Measuring the cost of running a full node.

[teukon]

We might approximate the total cost of all mining activity by summing all block subsidies and transaction fees.  If we'd like the entry barrier to mining to be low, we should surely like the cost of running a full node to be a tiny fraction of this total.  Can we automatically adjust block size limit to target a particular fraction?

Is it possible to measure the cost of running a full node?

One thought: Perhaps we could have a parallel, proof-of-work blockchain with no transactions.  A fraction of Bitcoin's mining reward could be channelled to this supporting chain.  The difficulties of the two chains could be used to calculate an approximation of the total "full-node costs" of all miners.

Does this work?  Can it be abused?  Is there an easier way to trustlessly measure the costs of running a full node?

Note: this type of block size limit targeting would be supply/decentralisation focussed (like BIP100) and so would be compatible with demand/fee-market focussed limits (like BIP100), acting simply as a ceiling.

[1714934188]

1714934188

[1714934188]

1714934188

[1714934188]

1714934188

[1714934188]

1714934188

[Buffer Overflow]

Is it possible to measure the cost of running a full node?

You could put an electric meter on the plug of your node and measure it's consumption I guess.

[teukon]

Is it possible to measure the cost of running a full node?

You could put an electric meter on the plug of your node and measure it's consumption I guess.

I get that.  I meant is it possible to create a trustless measure, something that Bitcoin itself could refer to for example in calculating an appropriate block size limit.

I vaguely recall Gavin was toying with the idea of trying to measure technological development trustlessly by, for example, having full nodes report how long it takes them to verify a block.  Unfortunately, this system was open to abuse.

I didn't keep up with this line of inquiry and learned that eventually Gavin turned to simply estimating future technology trends (e.g. BIP 101).  I can't prove to myself that a dynamic approach is impossible so I decided to think out loud here.  I guess what I'm looking for is for someone to show me how to abuse the idea above; this would definitely help me hone in on the problem.  If no-one breaks it then I'll expand the idea and flesh out a draft proposal to catch more eyeballs.

[tl121]

Is it possible to measure the cost of running a full node?

You could put an electric meter on the plug of your node and measure it's consumption I guess.

I have a small fanless machine that runs a full node.  During the August "load test" it was running at around 5% CPU load, and so would be able to keep up with 8 MB blocks without difficulty.  The machine with 8 GB of RAM and an SSD cost about $400 and uses less than 10 watts of power.  It runs off of a home office DSL connection. It has been running stably since November with an open port 8333 and a connection limit of 30.