningful information flow. So hardening the device itself. I think those of us that don’t think about space the idea that radiation impacts the computer in space. I would’ve never thought of that. So if you don’t mind, will you take two or three minutes and just give us how OrbitsEdge… What the devices are doing, what you’re supporting in space. How did this demand come up? I can understand why you need compute on a spacecraft to get it there and get it home. But what else goes on up there?
Rick Ward: So most satellites are actually quite dumb. They are remote control, basically. They’re quite blind. They depend on us to know where they are. And if there’s a possibility of a satellite colliding with another satellite, it doesn’t know that before the fact, because honestly, they’re traveling really fast. So by the time you would see another satellite it’s already too late. So we pay attention, we map all that stuff. We do what’s called space situational awareness, and this is all tracked and calculated. And they might say, so there’s a 3% probability that there will be a collision between these two satellites three weeks from now at this specific point where their orbits converge. And like I said, 3%, that is a point where they probably will cause one of them to move. And that’s actually a negotiation between two entities, two companies.
Raymond Hawkins: So I think of satellites and I’m going to use a term that just geosynchronous orbit. I think of a satellite that goes up and it’s supposed to turn at the same pace as is the earth. That’s geosynchronous, right Rick?
Rick Ward: Mm-hmm (affirmative).
Raymond Hawkins: Okay. And in my mind that satellite is now static. I know it’s moving, but it’s moving at the same pace as the earth. Are there satellites up there that are moving at different speeds? Is that how we end up with collisions? And I guess what I’m asking is, do they move at different speeds and do they make different tracks? Because my thought in geosynchronous is you put it up a certain distance, you get it to move at the same pace of the earth. And it’s kind of holding that piece of real estate for good. Does that deteriorate over time?
Rick Ward: So the one you mentioned really doesn’t. The geosynchronous satellites are generally pretty far out there and where they are it’s much less crowded than what we’re generally concerned with when we’re talking about collisions. The place where we’re most concerned about collisions is lower orbit. And that’s from about 200 kilometers above the surface of the earth to a bit over 500 kilometers. From there you get Neo that’s Leo, and then you get Neo, which is middle earth orbit. Not to be confused with middle earth, which is a different place entirely. And I’ve never been there.
Raymond Hawkins: Yep. That’s another show. We can get back to middle earth.
Rick Ward: And that is 600 and I’m not sure exactly where it ends, but then geosynchronous orbit is the one you mentioned. And that’s like 1000 plus kilometers away. I don’t really mess around with that one too much. There’s a current push towards doing everything in Leo. Lots of things are moving towards Leo because for instance, if you put a camera on a satellite and you take earth’s observational data, your camera works better if you’re closer. So you want to put your spy satellite or your any sort of picture taking satellite as close to the earth as you can. And below 200 kilometers, it’s going to fall out of the sky fairly quickly. You’re talking a matter of months.
Raymond Hawkins: Okay. And is that because the gravitational pull is yanking it down, it’s close enough that gravity continues to impact it, and it’s pulling it in?
Rick Ward: No, actually if you think about our atmosphere, you kind of say, well, there’s nothing above 200 kilometers, but that’s very much a gradient. You’re still going to have random bits of oxygen, random molecules of oxygen that are just up there and you strike them and you don’t notice it, but they slow you down. And if you strike enough of them, they slow you down enough to where you don’t have enough velocity to maintain orbit. In fact, if I was flying at like, I don’t know, 20,000 meters per second at like 10 feet above the ground, I would technically be in orbit, but I would be going through so much oxygen that I would turn the whole craft into a giant plasma ball.
Rick Ward: But if I’m going fast enough to not get slowed down by all the oxygen and trees and people and mountains and cows that I’m going to be hitting, I would technically be in orbit. So if you want something to de orbit, just get it down below 200 kilometers and it’ll be gone in a matter of months.
Raymond Hawkins: And it’s running into stuff.
Rick Ward: Yeah. It’s running into all that oxygen. And then it just slows down and falls out. And that’s where all the cube sets end up. They have these short missions that are at a hundred and something kilometers above the earth. It’s also cheaper to put stuff there.
Raymond Hawkins: All right. So we’re putting stuff in low earth orbit, stuff that wants to see the surface take pictures. And you say these satellites are dumb devices. So as we think about OrbitsEdge, wanting to put compute up there, because you’re telling me today we monitor those satellites from earth, having data, having compute power up in that low earth orbit, tell me what advantages are we gaining and what are we able to do by getting compute up there?
Rick Ward: For sure, there’s really two constraints that earth observation, and I’m using earth observation a lot because that is kind of our low hanging fruit as we see it, that is the first place where we feel we can make a difference. There’s a lot of other places we can make a difference, but those customers don’t exist yet. So we kind of have to focus on things that we can do today. So one of the issues that EO companies deal with and everybody deals with, is bandwidth constraints. It is not easy to get data down to earth. The frequencies are regulated by the FCC and other international bodies. You have to put in lots of paperwork, takes a year plus to get authorization to transmit stuff down and back up. And it costs a lot of money, they have auctions for bandwidth. And it’s not a trivial thing. So anybody up there-
Raymond Hawkins: So just like we sell spectrum here on earth, they sell spectrum into space?
Rick Ward: Exactly.
Raymond Hawkins: What I hear you saying is it as expensive? Is it a cost issue, a bandwidth issue, or just an entitlement issue by the time you get authorized to do it? Are you paying these government entities to move the data? So, I can already see where this one’s going, but help us get our arms around what the limiting factors are getting data down.
Rick Ward: You’re paying for the privilege of using the spectrum. And no matter how much-
Raymond Hawkins: You’re not getting a [T-Mobile 00:18:28] bill.
Rick Ward: Yeah. Actually, you also have that too. So you pay for permission, then you pay for service. So I can transmit my signal down at the earth that doesn’t do me any good. I have to transmit it down to a receiver. And that’s a ground station and ground stations, bear in mind as my satellite because you’re in a satellite now you’re a satellite. You only see one ground station for a few minutes.
Raymond Hawkins: It’s orbiting around the earth.
Rick Ward: Exactly. So if you have 10 ground stations that you have contracted with, or your company owns 10 ground stations, you have 10 opportunities to send data down. If there’s cloud cover, if there’s bad weather, if there’s for whatever reason, then you’re not getting data down to one of those ground stations or through you, whatever.
Rick Ward: So there’s always that constraint. And a lot of times the satellite doesn’t know whether or not the data ever made it down. It just sends down the stream and-
Raymond Hawkins: It broadcasts but it has no confirmation.
Rick Ward: Yes. And sometimes you have to send up and tell it to do it again. I actually had a conversation with somebody the other day that he made a good analogy. He said, “You’re filling buckets and pouring them out.” So yeah, you fill up the bucket of pictures, you pour it out and then you fill up another bucket of pictures and pour it out. So you don’t know whether or not anybody caught it. Maybe some of the water was dirty. Maybe some of the water just landed on the wrong spot, but it’s gone. You’re not worried about it. You get what you get and they get what they get.
Raymond Hawkins: All right. So I got low earth orbit satellites, doing earth observation, paying for the permissions to be able to broadcast back down to earth. They are then paying for a service for satellite stations, receiving stations to pick up that download. And the more of those that you have, the better chances or the more numbers of times in each orbit you can download. And where I see this going is you guys at OrbitsEdge, I’m assuming, your plan is to have sufficient bandwidth down to earth and sufficient service stations or satellite receivers on earth to be able to provide people quality communications. Is that part of the first stages of this?
Rick Ward: Kind of, not exactly. So there’s a few things we can do once you have compute on station. Maybe some of the pictures that your satellite took were of cloud cover. Maybe they took a picture maybe over a grid coordinate A, it was raining and there’s nothing to be seen. Maybe for whatever reason, the images is just bad. So we can filter that sort of stuff out. Say image A was bad but image B is good, which is, let’s say adjacent to A, but not A we’re going to send that down instead.
Rick Ward: Another thing we can do is change analysis. So you take a picture of my house and most of the time you’ll see that nothing has changed, the house hasn’t changed, the trees haven’t changed, but maybe you’re interested in whether or not I’m home or not. And you might see that my vehicle was gone or there, or I’ve put it in a different place, or I’ve done some sort of change to the yard. I’ve put in some flowers here. What if you could just cut out the parts that are different from the previous pass and just send those parts down. That’s called change analysis.
Raymond Hawkins: Change analysis. And you’re doing that in compute in space.
Rick Ward: Right. In that scenario, let’s say, again hypothetically, that 90% of the image is different, sorry, the same for what it was the previous pass for that same bandwidth. I could theoretically send down 10 images of change analysis versus one image of raw data.
Raymond Hawkins: Of 90% of the same stuff. So it’s a lot like deduplication in a storage device on earth. Why send over and copy stuff that didn’t change? Just send me the changes. I got it.
Rick Ward: Exactly.
Raymond Hawkins: Good stuff.
Rick Ward: And you have to look at it, the end user doesn’t care about the data. He cares about making a better decision and that decision is information driven and he wants, so what’s different? What’s new. What contextual clues do I get? The more complete the picture can be for the end user, the better the decision quality. And that doesn’t necessarily mean he needs the complete picture or she. So yeah, another thing we can do is quite simple, it’s compression and encryption. There is not a lot of capacity for either of those functions, given the current state of computing space. But if we can offer compression and encryption, then we can improve both of those things at the same time.
Rick Ward: So with compression, you can get stuff, you can get data orders of magnitude more dense than you could get it in raw formats. So that’s a big deal. Now you’re talking about hundreds of times more data imagery than you would have without compute capabilities. And in order to get hundreds of times, you would theoretically need dozens more satellites that are all equally dumb as the previous generation. So you can really turn it into a force multiplier.
Raymond Hawkins: How much is a consumer of earth observation?
Rick Ward: There’s actually a couple of dozen.
Raymond Hawkins: Weather.
Rick Ward: Oh, okay. So what are other avenues. There’s actually a couple of dozen companies that are working on that either with their own satellites or using somebody else’s satellites. The applications for it are farming, that’s a big deal. It’s good to be able to give the farmers insights, especially these large corporation farms. It gives them better insights on when to plant harvest and put down pesticides and stuff. For instance, you can do spectroscopy. You can do spectroscopy by satellites, which means I can see what kind of molecules are out there. To that end, they can actually see the ratio of crop to weed in a field. So that’s some pretty sophisticated stuff. And we’re getting into multi-spectral imaging and hyper-spectral imaging where all of that stuff is you’re talking like, again, orders of magnitude 10 times, 100 times more information per square meter than what previously existed.
Raymond Hawkins: Okay. What’s the future of OrbitsEdge? What’s the future of computing space? When we talk about edge computing, this is really edge computing. So tell us what’s coming.
Rick Ward: So first off, we need to go into space. So we’re not actually going to be the very first high powered computer in space. Our partners at HPE are currently flying a mission called Spaceborne 2, which is, you might’ve guessed, subsequent to Spaceborne 1, which ended, it was about a year and a half from ’17 to the end of ’19, roughly. And they put a high powered computer, a micro data server on ISS. And it functioned beautifully for a very long time. In fact, they got their ride back home bumped a little bit. So they had to spend six more months in operation they planned on and it did well enough that now they’re flying a second mission up there, and this is going to do-
Raymond Hawkins: 615 days in space?
Rick Ward: Yeah.
Raymond Hawkins: I think that’s the number.
Rick Ward: Yeah, about three-quarters that was turned on because they had to-
Raymond Hawkins: So more than a year, it was able to run in space?
Rick Ward: And it ran when they brought it back down to earth too. So that’s pretty remarkable. So their second mission is working in conjunction with other experiments that are being conducted on the ISS and they were wanting to do well, what can you do with processing? They’re wanting to do all of those things for other experimenters on ISS. One of the issues is they’re doing genomics testing. They’re wanting to sequence things. They’re wanting to see how astronauts are fairing. They’re wanting to see how experiments are going, biology experiments. So they sequence a thing, a genome, and it might take three months to send all those A-Ts G-Cs down to earth.
Rick Ward: And so the principal investigator of this mission, Mark Fernandez, he was talking with some of the people who are doing the work on the ground. He said, “So what do you actually need?” And they said, “Well, once we get the genome, we’ll put it through a computer and we’ll turn it into a spreadsheet. And we only need like a few kilobytes, but we have to get all these gigabytes or more might even be terabytes before we can get the spreadsheet.” And he’s like, “So what if we just sent you down a spreadsheet?” And that turns it from three months of transmit time to 10 minutes. So stuff like that is a big deal.
Raymond Hawkins: So the trick being they got to do the compute up in space. That’s the trick though, right? The computes got to reside up where the data got. Yeah, I get it.
Rick Ward: Yeah. So that means, instead of let’s do four genomes a year, we could do a couple of 100 if we wanted to. It’s a matter of how long it takes the sequencer to operate. So there’s a lot of things that you can do that we’re not ready for yet. We’re looking at ourselves as a piece of that facilitates other things. So it looks like we’re serious with Artemis. I’m very proud to see that the new administration is breaking with previous tradition of gleefully canceling the previous administration’s big moonshots last mar shot and redirecting to the opposite one so that nothing ever actually happens. I’m very heartened to see that the Biden administration is supportive of Artemis. And when you get to the moon, you will find that it’s about a 1.2 second the lag between information going here to the moon and back to here, or the other way around.
Rick Ward: And if you’re going to be doing a Rover or you’re going to have humans on site, that’s doable. That is sufficient that you can get some stuff done. If you’re going to have a Rover that operates meters per day, where you take a picture, roll a little bit and take it out of the picture and wait for feedback. Sure. But if you’re going to do something where you’re expecting to travel kilometers per day, you have to have onboard compute. It has to be at least a semi-autonomous vehicle where it can think for itself, it can make some decisions. It can save left or right. Faster, slower, all by itself without me having to directly joystick control it.
Rick Ward: And as you get farther than the moon, think about it, the moon is the closest target we have in the whole solar system. There’s not another body that is closer than the moon. So the farther you get, the longer that time lag gets. Mars you’re talking about from like 40 minutes to a couple of hours or so, depending on the time of year. So it really does become untenable to not have local compute. And the more we get out into space, the more our need for compute will grow.
Raymond Hawkins: Makes complete sense. Well Rick, I got to tell you this is awesome. It’s fascinating to understand and hear the idea of what the edge and what compute in the real edge, the way we out edge, makes me think of my time as a kid watching Star Trek and the places men have never been before. I don’t remember the intro, but that’s the kind of stuff we’re doing. And you guys enabling us to do it with compute power and all the things that compute power then enables, just fascinating stuff. Frankly, Rick, if you’re willing, I’d love to record another episode with you there because I got about 70 more questions, but can’t fit them into this episode. So thank you for joining us Rick Ward, CTO and founder of OrbitsEdge, literally taking compute to the edge out into outer space and fascinated to hear you join us another time and talk about more and more about where it’s going, what’s out there and what having computing space can do for us. Rick, thanks so much for joining us.
Rick Ward: Thank you for having me and I can’t wait to do it again.