HN Explorer

Ask HN: Any vague estimation if peertube has lower carbon emissions?

1 points by jokoon ↗ HN
I'm part of a small french group of technicians/engineers who tries to examine alternatives when attempting to transform the whole economy to limit carbon emissions. It's a very large subject.

Video traffic, would it be snapchat/youtube/netflix/vimeo/etc uses a lot of dedicated bandwidth with expensive servers which usually have a very good availability and reliability, but at the cost of servers and energy.

Peertube is a totally different model which doesn't really allow users to view a video with a similar quality or latency (the video might take more time to load), but it does work and the platform use client upload capabilities, which would be better if a video is very demanded.

The problem when studying carbon emissions of IT is that there are too many factors:

* equipment of the client

* equipment of the server (on top of housing cost, cooling costs, maintenance etc)

* equipment of the ISP

* Energy spent by the ISP/server/client

Those parameters are even more difficult to estimate in a P2P environment.

I'm curious if peertube would be more carbon/energy/bandwidth/hardware friendly in general. Do you also think that a P2P model would also be better for the environment, potentially or in reality?

9 comments

[ 4.3 ms ] story [ 11.0 ms ] thread
tsjq
The less we use, the more environmentally friendly it is. Anything.
phillipseamore
Energy use of video services is greatly exaggerated.

P2P likely just moves the energy use to less efficient devices. It might also use more bandwidth since re-transmission is common (peer goes away halfway through passing on a segment so it needs to be requested from another peer).

If a P2P network isn't aware of it's network paths the traffic might also have to traverse a lot more infrastructure than a nearby datacenter (e.g. UK viewer peering with DE, FI and FR peers).

jokoon
The statistics for the bandwidth taken by youtube/netflix are quite revealing...
phillipseamore
How? Both deploy caches at IXs and ISPs so most of the bandwidth is just going the last mile.

Bandwidth usage != energy usage unless the usage is such that it requires upgrading equipment or other infrastructure, but that might actually decrease energy use (a 5YO switch is less efficient than most current models).

The energy consumption of a network switch is about the same whether it's pushing 10gbps or 100gbps.

jokoon
The quantity of hardware required is also a source of carbon emission, energy is not the only parameter
phillipseamore
Netflix deploys three versions of their caches to IXs and ISPs (their called Open Connect Appliances).

The smallest are 1U, up to 112TB of storage, draw 250W and can push 10gbps (~1700 simultaneous viewers in HD).

The largest are 2U, up to 360TB of storage, draw 500W and can push 200gps (~38000 simultaneous viewers in HD).

That's comparable to just a few typical desktop PCs. And perhaps just a single decked out gaming PC.

Thinking about this, since most of the power usage is from client devices, I'm now wondering if gaming uses way more power globally than video (especially since bandwidth is not a good indicator of power usage).

jokoon
Gaming does use a lot of power because GPU are very power hungry. The Nintendo switch is one of the best gaming platform, peaking at about 50W when gaming PC requires about 500W to 700W PSU.

For video, most of the carbon is emitted while building hardware, so the question is mostly about hardware durability, which would be a good question. Another interesting question is the carbon cost of bandwidth:

* wireless antennas takes a lot of energy, especially 5G which requires a lot of new hardware.

* I'm curious about the energy cost of wired internet, optic fiber installation.

Generally, the problem with IT is mostly hardware and quick iteration of software that requires new hardware. It's a very chaotic and vicious cycle. Fixing it might obviously reduce profits of IT companies.

phillipseamore
A 4/5G base station setup uses from 1 to 3kW. That might be very little pr. user.

Fiber is very power efficient since it only needs powered equipment on each end and the fiber itself is passive (over extreme distances, 100km+, repeating might be needed but that's almost exclusively submarine cables).

POTS/Copper (xDSL etc) are not very efficient since every active (as in available, not as in being used) line needs to be powered on 48v.

In my experience HW is usually either renewed to get energy savings or because it does not meet demand anymore. A network is most often designed for more demand than is expected so it's usually about energy savings.

With P2P there would either be the same requirement for networking HW or possibly more. Demand for bandwidth would not decrease with more P2P usage but rather diversify (likely requiring more interconnections = more HW).