ipfs / notes

IPFS Collaborative Notebook for Research
MIT License
401 stars 30 forks source link

IPFS energy consumption #360

Open czanella opened 6 years ago

czanella commented 6 years ago

There's been a lot of concern lately regarding Bitcoin's growing energy consumption, which, according to https://grist.org/article/bitcoins-energy-use-got-studied-and-you-libertarian-nerds-look-even-worse-than-usual/ , should reach 0.5% of the world's electricity usage and would keep on growing.

If it's true, that's quite alarming. I wonder if there's ever been some study or estimate on whether IPFS is also a system that would consume tons of energy or not.

I know that IPFS is not exactly a Blockchain system, but I also know that fetching a file from an IPFS network is a much more complex process than requesting a static file from a HTTP server. Does anybody know if I should be worried about IPFS' carbon footprint?

Thanks!

Stebalien commented 6 years ago

In practice, using go-ipfs is probably less efficient than using most http clients and servers. However, that doesn't mean that the protocol is necessarily less efficient.

In theory, IPFS should be more efficient overall in because:

  1. If your neighbor has already downloaded a file, you can get it from them instead of fetching it from the other side of the internet again (routers do take power).
  2. Because we use content addressing, you can often avoid downloading a lot of duplicate data. This should actually help significantly in datacenters/clusters that need to distribute mostly identical docker/machine images.

Note: IPFS will definitely be less efficient for a simple get a new file from point A to point B over a reliable connection test but that's not a good real-world test-case.

novusabeo commented 6 years ago

I believe IPFS can and is more efficient overall depending on certain outcomes -- If nodes increase, success increases, exponentially with the growth of the mesh network. Efficiency is determinable based upon a transition from traditional PHP based server systems to code that enables the IPFS network to work and deploy without the need for server-side process specific code which is energy consuming. Many systems will benefit from a transition to this system and looking into new possibilities with a mesh network that can only operate within a distributed system as opposed to complete design and creation of systems, applications, etc based on existing infrastructures like HTTP and single entry point access to websites. Routers, web systems, these are a platform on which IPFS currently operates, but if a true transition were to happen, if the true possibility of mesh were realized and understood and utilizes and integrates more with less intensive blockchain systems then efficiency becomes scalable based upon use whereas existing systems are only efficiency scalable based on power consumption -- better processors, better servers, better networks, etc.

Elduderinooo commented 2 years ago

Hi here, I'd like to dig into this discussion as I'm looking for reliable data on IPFS (network of nodes) carbon footprint. IPFS as a protocol would'nt account for much but what about comprehensive data of the underlying storage solutions used to run IPFS nodes as services (such as Pinata). Anyone to direct me to some calculations or sources of data I could dig into ?