Well, I ordered a Pi today - exclusively for testing as a farm/vault. Only got the 2GB RAM version - seems like RAM won’t be critical. Not going to get an SSD for it yet … not sure what size would be optimal. Wondering how many vaults I can run on a headless Pi and what optimal size of vault might be. Time will hopefully reveal the ideal.
I’m not concerned with storage, Once I know how many vaults I can run reasonably without maxing out CPU, RAM, networking AND when I can make a good guess at what the best vault size is – then I can get the right sized SSD as attached storage.
for testing though I’ll just make tiny vaults on microSD card.
That’s my logic anyway, maybe I’ve got it wrong somehow.
Looks to be significantly cheaper than a Pi setup + it has a built-in battery backup … but quite a bit less CPU and only 1GB RAM – I don’t know yet the significance of the lower specs, but will have down-the-track once I get my Pi and can put testnet on it.
Thank you, yourself.
I haven’t tried the Freedombox. Please let me know if you do. Maybe you already did. I haven’t read this thread or the rest of the forum properly in a while.
Depends on how much resource is being used by each vault versus how much available resource exists where the vault is located. So CPU, RAM, Drive space, and Network throughput.
As a farmer you seek to maximize the utilization of all the resources for which you are paying to get the most bang for the buck.
My early speculation is that I will have excess of most things, so multiple vaults is better than one vault. Maybe I’m wrong. Will be testing to find out.
Probably another factor is that safe should become larger than your living room can handle… So with a couple of small heaters in your living room you provide resources for safe (+ don’t need to start worrying about space, cooling and backup power/network supply) but you shouldn’t be able to destabilise the network as a whole
+because the replication set stuff and raid like behaviour is all part of the network you can just use regular consumer grade hardware which is way cheaper than professional server hardware…
… So less expensive hardware combined with excess resources (cooling, unused network capability, space) while still providing reliable data storage should lower costs for society as a whole imho
A couple of issues.
In the same IP address, could you have several computers doing agriculture ???
Can you get the same rewards, with a desktop computer, a laptop or even a mobile phone ???
Yes this should be possible (using multiple ports on the same Internet connection)
The rewards depend on your nodes performance - your phone will be offline again and again and therefore probably perform badly… The details of reward will probably be tuned with the coming test networks… So I guess too early to compare laptop with desktop pc
Uptime is essential or you may get booted and have to que to rejoin.
Everything else you mentioned will no doubt affect earning rate, it will be a race to see who can serve the requested data chunks, you want to win that race to get rewards.
If by NAS you are referring to its RAIDing disk for better security of data. If so then this is like RAIDing a set of RAID storage units.
Safe is a massive RAID already and the data security is built into the network. Also Drives today are getting more and more reliable, so I doubt having RAID for your node will be an advantage in itself. Better to put each disk onto its own SBC and run multiple nodes. Any one disk failure only means you lose one node and have to rebuild the age of it. If your NAS unit fails (thus all storage) then you have no nodes earning.
But if you are using a NAS to give you more space then fair enough. Best to run multiple nodes off it anyhow.
Nobody knows yet. We have some expectations, but we have to wait how will the testnet perform on real Internet. read/write speed on hd - maybe not that important, hdd seek time
counts in the overal latency, but probably only as a small part latency - you want overal latency to be low (receiving request, getting data from hdd, sending data) network speed - saturated connection means latency goes up, but it is imposible to predict what will be the network usage, it depends on number of Safenetwork users and what they will do uptime - availability, not necessary uptime, short drop out (reboot, network reconnect) should not lead to big loss
It would be interesting to see differncies between ISPs, how well are they connected to other networks and not only the big players like Youtube, Facebook,…
@peca Unless your hardware is so 2010 then most of this will be overshadowed by the lag time across the wire to elders and nodes across the world.
As you mention a “saturated” link to your ISP will cause significant lag time that could be greater than any over the wire lag. But I’d expect that should not be occurring for suitable connections. Some previous testnets required 6Mbits/sec up and from those testnets that was more than sufficient. Unless you run too many nodes and saturate that way
I upgraded my old banana pi m1 (the one with the sata connector) to an odroid hc4 – “toaster” a few weeks ago. the hc4 seems to have a slightly slower processor/soc than the rpi4 but offers two very fast sata connectors (connected to the soc via a direct pcie link). the hc4 can power all sorts of hdds, even the bigger 3.5". I had a DIY power “solution” built for my old banana pi by soldering a sata power cable to an usb cable, but that only works for the smaller 2.5" hdds, these take mostly 5v (usb is 5v). Bigger hdds need 12v as well.
I had to remove the builtin bootloader – “petitboot”, to make it work with the latest armbian. petitboot is able to boot from a hdd directly, but it doesn’t work with armbian. With petitboot gone i had to use a micro sd card to boot the hc4, which then forwards the boot to the hdd (nand-sata-install). (sd cards are very bad for doing anything if compared to an ssd)
Another thing to watch out for is the processor frequency. armbian comes with a 2.1GHz preconfig, which is an overclock. I had run a stress test (stress -c 4) for hours without it crashing, but then it crashed the other day on light load. Changed the max freq to 2.0GHz (through armbian-config), it is running without a problem since then.
I wasn’t aware that the standard Rasberry Pi OS was 32bit. There is a 64bit but still in beta. It is considerably faster though, so hoping I’ll be able to use it with Safe.