I strongly support your project, although I´m not a programmer. In fact I´m coming from renewable energies.
However, the question:
Provided there are millions of end users on the safe network, using there consumer products or home infrastructure, what effect will the day-night-cycle have in terms of energy used for relocating data as a majority of the population of one region will more or less synchronously switch on or off their devices?
As far as I understand the safe network uses a mathematical relationship to identify neighbors for data storage. Although this seems to be by far the most secure approach, won´t it result in a huge data wave travelling around the planet?
At least there seems to be an optimization problem: How much energy usage generated by the network is justifiable for how much security.
Really good question and I think it would be desirable to have a discussion on this: optimising SAFE Network’s use of resources for sustainability.
I don’t have any immediate thoughts, other than to clarify one point. The closeness you refer to is not geographical, but is in xor space, so effectively random.
However, caching causes frequently requested data to move “closer” to the place it is needed in terms of number of hops - which would probably cause a net reduction in resource and energy use.
So I guess your question highlights the possibility of xor distribution being random, and therefore more “expensive” than if data could be stored closer to where it is most likely to be needed, at least in terms of the number of hops.
I don’t understand how feasible this might be, or how beneficial in terms of resources. My guess is, not so much, but it is just a guess!
I guess one thing we might like to consider is how well resource use is factored into node ranking, and how effective this is from a sustainability perspective, and whether this can be improved.
It is already catered for to some extent through human behavior (through our desire to reduce costs), but given how invisible these costs will be I think that could be improved on considerably in code.
Really good question and I think it would be desirable to have a discussion on this: optimising SAFE Network’s use of resources for sustainability.
I don’t have any immediate thoughts, other than to clarify one point. The closeness you refer to is not geographical, but is in xor space, so effectively random.
However, caching causes frequently requested data to move “closer” to the place it is needed in terms of number of hops - which would probably cause a net reduction in resource and energy use.
So I guess your question highlights the possibility of xor distribution being random, and therefore more “expensive” than if data could be stored closer to where it is most likely to be needed, at least in terms of the number of hops.
I don’t understand how feasible this might be, or how beneficial in terms of resources. My guess is, not so much, but it is just a guess!
I guess one thing we might like to consider is how well resource use is factored into node ranking, and how effective this is from a sustainability perspective, and whether this can be improved.
It is already catered for to some extent through human behavior (through our desire to reduce costs), but given how invisible these costs will be I think that could be improved on considerably in code.
When nodes come alive again, for all their being considered new, do they not still bring with them the data they had before?.. Will that not fairly neutralise such a wave??
The possible expenditure on relocating data is only part of the equation. Also would have to calculate the savings of double use the same machines, already in operation, to function as servers.
None. David said that offline vaults won’t be considered offline for many hours.
That’s by default necessary, for privacy and redundancy reasons.
And not knowing where the data will be requested from, the claim that it will cause a “huger” data wave than it would if neighbors were selected using some other approach, is not valid.
MaidSafe isn’t bitcoin, there’s no correlation between energy spent and work done.
Spreading data around doesn’t create any “problem”, it’s a feature and it doesn’t cost much more to send data from Europe to Middle East than it costs to send it to South America.
I think @janitor is referring to a recent comment by David in discussions about non persistent vaults. No doubt the details are still TBD though as part of the testnet.
Actually I was referring to an older discussion where someone asked a what-if type of question related to moving HDDs (with vaults on them) from one rig to another, and whether that would cause a loss of rank.