I would like the SAFE community's ideas about hardware

Hi guys,

After giving the hardware thing some serious thought, I decided that I’d like to commit to one design, crowdfund it, and then get it shipped. We’ll base from pre-existing hardware (unless volume is very high, in which case I suppose there’s the wole “stretch goal” concept). The first thing I need to know, though-- and please-- if you don’t think you’re going to contribute $x when the crowdsale starts, don’t say you want me to build the $1000 version of a safebox, just for the hell of it…

So, here are my queries to the community, with some background information. I’ll answer any question people have.

  1. What is the minimum feature set? I can build a minimal safebox for $20+shipping, but note the phrasing: minimal. At that price, there’s not even any storage in the thing. You’d need to go out and buy an SD card, set the SD card up, and get the board moving. I’m assuming this is not what people want, but I’m throwing it out there in case it is what people want.

  2. Storage out of the box:

  • Lowest End:
    • 8 or 16GB SD Card
  • Better:
  • 128GB SSD
  • Bigger (consumes more power, and can break though):
    1TB HDD - 3TB HDD (Largest for laptop form factor)
  • Seriously expensive:
    N 5-10TB Disks
  1. Your home router doesn’t have to deal with NAT. Should this device double as a router (Maidsafe team, I’d love to get your opinion on this part specifically).

  2. What are your thoughts on the software for this device? I’m thinking a basic, clean linux-based webUI, but I’d need to check with the team to ensure I know the best way of making it so that devices “behind” any hypotehtical router can access the safe network. I don’t know if PCs should always have client software on them, or if it’s fine for them to connect via a “beefy router”. A beefy router can get as powerful as a Xeon Octo-core and 64GB RAM. Yes, I know at that point it’s much more a “computer” than a router, I’m just calling it a router because it’s got many ethernet ports.

  3. Form factor:
    a. Raspberry pi
    b. stick pc
    c. router
    d. you tell me

  4. Interfaces (Thanks @quippu!)
    -Minimum: 4x USB 3.0
    -Preferred: 4x USB 3.0 & 6x SATA3
    -Insane: Full RAID controller w/ 12+ SATA3 (these cost a thousand dollars or so each)

My end goal with all of this remains a modular generic compute system, but that will take some time. (or, it won’t. Sometimes things come out of left field in a good way, too. I do have my eye on the 3rd-gen Rpi modules…)


I think a smaller chromebox/micro style would be great…I think for storage 250GB would be the lower end I personally would like to see…depends on the usage though…farming only vs full fledged usage machine…

1 Like

in any of the cases such a system should have fast interfaces to attach more storage later on.


My views, not saying yet that I would buy even one. Said that so that I am not counted as a sale (yet)

Option 1

Personally I would not mind this, it allows a backup unit if one fails. Or I can shove a SD card I have sitting around collecting dust.

I would hope that there is a bash script or similar to do the work of formatting and setting up the vault. You need this anyhow since people will need a way to “reset” a card after either a failure of the SD card or upgrading to a bigger SD card.

Well I have been playing with microprocessor (& microcontrollers) systems for over 40 years now.

Option 2:

I would expect that you would market it as zero storage and the supporter specifies the storage size by which reward they want or use backetkit after the campaign to add on storage to their reward/pledge.

So then Option 1 & option 2 become the one option, just option 1 is option 2 with zero storage supplied.

Also I would not supply any storage above what can be supplied with a SD card or USB stick. The costs of transportation of disks would be beyond reasonable. Buying the disk locally for the supporter would end up far cheaper

Option 3:

My home is cable, I would have to put the modem into bridged mode to make special use of this. Also ADSL it is typically a modem router and the same would have to happen.

I’d say that marketing it as not having to deal with NAT is a drawback. The SAFE s/w is going to do this anyhow and to remove it would mean more code testing and another piece of code that can break.

Having it act as a router maybe useful to some, but if you have linux then you can make this a option that can be enabled anytime, if you add a switch.

Only if you are running on the edge with performance would it be something to make a purchasing option (to get a larger unit). But we have the Pine64 that is quad core running linux now and has the options of SD and USB storage, costing $15/$25 on a single unit basis. It will do everything once SAFE is compiled for ARM

Option 4:

What OS were you thinking of?

Linux seems to be the best option and can run on very minimal hardware. Once you go Linux then its a matter of what RAM you give it. Then it can run standard systems (routing, SAFE, shell, etc)

Not sure what you mean here. Client as in SAFE client? APPS? Browser? And beefy router? Most routers today should handle the 100-150 connections that SAFE will require to connect to its group and near nodes

Maybe have a 1GB RAM option for a plug in unit that is a vault and simply upload a file with the SAFEcoin address for the vault. Monitor & keyboard can be connected but UI only.
And a 2GB RAM option that allows normal linux operations and UI

Option 5:

Probably depends on the options.
One other possible is a hdmi stick so it can be connected to a monitor or TV using a BT keyboard/mouse (sell as optional extra)

One as a router if demand for it is high enough. Would your router have WiFi?

Option 6:

The minimal system I thought would be a vault only unit. So a SD card slot and ONE USB port is all that would be needed. Plug it into the local router plug in the SD or Disk and power then its running. Cheap n nasty vault that has minimal work to be done, just set the SAFEcoin address.

Maybe have the potential to plug into a daughter board to give access to other USB ports / SATA.

So the minimal comes with two versions


1 x SD card slot (micro)
1 x USB 3.0
1 x gigabit network port (unless you are going WiFi only)


1 x SD card slot (micro)
1 x USB 3.0
HDMI plug - the unit is a HDMI stick to plug into a monitor/TV  (if desired)    
1 x gigabit network port (unless you are going WiFi only)

have a “daughter board” that can be added to provide the other options of multiple USB, SATA, whatever and only plugs into the second board option (2GB model)


As far as storage speed is concerned one has to remember that the greatest delay at this time is your upload link and lag to the node it is talking to. Average this lag will be ~ 1/4 way around the earth.

1 Like

Neo, your feedback was extremely useful, thank you. A few things:

  • Of course I’d make a script :smile:
  • I know CRUST deals with NAT, but… the reason CRUST needs to exist is that nothing-- not even CRUST-- can deal with all NAT all the time. If I’ve got that wrong, please tell me in what way, because then I’ve got a deep misunderstanding of something.

You’re one user-type. I assume there are others who wouldn’t mind spending more for something that’s already configured, but I don’t want to make assumptions (and that’s why I started this thread) so again, I really appreciate your taking the time to give such a detailed, well-thought-out reply.

PS: Not counting sales at this point-- more like judging interest-- anyway, thx again!

1 Like

That"s me! :slightly_smiling:

Yes, very true.

My thought would be that the above hardware options could have a third that does away with the daughter board and is option 2 with the extras.

So each of the h/w options allow you to provide customization, but this customisation is more on the s/w level and what you provide that plugs into the SD/USB/SATA ports.

The s/w options would be available for whichever board can handle it and as such not really needing to be part of specific pledges, but rather the particular h/w board allows particular s/w to run.

By having the 3 hardware options you provide for the different scenarios that are likely to occur.

h/w opt 1 is for the person who wants some vaults and does not want to do anything else with them. Just have them run and provide storage to the network earning coin. These might be their 2nd/3rd vaults.

h/w opt 2 is for the person who may want the unit to perform additional tasks. Whatever linux offers and yourself. Maybe a status screen in the UI so they can see how the vault is performing simply by plugging it into their TV/monitor.

h/w opt 3 (opt 2 + daughter board or its own board) is for the ones who want more and obviously more expensive. I expect the typical situation is that the user would have only one or two of these and the rest opt 1 boards.

Each h/w option can come fully configured only requiring the user to supply their safecoin address.

You can then design s/w and/or UI for the users. Each piece of s/w, UI is suitable for any h/w option that can run it.

So really the question in my mind is what to have for h/w option 3

For the bigger farmers.

12x sata
2x lan

no aes-ni.

1 Like

This is somewhat off-topic apologies. My current connection $30 p/m uploads at around 6meg. For an additional $10 I can upgrade to 10meg up.
Or… For $299 there is 2000meg up option.
I’m guessing that the extra $10 a month would benefit me whereas the $299 would be a waste of money?

1 Like

I’d imagine no one is going to know for sure until farming is tested but I would think that 2000Mb is going to be overkill and probably will waste money. But the 10Mb up seems pretty worth the $10


The Minimal Case

I’m interested in the minimal model, as a proof of kinds for what safenetwork could do for the developing world. If you consider the use case of someone just wanting to have a working machine on which they can do work and communicate freely and safely then that would be satisfied with a Raspberry Pi, if there were a set of client binaries available for 32 bit. I’d like to play with that, but really don’t have the skills to do a build. A concept demonstration for your project, at the entry level end, could be a Raspian dist with the safenetwork client stuff installed and ready to go. If you can be bothered, I’d be very happy to test that, have a Pi ready to go with 8G SD. Happy to get a larger SD Card for the purpose.


Yours is a problem I’m desperate to fix. can you tell me more about your fconnectivity situation?

I’m not sure that I understand your question correctly. I decided to do the upgrade and got triple the download and dbl the upload for the same price I was paying before the upgrade and from the same provider. Not sure how that happened but I offered to take the lady on other side of the phone out for dinner… Typically after those calls I need a stiff drink. :grinning:

Ideally it would be awesome if you did one of two things;

1; offer something basic that can be custom scaled (maybe limited options if its easier for you, no doubt this is a learning curve for all involved)
2; maybe like 2 or 3 options where you have a basic, advanced and a super duper machine.

But none of that is necessary.

For me personally and I know I keep saying it all I want is something that is kind of idiot proof out of the box, I can work the rest out if I have some good support or literature. Basically all I want mine for is farming so yeah a basic clean and most importantly easy to use UI, some specs that meet the demand of the network etc. I would not need anything development wise like the 21 BTC machine.

Cost wise under $300 USD

Now, Idon’t want to be too optimistic, but I’m naturally optomistic, otherwise I wouldn’t be trying.

Four months form now. Kickstarter beginning in 30-60 days.

Something freaking awesome.

Something moderately priced.

Something fast as hell. …and they are all the same something.

All I will say for now, since I like my foot not in my mouth ;).


Will it have cjdns? This way we can start building meshnet ground up, while using safenet over the clearnet. It will be a double kill with single stone.


+1 to this idea, also adding other functionalities to the higher spec models or even lower specs depending on cost prices will help in the future and give it multiple uses .

like expanding wifi access in towns to web or just maidsafe once it is filled up with sites, allowing other companies to buy them and have them set up in store , offering wifi access and earning potential rolled into 1 . (this idea would most likely be custom code rather than added hardware or other goodies, but is just a random example)


The beauty with cjdns is that it works well with safenet. Safenet has “router” which is required to jump into the safenet. So in this sense, you pick your most trusted friend router. In exchange, the friend router will give you access to the safenet, xor address. Bomb proof!



You’ve now made me curious. Do you know how this works out in the real world?:

CJDNS - SAFE - Plain Old Internet

Eg: Will SAFE ndoes on the CJDNS reflect changes in the SAFE content from “Plain Old Internet” provided that there are nodes attached to both networks?

1 Like