AWS buys 100 diesel generators and that's just for Irish datacenters | Hacker News Reader

AWS buys 100 diesel generators and that's just for Irish datacenters

theregister.com

27 pointsBender29 comments

I really do believe that there have been millions of generators and solar panels sold, just because there "might" be a power outtage.

Like for instance here in Germany, they warned us of sad winters and blackouts all around and as it stands right now, there is nothing going to happen to either power nor gas supply.

So I think these have been bought under fomo and will most likely not be needed and sold again.

Solar in Ireland is going to be difficult, but wind is not. Its just a bigger investment that takes more space for something that "might" happen.

>Like for instance here in Germany, they warned us of sad winters and blackouts all around and as it stands right now, there is nothing going to happen to either power nor gas supply.

Can you elaborate on this? As an American I'm curious what your perspective is as someone who is currently living in Germany. What we're hearing (at least from commodity analysts) is that there is will be major energy supply shortages throughout Europe, particularly Germany. Not arguing one way or another - just trying to hear more perspectives.

Other German here. I can't speak for every country in Europe, but right now it looks like energy supply shortages in Germany are unlikely.

Gas storages as of today are 97%, and Europe as a whole is at 93% [1]. This is also the reason the gas price shoot up this much, because everyone was filling up the storages as fast as possible, no matter the cost. But due to the full storages gas prices are falling. Spain is even turning away LNG tankers because there is no more capacity [2].

Without any gas supplies, Germanies storages will last 2 months. This is incredibly unlikely to happen. There are LNG terminals being build right now and one is about to go online [3]. Besides those, there is still gas coming in from the Netherlands and other parts of Europe.

When it comes to electricity, there may be some trouble ahead if France doesn't get its nuclear plants online in the next months [4].

Another factor which helps right now, is that it's still kinda warm for this time of year and this winter is predicted to be on the warmer side.

The last point to make is that there are laws which mandate the shutdown of gas- and electricity-heavy industries to prevent any residental shortages. In an emergency nearly 50% of electricity and 50% of gas could be saved.

[1] https://agsi.gie.eu/?s=08 [2] https://www.dw.com/en/lng-tankers-stuck-off-spanish-coast-am... [3] https://www.uniper.energy/projects-and-cases/lng-terminal-wi... [4] https://www.euronews.com/2022/09/04/frances-edf-plans-to-res...

That's an interesting point about the mandates regarding industry. I never knew about that.

Thanks for taking the type to write this. Good info

If you've bought the panels, presumably you'd use them?

Often, utilities and laws make it irrationally difficult to use solar and grid at the same time. You'd need a grid tie system (if permitted), or loads that you can run off solar alone (not connected to your home wiring).

I can totally see the rationale for buying panels, and an inverter, such that I could feed into my home if the power went out - but not going through the mess of permitting, just planning to hook it up if the apocalypse does occur.

Assuming you want to use solar and grid at the same time, how would you avoid using a grid tie system?

You definitely need to match phase with the grid, or bad things will happen, and you definitely need it to not back-feed current into downed lines. Is there some other expensive thing the grid tie is doing, or are you proposing some other setup?

You’re correct - I’m addressing why someone would buy a solar setup and not use it. You’d need a grid tie to use it, and often the permits for those are slow to get.

But if the power goes out for an extended period in a real apocalypse scenario, and I’m getting cold, guess what: I’m turning on my solar and disconnecting the grid entirely.

You could have both grid and solar feed into something less regulated with no back feed. A stupid but easy to think of example would be Solar-battery + inverter for your load with a DC grid powered charger when batteries are low.

> Is there some other expensive thing the grid tie is doing

Legal permits and network regulations are the expense I would guess. Probably nothing particularly technically difficult.

And fuel to power them, and oil for preventive maintenance.

The title/article is clickbait-y. Data centers normally have diesel backups. The article is portraying information that is completely normal as exceptional due to ongoing circumstances. It'd be far more interesting to learn that Amazon DCs do not have diesel back ups as that'd be a huge deviation from the norm.

Most DC's I've worked on have back-ups that are roughly 2-ish megawatts per gen-set so AMZN is buying enough gensets for 200-250 MW of load.

Why go the petrol route rather than invest in batteries + a huge solar array or wind farm?

Are economies of scale not there yet for Amazon's data center power requirements?

Think high level, and review in low level. Battery’s energy density is not nearly closed to fossil fuels.

Question like this particular in a tech forum shows how much trouble we are in when actually dealing with environmental challenges.

Honestly what would be better for the environment? A lot of batteries that rarely if ever get used but still need to be replaced? Or storing a bunch of fuel in a tank?

The backup generator should rarely be used. The resources that went into the construction of the generator and tank is probably much less than would go into all those batteries.

I'm sure you're correct, but I'd love to see numbers to back this up.

The union of concerned scientists has done this for electric vehicles, which is really useful.

I'd also like to see it for low-hour-usage stuff, like tractors, lawnmowers, chainsaws, etc.

(Though, honestly, electric chainsaws are way, way better than medium sized gas ones, CO2 footprint be damned.)

The CO2 for the fuel is pretty easy to calculate. It's the 2-4kg or so per litre (depending on where it's sourced). We'll call the carbon footprint of a generator negligible as it's mostly Aluminum (20kg CO2e/kg new or 1kg/kg recycled and steel (2kg/kg), it's maybe 10kg/kW so max 200kg. It will be lost in our error for the dirty fuel source once we use 50kg or so of fuel. The carbon cost of keeping some fuel out of circulation is hard to quantify, and tanks are essentially free so I will call the capacity cost 0.

~3kg/kg for fuel is really convenient because when combined with ICE efficiency, it's about 1kg/kWh

This https://climate.mit.edu/ask-mit/how-much-co2-emitted-manufac... says every kWh of NMC causes 2-16t of emissions. I'm sceptical this applies in 2022 given that $130 only buys 5t of coal, but we'll use it as indicative.

So for our generator we have 200P + 4E + 0C where P, E and C are power, total energy used and hours of capacity respectively.

If we assume 2C charge rate and 4000 cycle life, our battery is 2400 * max(P/2, C) + 2400/4000E

So E = 600C + 6/40E Assuming P is small compared to C, then any capacity used less than ~550 times in the 8 or so years it lasts is better off as a generator.

This is a function of the carbon intensity of battery though. At 200kg (https://spiral.imperial.ac.uk/bitstream/10044/1/79316/2/Revi...) it's about 50.

In this case though the generator won't be generally running unless external power is cut off. So it won't really be creating CO2.

> that rarely if ever get used

Oh, they would be charged and that degrades them.

Also you can just buy more diesel and fill up the tanks but you can't do that with batteries.

These are just backup generators. Cheap to buy and own but not cheap to run. A sufficiently large pv/battery array would probably be huge.

The price of gas and electricity is so high that diesel looks relatively cheap.

The data centres are being located in the built up East of the country, near Dublin the capital city.
Land for solar farms would be expensive and to be honest I'm not sure if the sunshine here would make them viable.

I don't know why data centres are nt being located in more remote regions of the small island, presumably there are network / technical reasons.

It's financial. Land gets cheaper the further away you go but the resources (like people) get more expensive the further away you go, so it's a balancing act. Some will be in the build up parts, and some will be in the remote regions. Take a look at Sweden, there's expensive DCs in Stockholm, but there are also remote DCs way up north in Luleå.

Solar farms and batteries are a 5-10 year project.

Diesel generators are for powering your data center tomorrow.

Really? They're equally difficult for a homeowner to get installed.

@ChuckNorris89 wrote:

> huge solar array

> Ireland

Yeah, I figured there could be land mass availability issue.. maybe better to beam it from outer space or across the ocean from a different *-land?

I love the idea of a self-sustaining data center, hopefully someday we'll crack it in real life.

Capacity factor is still 12% in ireland. 7% on a mostly clear week in mid winter. Your solar array has to be about 30x as big as the floor area to cover net daily energy on most days.

Maybe down to 20x with bifacial modules and relying on grid or backup for a few hours a day during bad days in winter. Eyeballing some photos in the US, including the parking and setbacks you might be able to cover 10% of the power in winter or 15% year average

In some areas you can probably put up a 3MW wind turbine or three. That would substantially lower the cost and size of storage. Max density for an entire area is only 10W/m^2 though so it's only another 4% assuming the available land looks like some random amazon datacenter I glanced at.

Still, 15% ish net annual is a dent for doing the extremely energy intensive thing without expanding the land use (assuming it's a good wind site).

I wonder if in some future where 70% efficient five junction panels are available whether it would make it easier to cool. You'd effectively be taking the sun heat and putting it inside so the outside would be cooler.

I love that idea. I bet Amazon could plunk one down somewhere sunny, like in Arizona, US if they wanted to.

Reliability trumps all else in datacenter workloads.

Reply on news.ycombinator.com