18:45 emission gone?

Sometimes it just said running, but not a real measurement of if everything works fine. Especially on UDP, hard to know if it works fine. Since I did heavy tuning to run 800 nodes on SBC @ 3W, 4TB nvme. I am planning to scale up the machine. So the emission is a good indicator to show if it working fine.

I am still wondering the network latency that kills the rewards as well. Like pi-hole, already add few 10ms unlike OEM router. Especially reading some guys running fewer nodes with more rewards than me, I am not blaming unfair or what, I am looking for improvement. Maybe the geolocation is an issue as well🤦🏻‍♂️

1 Like

Where are you geolocated?

Can you share your tweaks? Please :slight_smile:
I am about to butcher an old PC PSU to run several SBCs.

At northeast England, maybe that’s why :joy:

For the SBC, I choose ARM rather than x86, since ARM much more energy efficient. I tuned the Memories management, which allowed me to handle even 100% more nodes without overloaded the cpu. For the cpu, I actually reduced its scaling frequency down to 1.8, quite energy efficient for this. Never over 45°C, no fan, all passive cooling.

However, the rewards still not showing that what I expected. I am quite sure I need more improvement on the networking.

3 Likes

Ah the Geordie Jinx…
What model of SBC are you using?

I bought a couple of Banana Pis a couple of years back, just cos they had SATA connectors.
I need to power them up again and see how many nodes they can handle these days.

Banana Pi too expensive for this application, maybe 150 nodes maximum with around 3-4GB ram. And the performance not quite nice🤦🏻‍♂️SATA need quite a lot of energy to work with although quite cheap. I saw 6 sata port version of Banana Pi, but not working well with using all those ports.(maybe they reduced the bandwidth of the data transaction, like the whole system freezing and needed a hard reset).
I choose 4 core A55 arm cpu sbc, I have tried 8 core A55, but failed. The cpu load balance failed on 8 core version, maybe the sbc not in well design. I have tried on 4 core A53, still good, under 500 nodes fine@2.5W with heavy tuning.

I sure try pi 5 to see how much nodes it can handle. But too expensive if it just handle under 2k nodes😂

2 Likes

A fine tuned memory management can reduce cpu load quite a LOT! I am not sure if it still the same thing on x86 or AMD, need some testing. Another limiter should be the physical storage drive, the limited bandwidth also freeze the system if too much read and write at the same time. So I don’t think over 10k node on single machine would work well or need more latest generation of storage🤔
Turn off the journal maybe…:thinking:

1 Like

Here is the network for around 3k nodes, 105Mbps down 80Mbps up. Doing the nodes while watching 4K and torrenting still fine. So plenty of room for scaling up those nodes.

1 Like

with 4TB nvme, the limit will be your network bandwidth before the sdd (unless you have a very cheap ssd or your nvme port is quite limited).
800 nodes @ 3W is excellent, is it with 0.3.9 or 0.3.8?
With 0.3.9, I can run 1500 @ 45W on old skylake mobile 4 cores with 64GB ram…
What memory settings did you tune exactly and how many nodes were you able to run before tuning? (and are you sure your node are active? at some point, I was able to start way more nodes, but it turns out I made a mistake in my port forward conf and they were not earning).

0.3.9, I don’t think network bandwidth is a issue while Mpbs vs MBps already having 8 times different. And I still have plenty of network bandwidth. Nvme bandwidth much much larger than giga Ethernet.

Before tuning, 300nodes ~ 400nodes. All nodes are working right now, although the rewards lower than expected, that’s why I care about the emission to check if it working fine.

What I have found was the memory speed doesn’t matter much on handling nodes, so reduce cpu load as possible

Anyone else’s emissions still way down on 0.3.8? I’m getting about 10-20% of before and running about 50% of nodes (due to increaed load).

  1. Emission cut half
  2. I don’t think it is now @ 4M nodes
  3. Not really cover the cost if you want to earning from it (mine need at least 4-6months to cover the cost)

Similar rewards rate to me, just around 20% of my expected from the emission

I’m getting about 2 ANT per 24h for about 1600 ant nodes. I was getting about 18 for 3200 with last version. :thinking:

I have one sbc with 0.3.8 and it earns same as the other 2 on 0.3.9

That is around the rate I am getting per node

1 Like

Routing tables are updating now - no more super earners

Edit: The only funny thing now is that with 2 ant per 1.6k nodes and 16k ant distributed daily this would hint to 8k*1.6k=12.8 million nodes.
So if 5 million is correct there’s quite some super earners left (who didn’t update yet) until the next release reduces their earnings

5 Likes

I’m guessing / hoping it means the rewards are being spread around more effectively vs the uneven distribution towards a few favoired old nodes previously.

So, more nodes are earning due to improved connectivity / network visibility, hence rewards are lower.

1 Like

Currently I am seeing the rate of around 1 in 40 nodes getting a emission unit per day

with 288000 units that gives around 11.5 million nodes on version 0.3.8 or 0.3.9

My nodes are estimating around 10 million nodes

1 Like

So, it sounds like my nodes were over achievers before then? :sweat_smile:

2 ANT per day is about 15p* per day or about £55 per year. At that price, it would take a used mini PC with a 2GB disk over 10 years to repay.

Obviously, machines which can run nodes in the background will be fine, but I suspect that would even make it tough to pay for the electricity in a data centre.

It will be interesting to see where all this balances out when most folks have upgraded.

EDIT: * That is less that the cost of 1 unit (KWh) even in off-peak for me. Maybe some suppliers can drive that lower, but at 80W, without solar giving me free power, it would actually be costing me to run the mini PCs now.

1 Like

One would expect that once data is being stored and the network is becoming useful that the price will increase. I would expect your 730 ANT per year to return more than 55 pounds.

For my SBCs its a 5 year investment at these low rates. But if ANT goes to 0,50 USD then its one year.

1 Like

True, as ANT increases in value, the economics change substantially.

Indeed, as emissions dwarf data fees/rewards, really it is only the token price that dictates the value of the rewards.

This will surely put upward pressure on the token too, as we’re pretty much at the cost of power or less for many. I can’t imagine those 5m nodes are on all running on spare laptop/desktop capacity either.

It will be fun to see how this unfolds! :sweat_smile:

2 Likes

That is why I said the 4M nodes in total is not trustworthy. Or my nodes underpaid😂
So power efficient much needed to keep everything low cost, sata hdd definitely cost more to run. At least ssd, powerstate 2 to keep low power consumption. Else, running cost > rewards.
Network basically is an already expense for home user, so only SBCs in low power can earning something right now. But long time to cover all those cost tbh
In order to keep it in higher price, mining all those emission and reduce the market supply, or buy all available token on the market and hodl would help😂

1 Like