Project Natick – Underwater datacenter with live cameras | Hacker News Reader

Project Natick – Underwater datacenter with live cameras

natick.research.microsoft.com

81 pointskerng33 comments

It's intended to run without maintenance for five years. That's been tried before. InfoSeek did it in the early days of search engines. They didn't repair machines; the entire module of some number of racks was replaced at end of life. But that idea seems to have fallen out of fashion. That Microsoft is trying this again is significant.

It's running in dry nitrogen. That has advantages over air. No corrosion, for one.

So where does the heat go? Judging from the FAQ my assumption is it's transferring it into the surrounding water? How many of these before it starts becoming a local ecological problem?

There is dramatically more heat capacity in the ocean than in the atmosphere. Many power plants reject heat directly to the ocean, and clearly they do produce a local hotspot, which is much argued over by ecologists.

There can be some valid local ecological concerns. If it emits into a spot which is particular important to some species it can become a problem. It can also create ecological activities, where some species congregate around the outlet because they want the hotter water. This isn't necessary good, but again, it's a local occurrence.

Globally, using ocean water for cooling is almost always ecologically better, simply because it's more efficient than alternatives that are either constrained in terms of the usable volume flow of water, or must use a less efficient medium for heat transfer. Eventually, all heat we produce will be radiated into space, but if an activity indirectly produces greenhouse gases, then it warms by a very large multiplier compared to its direct heat production.

You seem to have a good understanding of this subject. Could you explain to someone who is much less well-versed, why it matters if some species has an important area that it can't go to anymore and, say, stops producing in that area? (Goes somewhere else.) I mean is it some great tragedy not to have some species in some particular place?

I want to kind of add a disclaimer to my question. At the limits of course I am not questioning ecological conservation - the Earth would both suck if it only had humans and our crops and livestock in it, and be thrown completely out of equilibrium so that it would suck for humans too. But I assume that the effect you're talking about is where over time some tiny little place becomes bad for some species, so they stop going to just that one place. Likewise you mentioned the opposite effect where "It can also create ecological activities, where some species congregate around the outlet because they want the hotter water" but this doesn't seem like some objectively great thing - wow, we have all these fish hanging around what amounts to a thermal spa for them, fantastic - but more like just a neutral fact.

This is my impression, that it is what it is, so I'd just like an explanation of the strong moral component or why this is important - since you seem to understand this. (Again, I can do part of the reasoning myself since obviously I wouldn't choose a barren lifeless planet with only humans and stuff we eat, on it, I'm just having trouble extending this reasoning to tiny local perturbations.)

Not an expert by any means, but do dive some. I think one concern is that not all life can easily move or be recreated. Coral reefs probably being the biggest one there. They have huge biodiversity and are incredibly important in the health of the oceans but are also incredibly fragile and sensitive. It isn't hard to imagine a few degrees difference in water temperature affecting these and they don't just "pop up" again somewhere else, they are often thousands of years old.

Most of the ocean is pretty uninteresting though, so I would hope this project is looking at installing these datacenters in deeper waters (where as you said the life can more easily migrate) rather than the shallower waters that reefs inhabit.

I have no experience with data centers, but a little bit of experience with ecological impacts of pass-through nuclear cooling, i.e., water is pumped in from a body of water as cooling for a nuclear power plant, then pumped back to the body of water.

My guess would be that the heat diffusion system for the data center could be designed in such a way that you could have a very large installation without a large ecological impact, as long as a modular footprint like the one pictured is used (think of reef systems that spring up around shipwrecks). However, this will very much depend on the ocean currents at installation to quickly transport heat out of the area. Most species can handle a couple degrees (Celsius) variation in water temperature, and given the pure huge heat sink that is available by being completely surrounded in water. Species that prefer slightly warmer temperatures would probably establish a new foothold in the areas around the datacenters.

As for the "why it is important" bit. My conservation stance comes down to the precautionary principle. We still know so little about the ecosystem of the ocean that has been working pretty well for millions of years. We should endeavor to change as little as possible because we don't want to disrupt the complex feedback system that exists. Despite all the environmental impact studies that would be performed, we don't know what the possible cascading effects could be, if any.

It makes more sense to heat things that need heat. For example we have a few public pools heated by datacenters [1]. Alternatively you can also heat homes like they do with the heat generated from burning trash. [2]

[1] https://www.tecchannel.de/a/oeffentliches-hallenbad-von-rech... [2] http://www.kezo.ch/technik/produkte/

I know that Google also has water based installations where there are basins to let the water cool down before it is reinjected into the local system.

Few companies could and would throw money at a project like this. Environmentally friendly datacenters are needed.

Dumping heat in the ocean is not environmentally friendly. That heat could have been used to heat up buildings.

Almost all the energy we have, ultimately comes from the sun. As long as we can covert energy in a clean way, there is nothing inherently environmentally unfriendly about spending energy on one thing that could have been used for another thing.

Yes, but we need to be careful when we discharge excess heat. The sudden addition of unnatural heat can be an issue for ocean ecosystems. https://en.wikipedia.org/wiki/Thermal_pollution

Something tells me that thausands of heat exchangers where you produce hot and exachange it for cold, at the bottom of oceans which already are recording highest temperatures in modern history, is not much of envrionment friendly.

I would be very surprised if the waste heat had any significant impact more than a few meters/yards from the container. Water has a really high heat capacity and the oceans are large.

Almost all the energy that is heating up the ocean is coming from the sun - we've just put a thicker blanket round the earth to trap more heat. Waste heat from human activity is a tiny part of the global energy balance.

Seems like it would be really annoying to service if something breaks and needs to be replaced.

Guess you would just remotely disable any part that breaks. Once enough stuff has broken you just replace the whole thing.

Designing it would be a little bit like designing something that is going into space. You'd need to be very confident in your ability to remotely manage it over a long time frame.

I'd imagine they would want to build this cheap enough to just replace the entire thing when something goes wrong. Anything else seems very impractical.

It fails in place, then you replace the internals every five years as a single unit. After 20 years, you recycle the whole thing.

They're going to save soo much money not having to lease space in major cities along the coast because of this. Hardware is pennies on the dollar in comparison.

The leasing cost is buying land (ammortized over 30 years) + maintaining land and then compare it against buying specialized equipment, hardware, and maintenance staff for underwater projects.

I guess with scale, the price goes down.

I'd assume you still need a small datacentre close to it, i.e. on the cost, for routing, backup generators, maintenance, etc.

"Small" being the operative word. A room vs. a building.

Well, that depends on your definition of small. This is what I believe the diesel generators of a typical datacentre look like (+fuel tanks):

http://www.dchuddle.com/wp-content/uploads/2014/08/Backup-ge...

I'd imagined that it was a shipping container on floats, but it's a pressure vessel (keep that seawater out) filled with nitrogen (to stop corrosion).

So many fish visible on the live cameras. I suppose the container surface has become a source of food for the local ecosystem now.

If you drop anything interesting he ocean it will become a rich ecosystem.

https://en.m.wikipedia.org/wiki/Artificial_reef

Very interesting biology and physics project. I wonder how the layer of life building up on the warm surface will affect the heat flow.

Is it just me or does that website load insanely slowly?

Wow, yeah. I counted out 16 seconds. I'm using umatrix, so I was starting to think I'd just blocked something needed for the site to load by default, but decided to give it a bit just in case.

Still don't see the benefit of submerging and dealing with all of the engineering challenges that causes.

At 240kw of power consumption, that level of energy density must be using some sort of forced air/water cooling system.

So if you're already pumping/circulating the water from outside, what is the benefit of submersion? Why not just float it on the surface?

> So if you're already pumping/circulating the water from outside

»Natick datacenters consume no water for cooling or any other purpose.« (From the FAQ)

I guess if you're able to get the heat to the outer walls, then being submerged in water should take care of the cooling for you, without the need to pump in ocean water and circulate it.

Recirculating the water wouldn't be considered consuming.

They do have flat plate heat exchangers, commonly used on boats and either in direct contact with the water, or bonded to the hull, that have a closed loop cooling system to conduct the heat to the water.

But submersion of the entire vessel isn't needed for this to be effective, just sufficient surface area in contact with water to effectively disperse the heat.

There are many safer ways to do this than submersion. I will do the calculations later but I don't believe even the entire surface of the submersed data center would be sufficient to disperse the heat using natural convection and conduction. It takes 1000s of feet of submersed(or buried) tubing to effectively reject the heat in common geothermal/lake source home cooling systems, and they are typically in the 60,000 btu range(~18kw)

A submerged heat exchanger in a closed loop cooling system seems like a much better solution, if the reason for submersion is cooling related.

Weather.

Reply on news.ycombinator.com