We've never even built datacenters using robots here on Earth

Interesting Moon experiment raises cool questions about disposal of hidden value


Opinion Making a call on the quality of a new idea in tech can be hard. But if you ask me, not in the case of Lonestar Data Holdings, whose plan to build datacenters on the Moon is literal lunacy.

Every detail of the roadmap, from tentative tiny proofs of concept to massive underground server farms built and tended by Moon robots, is priceless nonsense. From Apollo onward, every spacecraft has had data storage and network access. We have retrieved data held in New Horizon's 16GB filing system from Kuiper Belt object Arrokoth, 16 thousand times more distant than the Moon. Concept bloody well proved.

As for building bit lairs in lava pipes by robot, nobody's built a datacenter by robot on Earth yet. And nobody seems minded to try.

As the current rule of thumb is that landing a kilo of stuff on the Moon is around $1m, with an HP ProLiant DL380 G5 Storage Server clocking in at around 30kg, even FedEx suddenly seems like a bargain. Everything about space is very expensive, very difficult, and absolutely reserved for things you can't do any other way.

The biggest mystery about LoneStar is how on Earth it got $5m in startup funding, just enough to land a laptop, yet addressing that mystery is surprisingly useful.

The proposal has resonance because it seems to be an exciting science-fiction answer to problems we instinctively know exist. What LoneStar is good at is provoking a thought experiment. Its ultimate justification is that our home world is too dangerous a place to keep our most valuable data. Bombs! Earthquakes! Hackers! It's worth adding an unimaginably expensive extra tier to Amazon's storage offerings to make things really, really safe.

You might think you'd be better off with a bunch of high-reliability datacenters underground all over Earth, and you'd be right, but that's not the point. Everybody – individuals, companies, nations, and cultures – has their data crown jewels, and everybody is instinctively prepared to keep them safer than the more mundane. Identifying what those jewels are, how much they're worth, and how best to guard them is a very worthwhile exercise.

The actual Crown Jewels of the United Kingdom, a set of royal ceremonial objects kept in the Tower of London, illustrate the problems wonderfully. Their monetary worth is unknown but given the British Crown is the symbol of national authority, they're beyond important. Although easy to identify, it turns out that sticking them in a big castle marked "Crown Jewels Here" isn't entirely safe. They've been stolen, melted down, and sold, with a near-complete reboot needed after controversial 17th century leader Oliver Cromwell disposed of the lot. (The Scots cannily hid theirs over the Interregnum*, but bits kept getting stolen thereafter.) Putting your data equivalents in one super-safe place, even if it's on the Moon, will attract attention.

Expensive

The biggest disincentive to extra-special safeguards for extra-special data is cost. It costs a lot to classify and freshen the categories of data worth attention. Conversely, as bulk data storage pricing is inversely proportional to latency, the very slow storage tiers used for disaster recovery backups are the cheapest per-byte, so you might as well treat everything the same. And since there are legal limits to how long things must be kept, that's often the event horizon of the whole affair.

That attitude is shortsighted. It's perfectly fine, if you can't identify your crown jewels now, to keep everywhere they might be for later. But what? Some data might seem useless after milliseconds, like the market conditions you got fractionally before your competitors that let you do an advantageous trade.

Ditto last August's grocery bill, or the metrics of an app you've just retired. Yet data is never just data, it's also context, and as any historian will tell you, as data ages and its primary utility fades, the picture it paints about what was going on at the time is often increasingly valuable. That August grocery bill is a lot more interesting now than it was in January.

The terabytes of the daily grind encode signals about what an enterprise is doing right and wrong, not just for today, or the quarter, or even the last five years, but signals that can be read in decades to come.

Countries jealously guard their national archives; we have the tools now, from individual through SME to enterprise, to think the same way. Data that costs so much to obtain and process does not lose its value lightly, but it transmutes over time. Your future AIs will love it.

Even if you change nothing about data storage, thinking this way gets us out of short-termism and sharpens appreciation of long-term strategy and abiding value, and there are very few organisations of any size who don't need more of that. But the tangible benefits of having substantial archives of data for future AIs to sample, while hard to define, are equally hard to deny.

Companies thrive or die on the quality of how they perceive the market and the decisions they make as a result. The very finest record of how your outfit has evolved is there for as long as you choose to keep it.

But beware. Embedded context may not go your way. The finest, most gaudy gems in the UK Crown jewels, the Koh-I-Noor diamond and the Cullinan diamond, are there by dint of historically recent conquest, problematic symbolism that will in time make the controversy over the British keeping Greece's Parthenon Sculptures in the British Museum (where they are known as the Elgin Marbles after the British ambassador to the Ottoman Empire who took them) look like a game of cribbage. The current emperors of data, the Googles and the Metas, may like to meditate on this.

The ethics of where your most valuable data comes from is up to you to deal with, but thinking about it is also a natural byproduct of asking yourself what the true value of your data is over time, and what may be worth a moonshot to preserve.

There will be datacenters on the Moon one day, but only when they're the best answer to questions, not just the silliest. For now, the questions themselves, of risk versus value, of long term versus short term, of evolution versus mere survival, are worth every moment spent on them.

That's as true of a harassed data manager in a 10-year-old hundred-person company as it is of a 96-year-old head of a 966-year-old dynasty. Long-termism has its merits. ®

* The period in English history when the throne was vacant between the execution of Charles I in 1649 and the Restoration of Charles II in 1660.

Broader topics


Other stories you might like

Biting the hand that feeds IT © 1998–2022