Principal Program Manager Nuclear Technology | Hacker News Reader

Principal Program Manager Nuclear Technology

jobs.careers.microsoft.com

Wonderful! With investment into SMRs hopefully the technology can mature and achieve the economies of scale needed to be cost competitive for replacing oil and gas. We need all the investment we can get into next gen nuclear in order to decarbonize at scale as quickly as possible in coordination with solar & wind.

Previous discussion about 7 hours before this post: https://news.ycombinator.com/item?id=37627697 (thought my comment had disappeared, it's in the other thread)

IAEA Small modular reactors: https://www.iaea.org/topics/small-modular-reactors

"Small modular reactors: flexible and affordable power generation

... Global interest in small and medium sized or modular reactors has been increasing due to their ability to meet the need for flexible power generation for a wider range of users and applications and replace ageing fossil fuel-fired power plants. They also display an enhanced safety performance through inherent and passive safety features, offer better upfront capital cost affordability and are suitable for cogeneration and non-electric applications. In addition, they offer options for remote regions with less developed infrastructures and the possibility for synergetic hybrid energy systems that combine nuclear and alternate energy sources, including renewables. ..."

Time to put the cloud in space? Imagine an Azure datacenter on Ceres, fueled by fission. No used fissile material to dispose of (on earth)

1 hour RTT, compares well with Canada Central.

I know this is a joke, but I want to reiterate some of the challenges and shortcomings of this.

* Ping time would be roughly 1 hour

* Would require high-power wireless communication link

* Would take over 1 year to ship new hardware to datacenter

* No magnetosphere means increased hardware failure due to radiation.

Couldn’t you replace the magnetosphere with some sort of faraday cage-type setup? The other things aren’t necessarily deal breakers for various use-cases (e.g. off-site backups and other sorts of cold-storage situations where)

I'm being a bit hyperbolic, but Ceres is a ball of ice so large it's rounded by its own gravity [0]. I'm pretty sure you could engineer a shelter by using the waste heat of the nuclear reactor to bore a tunnel into the ice.

I wish I knew the math to calculate how much energy would be needed to send data to Ceres and back, because I think the transmission energy might be large enough that it would prevent the cold-storage use case. Reading and writing from flash storage probably requires less energy. That means this Ceres data center would be best suited towards low data, high computation, latency is irrelevant computation. Maybe brute forcing encryption keys or advanced fluid dynamics modeling.

[0] Ceres is 50% water by volume.

Perfect for cold storage then?

The entire benefit of Ceres is that you get access to nuclear power without having to risk radiation exposure for Earth. Cold storage is famously low power usage, the storage drives are powered off. It doesn't take advantage of the entire point of building a datacenter on Ceres. It takes orders of magnitude less energy to fill a SD card full of data on Earth and leave it here than it would to launch it into space in case we need to read from it in the future.

I was trying to be funny in my comment. But thanks for a very informative reply!

I mean with a fission reactor right there now much does the magneto sphere really matter?

Or are fission byproducts just that low energy when it comes to flipping bits?

Except through a decaying orbit.

Pinging microsoft.com [20.236.44.162] with 32 bytes of data: Reply from 20.236.44.162: bytes=32 time=2767923ms TTL=56

I am very impressed by the fact that this is an actual Azure IP.

Even if we could scale to Cosmos web services we’re still stuck with IPv4

You'd want to put them in Low Earth Orbit, not Ceres. Data centers (CDNs) in space would have 10ms ping times in basically the worst case (assuming a relatively high low earth orbit of 1000 km and hitting satellites at approximately 45 degrees elevation). In the best case they'll have ping times possibly as low as 3ms (satellites that happen to be directly overhead in their 400 km orbits).

You get direct connections to the servers without going through multiple local jumps and you aren't slowed down by the slower speeds of light in fiber.

Error correction, mass limits, and power generation are the difficult parts.

Deep Space Network needs to be a thing

You'd want to put them into Low Earth orbit, not on Ceres. We already have a bunch of old soviet nuclear reactors in orbit of the Earth.

[dupe]

More discussion earlier over here:

https://news.ycombinator.com/item?id=37627697

Reply on news.ycombinator.com