Data persistency and APTs

This could be a difficult question to answer. So, please brace yourselves:

What if whatever entity throws a billion dollars to support an APT to make sure they create, over time (months or even years, perhaps?), a huge number of nodes under a regular appearance (even through regular users/people/spies) to hold a significant amount of data. If Maidsafe has quantum-resistant E2EE, the problem of the data being accessed could be not in question, but then there could be the risk of a sudden black-out if they all stop their connectivity at once, right? Or worse, if they deleted their storage altogether at once. Like, the data could be lost forever for the victims of this APT. Had you thought about this before? Like, the world’s future could depend on the answer to these questions, if Maidsafe gets big enough.

1 Like

The same problem exists for a country like China when finding it citizens have massively adopted Safe as their way to access web pages/data/news decides to totally whitelist their internet to the outside world. Statistics show some very big files may lose some of their chunks. But even China is not going to be massive enough to destroy the usefulness of the Safe network. And they would represent a bigger threat than any billion dollar investment.

Lets take just a billion dollars and see how that would compare to 100 million people running 20 nodes each from home.

Some of what is needed for billion dollar plan

The billion dollars has to buy

  • machines to run nodes
    • maybe RPis at 25-50 dollars each running 5 nodes each and SD cards that die each year with the load
    • maybe small PCs at around 150 dollars including drive and each running 20 nodes
    • larger at >500 dollars running 40 nodes (there are limits and diminishing returns )
  • networking infrastructure
    • one switch for each 8 machines at 100 dollars (not the cheap and nasty as these have to work long term
    • one router/switch combo at 250 dollars for each 8 switches
    • one internet connect for each router with enough beef to handle the 40 machines with upto 2560 nodes
  • staff to not only install machines but to fault find when monitoring s/w detects an issue, or internet issue occurs
  • Aircon, buildings, substation for electricity, power backup
  • electricity
    • even at the world’s cheapest of 8 cents per KWH it will still rack up a decent bill. Even RPis are 5-10 watts each and switches and routers are more
  • corporate fees, tax, etc
  • Yes earnings will offset some of the recurring costs that home users already pay for so nodes just use that.

what is needed for average home user with a PC (a large portion of homes in many countries)

  • their PC to run 20 to 40 nodes
    • this they already have - no cost
  • their networking infrastructure
    • this they already have - no cost
  • their internet connection
    • this they already have - no cost
  • rent/electricity/air-conditioning/etc
    • they they already have
    • electricity cost maybe higher due to PC on for longer. Lets allocate 0.1 KWH per day for extra usage and average at 20 cents per KWH. 2 cents per day giving approx 7$ per year. But this should be more than covered by the earnings.

Cost per internet connection for Billion dollar plan (one year worth)

  • 320 nodes for RPi solution
  • 1280 nodes for small computer solution
  • 2560 nodes for better than average PC (better CPU/Ram speed than average home PC)
  • 1 internet connection - $1000 (cheap plan for reliable internet)
  • 1 router/switch - $250
  • 8 switches - $800
  • 64 computers
    • $1600 for cheapest suitable RPi solution. - $5 per node
    • $9600 for small computer solution - $7.50 per node
    • $32000 for bigger PCs solution - $12.50 per node

Totals are $2050 plus machine costs for hardware.

  • $3650 for RPi solution - $11.40 per node
  • $11650 for small computer solution - $9.10 per node
  • $34050 for bigger PC solution - $13.30 per node

Ignoring electricity and staff and rent the RPi solution is not the cheapest when scaling up, but the small computer. Even when you consider the sizes, it takes 4 RPi to match one small computer and the small computer is similar in size and power of 4 RPi. So for this exercise the small computer wins hands down for costs (elect/space/heat/etc)

What compute/network can we get for 1 billion dollars

Now since the billion solution sees bulk purchases we could reduce the cost per internet connection by around 20% on the already minimal costs I used. For this stage we forget things like building and staff etc. Add them onto the billion, lets go for total being larger than the billion

  • LOL I forgot Racks, cabling. Lets allocate a rack per internet connection (about 700-1000 ea)
  • $9300 for computers and network gear per rack
  • Aircon at 5k$ per 10KW - 64 small computers at 20 watts ea is 1.28 KW
    • $640 for installed aircon per rack

So $11,000 per internet connection - one rack - (1280 Nodes)

  • 1.28 KW power (this helps determine building infrastructure costs)

$1,000,000,000 gives around

  • 91,000 racks
  • 116,480,000 Nodes
  • 117MW power required.

From this we get real life costs of $7 to $12 million to build a data center per MW.

  • 117MW is 800 to 1400 million dollars based on power usage only.
  • Actual cost would be much higher due to the low power per rack. 91,000 racks is a lot and would need multiple buildings.
    • 4 billion would be a closer estimate to house 91,000 racks since the MW estimate assumes a much higher power density.
  • with a 20 year life span on 4 billion gives 200 million per year for building costs.

The staffing for this is fulltime job for more than one person and also repair by contractors need to be factored in. Not unreasonable to say staffing is in million dollar range per year

As a rough estimate it would be 1.5 billion for one year of 91,000 racks (116 million nodes)

  • 1 billion Compute/networking
  • 500 million building (per year)
  • staffing (insignificant compared to building)
    And 600 million each year after
  • 500 million building per year
  • 100 million approx repair costs
  • staffing

Now for 100 million node operators running 1 PC

  • zero for equipment, rent, aircon
  • approx $7 per year extra for electricity because they leave computer on for longer to earn
  • 20 nodes per PC for a minimum of 50% of the time.
    • because it is those who leave their computer on for longer periods that will be more attracted to running nodes then it’d be more like 66-75% of the time. But lets run with 50%

So at any time for the 100 million users

  • around 1,000,000,000 active nodes
  • $70 million for extra electricity cost per year

From the above we see that just 100 million people supplying nodes with a PC would swamp the billion dollar project. Now we expect that there will be more than 100 million people with PCs and similar people using their laptops, small computer, SBCs, NASes, and so on proving even more nodes. You could estimate that if only 100 million people with PCs ran nodes then 200-400 people also are running nodes. 2 billion nodes at any one time would not be unreasonable.

Using the calculation for lost chunks and using the current 5 replication node structure and assume the billion dollar project can reach 5% of nodes we get
0.05^5 = 0.0000003125 chance of any one chunk only existing in that 5% of nodes (the billion$ project). This is a file of 3,200,000 chunks that might lose one chunk. or one in 3 million chunks lost.

So yes there would be some loss but not as massive as one would think by spending 5 billion dollars (or 1.5 billion using loans to build buildings)

And China would have a bigger effect.

*There maybe typos above, but this back of envelop estimate would hopefully give you info to base other more serious scenarios upon.


What about a trillion dollar project? Wouldn’t it represent a much bigger threat?

Now you are into large countries trying this. What country has 1 trillion spare to build out the largest internet infrastructure by any one entity. I’m not even sure the NSA’s massive datacentre for spying on their citizen’s packets crossing boarder routers to external destinations cost that amount, is their whole budget that big?

trillion dollar in todays dollar is still a massive exercise. And to have it spare just to try and pull a bait and switch event. Honestly I doubt any country could. Also the building program would chew though so much of the trillion budget since no one will lend money for building project like that. In the above rough analysis the components bought were 1/5 of the total outlay but loans can spread 4/5ths over 20 years. But not for a trillion dollar project. So its 200 times the size in effect and yes the bait and switch would perhaps make Safe unusable, but then again it might end up promoting Safe so MUCH that everybody runs nodes that can, its included in standard OS installations so the 5 billion computers out there end up running 5-50 nodes each and swamping again the huge project.

That is the trap that the billion project can also fall into. It becomes the fuel to propel Safe into the front and centre of storage strategies of OS suppliers and large businesses and a trillion dollar project is almost guaranteed to do that. Thus defeating the very thing it wants to eventually do.

I said bait and switch since they provide so much cheap resources to people for storage since it’ll help keep the price low (the baiting) and then switching to not being there causing the prices to rise for new data.


:exploding_head: :exploding_head: :exploding_head: :exploding_head:

Wait, do you use A0 sized envelopes? :rofl: :rofl: :rofl:

1 Like

Why would you use any other envelope for estimates

1 Like