The meetme rooms are expensive, best to find a small colo, rent a 1/4 cage w/ power and get a leaf drop 10GE access to the Internet, you have to run something else I agree, like your biz website on the same system to make it work, also even sublet part of the 1/4 cage to others you know. a co-op of Autonomi Node Operators, sharing the monthly cost of the 1/4 cage power and internet access would be a good thing.
1 Like
I do understand where your coming from.
But isnt that just putting lots of peoples eggs in one basket.
Great for the network while its working, but if it fails its a compounded problem for the network.
2 Likes
Itās a reliability thing, it makes sense to co-op host in groups of 4 or 8 operators, each running independently and in parallel on different IP addresses at the same time sharing cost to get the higher reliability and higher speed at a lower monthly price.
This will happen because operators will find a way to save money and improve returns.
Doing the above provides Autonomi Network DAO a much higher degree of stability in a distributed fashion than otherwise possible from home setups.
Also there are relay node and private DNS node functions coming which could benefit the entire community of nodes with such reliability.
Pragmatically speaking, itās going to happen in order to properly underpin the viability of the whole Autonomi Network community @ scale to make it attractive to the masses, and we are talking about multi-millions of nodes, if not over a billion nodes being the medium term goal.
Such numbers of node operators will want speed and reliability for their uploads and downloads and their peer to peer communications, so having āthickerā multiple parallel node relay paths and super responsive DNS services coursing through the Autonomi Network DAO is a good thingā¦ a really good thing for everyone.
The other thing to think about is that next gen software tools like SpinKube deploying WASM versions of the safenode code in WASI shim enabled containerd enabled containers is both super fast and highly secure.
We will see imo, a lot of this type of node infrastructure deployment get automated by SpinKube is my prediction, to run 5000 instances on a Kubernetes cluster with super fast binary execution speed at really low cost , in a much smaller hardware footprint than the current safenode executable written in rust.
Rust compiles to WASM, the safenode is rust open source, so its gonna happen sooner rather than laterā¦
No one in the Autonomi Network DAO Community should be worried about the above outcomes. 
1 Like
case in point, my current 8GB RAM intel 10th gen processor with 4cores and 8 threads with Windows 11 can at best run dedicated 50 nodes because the safenode needs 150MBs. Itās likely compiled as WASM you are looking at 5MB to 10MB, which is 30X best case, 15X worst case more nodes on the same system. Thatās best case dedicated on this old PC 1500 Nodes in theory, with the same performance I am getting today. This means the home network better have a 30X faster up link. Right now I run 80MBsec download on Cogeco uplink and maybe 20MBsec uplink on a good dayā¦, its cableā¦ , on average I am consuming 1-2% peak of the network bandwidth with 5 nodes. WASM compilation could 30X that, to I am running 150 nodes, but I would need to upgrade my cable service, which costs me 85.00 CDN/month to CDN $125.00 a month ., doable.
That said I could partner with three other people and rent a colo 1/4 cage with 10GE access , for CDN $300.00/month, power included which feeds up to 4kiloWatt peak to cover my power needs , so 1 kilowatt per renter, more than enough to run a serious hyperconverged DAS system with in a 1U with 8 NVME X 2TB drives in RAID, and 4 ports of 10E each running at a leisurely 1Gbit/sec on different IP addressesā¦, easily scaling a single 64GB RAM box to 8X150 safenodes running as WASM compiled safenodes, contianer deployed to create 900 safenodes using 4.5TB of RAID protected NVME flash and still have about 3.5TB left over to keep the FLASH Fast , all of it operating in a 1U 450Watt Gen5 package with the lowest cost , lowest power gen5 processor from Intel.
1 Like
Havent you just compared an old pc at home to a better pc in a dc.
Of course youll have different results lol.
My point is still valid.
Currently my nodes are using ~70MB RAM each. What is your experience on that figure. Is it quite constant or it gets bigger with time?
It fluctuates but should not increase indefinitely.
Mine started off at around the 50-70MB each and grew to 150 approx yesterday and now 24 hours later are still around that amount. I feel it has to do with the increase in connected PEERs which also increased
After running for 10 days my nodes are between 70-230 MB, most of them around 130 MB.
1 Like
yup, thatās the point, why would you run a home PC that in reality costs 5X to 10X more per day to operate vs co-op share of faster machine with more earning capability?
Itās a different story for me. The more nodes I run on 1 computer, the more RAM they start to eat, until it becomes unstable and the system crashes. 300 nodes are currently occupied 73,6 GB RAM with the system.
Privacy. Security. Freedom
4 Likes
Either you are hitting a bug or you have bottleneck somewhere. It may be that nodes cannot do all their work and it slowly piles up until it crashes.
120/112 Mbit seems suspiciously low traffic for 300 nodes, my average for same count is around 170/170 Mbit.
3 Likes
This really doesnāt make sense since you are comparing apples to oranges. Yes both are fruit (computer processing & talking on internet) but they are not the same. One is for home use and used by the people to do what they want, the other is purposed for its task in a datacentre.
The home computer is on anyhow, it is consuming power anyhow with or without nodes. The tiny amount of extra energy in CPU and drive usages is a fraction of energy used. If it is anywhere near 1% then Iād be surprised.
The metric you used to say 5x or 10x is not there. You are taking two different metrics and equating them. The metric of cost to node earning power does not have the same basis for the home user to the data centre computer. The data centre computer is there to earn through use. The home computer earns as a tack on.
For the couple of extra watts a home PC uses to earn from 5 or 10 nodes is nothing compared to the hundreds of watts needed for 500 nodes, + costs of computer over time, + overhead costs the co-op incurs on behalf of that machine.
EDIT: This isnāt to say that the data centre idea cannot be profitable, just the costs structures and relative ROI will be less in general for the home user who just tacks on some nodes to their computer they already use.
5 Likes
An old laptop with 30 nodes burns 1.5 euros of electricity per month. The return is infinitely greater than anything you can put into a data center at this level, which will be the mass case. But if you want more nodes from home, things change and there is a point where it is more profitable for you to put a server in a data center.
Privacy. Security. Freedom
5 Likes
And that is actually apples to apples with the co-loc example since your old laptop us only earning and not being used as a home PC that was being used.
Not much electricity really
My Odysseys are about 15 watts and can also run 30 nodes. 15 watts is 10KWHr/month and we pay like 0.20 Euros (converted from AUD) per KWHr so it is 2 Euros at our rates of electricity
Yes and that point is decided by the router/bandwidth
But if the nodes are being run on your spare resources that you use anyhow then it cannot compete. 30 nodes on my PC that I use will be less than a watt or 2 extra if its even one watt. Something like 0.1KWHr to 1KWHr per month for the 30 nodes. IE 0.02 to 0.20 Euros and that is the total costs for running those 30 nodes
The data centre could not even pay the monthly location (the computer share of) costs for it running 600 nodes. 1/20 of the location costs would still be higher than home using spare resources. Then there is electricity, the cost of the computer split over 36 months, the cost of internet. There is no way a data centre can complete node for node against home computer using spare resource.
Where the data centre helps is if a person wants to run a lot more than they already use. Like 600 nodes etc. Still not as profitable node for node but 20 times lower profit maybe OK
4 Likes
of course its an apples and oranges comparison. The co-op sharing orange model lowers the cost to run a safe node substantially while improving the overall performance of the network. Itās not an argument for one or the other, its an argument for both co-existing, market forces will decide what the proportion looks like over time.
1 Like
Because safenodes arenāt registered services with MS , MS Defender gets in the mix checking everything treating safenodes be default as suspicious, so that eats up a lot of clock per core, Just watch the Anti Malware service rise to the top of the Task Manager and see how much clock (CPU %) it easts up when the safenodes get busyā¦
1 Like
I didnāt have ārouter hardwareā on my launch bingo card. Need to look in to it.
2 Likes
I have for 110 nodes about 35-40Mbps up and down. For 300 could be 100-110Mbps.
1 Like
what OS have you got that Pi4 running and how are you set up?
i would think āBest Node Hardwareā would definitely invite/require setup and usage info to be relevant and useful.
i get fairly regular twitch and regurgitate running more that 12 nodes on PI4 8G with a fresh bookworm install booting/running from 2T 3.5" HDD plugged into USB3.
i do NOTHING with this machine except run nodes and what i need to send screenshots over to THIS machine which is my dailydriver: Pi4 8G fresh bookworm on 160G HDD.
This guy, while kinda slow, usually has a dozen nodes that mostly stay running and it has yet to overwhelm and reboot itself.
likewise, iām on limited ISP bandwidth, 20/20 on a good day.
so, given the premise of Autonomi running nodes on excess user space on our daily drivers, and given that some of our daily drivers are tricked out beasts and some are humble Piās, i think a more useful discussion would be HOW you use WHAT YOUāVE GOT.
So please, oh master of Pi, do tell how you get 40 nodes āno problemā on that system.