Cloudy Skies: Cloud Computing in the Era of a Space Internet

Cloudy skies have often been a metaphor for sadness or an unlucky streak. Just ask Eeyore of Winnie the Pooh fame. Conversely, the clearing of the clouds heralds new beginnings. Far above our atmospheric clouds, plans are underway for the installation of an internet in space. What may sound like the plot to a science fiction movie is something for which work is actively underway. NASA’s LunaNet is envisioned to support Artemis astronauts and future semi-permanent habitations on the moon. Its work on Delay/Disruption Tolerant Networks (DTN) is paving the way for an interplanetary internet to be used for communication, data transmission, and position, navigation, and timing (PNT) for space operations. As human space flight begins its resurgence and space exploration tools become more sensitive, a logical question about cloud storage in space must be asked. Will it be feasible to send all of this data back to Earth understanding the communications lag realities that will only get worse as we travel farther? Envisioning an internet on the moon or in the broader solar system necessitates envisioning a new kind of cloud computing, one whose space will be at an ultra-premium and around which we should be designing computational and artificial intelligence (AI) resources for future missions.  Clouding our View   Cloud computing is so ubiquitous on Earth that most don’t realize when they are using it. According to CloudZero , the cloud computing market experienced a 635% increase from 2010 to 2020. Probably because of the 2.5 quintillion bytes of data the world generates every day. Individuals, startups, governments, and multi-national corporations use cloud services to reduce their costs of having on premise servers. However, the “cloud” is not this ephemeral mass that passes harmlessly overhead. It is really racks upon racks of servers housed in one of countless data centers that sit on physical space on our finite global land. Far from a comment about their environmental impact, the point is that having a cloud means putting those 2.5 quintillion bytes of data physically somewhere. On Earth, we’ve scaled this operation to the point that individuals can gain easy access to at least some cloud space. Even a free Gmail account comes with 15 gigabytes of cloud storage, also for free. This scaling has made us forget what it really takes to have cloud computing and storage because it is so easily accessible to all of us. But just like most things in space, cloud computing will not be so simple.   Clouds Gathering   In space, our activities have been growing since 2010 as our satellite launch rate has increased exponentially. The majority of that activity has been to launch into Low Earth Orbit (LEO) with the intent of relaying information back to Earth. Communications, Earth sensing, and other services come from LEO and in some cases direct to users. There is currently no need for a scaled cloud computing option in space for this reason. Images from the James Webb telescope do not need to be stored in space, they are transmitted back to Earth and stored here. This is changing.   Frontier Foundry’s Substack is a reader-supported publication. To receive new posts and support our work, consider becoming a free or paid subscriber. As we push beyond LEO into cislunar space and onward to interplanetary space, we will do so with both humans onboard and with a new generation of scientific sensors that will provide us unrivaled data and insights about our universe. Many of those insights and data will be sent to Earth so humans can analyze them. We must also leave open the possibility that data collected in faraway missions will need to be used by the astronauts on the mission as they carry out their duties. Payload constraints do not permit large server racks for data storage at scale to be included on spacecraft. Modern spacecraft of course have much more computational power than previous generations such as Apollo, but for the scale of incoming data, our concept of data storage and processing must change.   The logical answer to this challenge is to leverage the coming lunar and solar system internets to create cloud computing in space. Just like on the ground, this will mean some number of physical servers that must be hosted somewhere. Perhaps a satellite that orbits the moon or Mars. Perhaps servers that are part of the semi-permanent habitations in the same places. But given the costs per kilogram to launch ruggedized servers into orbit, such cloud space will necessarily be at a premium. We will not be able to absentmindedly upload photos, videos, or other data to the cloud and forget about them as we do on Earth. Cloud in space will be reserved for the most mission critical data sets and may not be available to process all the data that human spaceflight is certain to generate.   AI to Support Space Missions   As data is collected especially by missions with humans onboard, there will be a need to analyze quickly and accurately what is collected to make mission decisions reserved for the commander, not the teams on the ground. On Earth, we can count on endless cloud space to house data that can be analyzed as needed. Instead of depending on a future cloud in space, mission planners should be looking toward non-cloud native AI solutions to preserve computational capacities and increase security.  Computational Capacity : Just as it is not reasonable to imagine that data storage space will be immediately available in future space internet architectures, it is similarly not feasible to assume that computational power will be immediately available. The use of AI in space has for years been something that sounds great, but often suffers from a lack of a true path to implementation. Some have blamed payload constrains but the new generation of AI is smarter about computational loads allowing it to run even on an individual laptop without connectivity to public cloud architectures. Non-cloud AI integration into future space missions will give commanders the ability to run calculations and quickly analyze data that can make the difference in successful and insightful missions. Doing so in a way that minimizes the need for limited computational power will directly contribute to the sustainability of our new era of spaceflight and to the space economy overall.   Security : Space is militarizing despite the best efforts of many over many decades. As space is now viewed as a contested military domain by the US and other countries, mission planners must consider the potential for cyberattacks against space assets as a threat vector. That means that the data we are certain to collect in our missions is at risk and so would be a future space internet and cloud architecture. Mission commanders can introduce a measure of security into their operations by using non-cloud AI to do their calculations without transmitting data in a way that it may be intercepted or corrupted. The final output can still be transmitted to Earth but the processing of it can be done locally in a way that increases security at a critical moment in our space activities.  The gathering of clouds around a future space internet need not induce Eeyore reactions in us. We should be excited about what we are building and the movements toward further exploration of our solar system. Cislunar and interplanetary flights necessitate communications, computation, and data storage, but we should do so smartly. We’ve learned lessons on Earth that we should apply as we build in space. Chiefly, we cannot depend on an endless supply of cloud architectures in space and that will impact our ability to store and process the data we are soon to generate. Security is also a primary concern as we see other space powers making provocative moves in space. This would lead many to believe that some capabilities should not be used due to the environmental constraints. But space as always challenged us to think differently and as we think about the use of AI in space, we need to do as we’ve always done. We need to think about AI not as we think about modern, cloud-based large language models (LLM) but in an innovate way. AI is being built that is secure, privacy preserving, and non-cloud native and that’s how we should think about this issue.   Our space internet will look different than our terrestrial internet and so will our future space cloud. AI in space will likewise be different and we need to embrace it now so that we can build the foundations of our new era of human space flight and exploration in a way that assures the security of our brave astronauts and brings humanity the maximum benefits.   Frontier Foundry’s Substack is a reader-supported publication. To receive new posts and support our work, consider becoming a free or paid subscriber.