P2Pool and low power miners

Recently, we got a new insight into 21 Inc's plans for its mining computer / chips - allowing the device to connect and mine on any pool, and ultimately - mining on a P2Pool-like network to further "redecentralize Bitcoin". This got me thinking about whether P2Pool would actually be compatible with a potential large swarms of low-power devices mining together. Lets see how it might work.

What is P2Pool?

P2Pool is an interesting idea that came about around 2011 to address the growing centralization of Bitcoin mining in mining pools. Instead of connecting to a centralized pool, a miner would instead join the P2Pool decentralized network and start mining there. The block reward would be split between peers based on how many "shares" they contributed to the decentralized network - essentially creating a decentralized "Pay Per Last N Shares" mining pool.

What is very interesting about P2Pool is that it allows for the decentralization of mining - anyone can join the network and contribute, you are free to mine for any valid block as long as you respect the mining reward distribution and all in all it once again allowed smaller miners to mine for Bitcoin blocks without relying (or giving power to) any centralized third party.

However, P2Pool is not without its disadvantages.

It is reportedly underperforming / being "unlucky", indicating that it might be experiencing a higher rate of orphaned blocks. This could be due to traditional mining pools optimizing their new block discovery time (I heard someone mentioning a dedicated communication network for the mining pools, but I can't find a source for that claim currently), while P2Pool might be reliant on the Bitcoin network itself, which can take a few extra seconds to populate.

P2Pool coinbase transactions are pretty big in comparison to the traditional mining pools' transactions. This means the blocks themselves can process a few fewer transactions, and there is a practical limit to how many outputs one can reasonably fit into a transaction to pay for the last N mining shares.

Mining at a traditional pool uses about 20MB per day, or 600MB per month. In comparison, P2Pool puts a much higher burden on the data transfers at about 38GB per month before we start taking the resources used up by BitcoinQT which you also have to run.

All in all, if you are mining on a computer with a good internet connection, a reasonably powerful set of miners attached and you don't mind earning a few percent less than you otherwise could, then P2Pool is not a bad choice.

However, what if you are dealing with mobile devices equipped with low-power mining chips?

21 Bitcoin Computer with P2Pool

Looking at the limitations of P2Pool and what 21 Inc is aiming to do with their mining computer, there are a few problems that stand out.

Currently, P2Pool pays its miners directly in the block coinbase, while 21 Inc prefers to buffer the balances at its shared wallet before letting you withdraw the mined dust to a wallet. If the 21 Bitcoin Computer was instead to be paid directly with the coinbase, you might quickly run out of block space. Looking at some sample P2Pool coinbases (1, 2), we can see about 200-250 outputs being included on average, taking up about 8kB of space. This roughly puts an upper cap of 32'000 outputs on a transaction before a whole block is filled with only the coinbase. Equally divided, every output would receive about 78125 satoshis, worth about 27 cents (at  current 356 USD/BTC exchange rate). This would represent about two days of mining for one of the 21 Bitcoin Computers.

32k computers mining a block every two days is fairly incompatible with 21 Inc's vision of "buffered pool mining" (quick way of mining coins to use for transaction) and putting a mining chip into every gadget.

Based on the amount of unique entities you want on the Bitcoin network, we can start extrapolating how often they could get paid on average. Sticking with the 32k outputs per block, we would have 4'608'000 daily outputs. If we looked at the sales of only iPhones in Q4 2015 (48.05M), we would require over 10 days worth of blocks to credit each of those devices individually. This is all before those transactions are again spent, before taking into account all the other smartphones, quarters and every other potential device one could think of in the Internet of Things world. All in all, Bitcoin couldn't handle this level of spam even if the blocks were increased.

Looking at the mobile data plans of a company like AT&T, 40GB/month would cost one about $300, or about $10 per day. The data price for P2Pool alone is 37 times more than the 21 Bitcoin Computer would earn. Mining at a pool would cost somewhere between $20-$30 worth of a data plan, making the data only twice more expensive than the bits one would earn before taking electricity costs into consideration.

Optimizing for your needs

All in all, it would appear that with mining, like with project management, you have three variables:

• Centralization vs decentralization
• Low vs high variance
• Whether small devices can efficiently mine or not

But we can only pick two of them. Decentralized low variance mining but not good for small miners? That's P2Pool. Centralized low variance mining for any device? Centralized pools. Decentralized mining for any device but with high variance? Solo mining.

Possible solutions?

While with the current technology it might be rather impossible to achieve what 21 Inc is aiming to achieve in full, there are some ways one could compromise while still achieving some of the desired outcomes.

First of all, one could try creating an intermediate solution between a fully decentralized P2Pool and a completely centralized mining pool. Perhaps we could see a lot of new, smaller mining pools popping up based on carrier, manufacturer, geography, etc. that the devices could connect to and contribute the mining power to instead. This would allow the balances to be stored on shared wallets and used accordingly, perhaps aggregated into bigger payments or some off-chain settlement between those nodes (and oh god, we're coming back to the tired 2013-era block size debate and ways of settling without bloating the blockchain...).

We could focus on creating bigger mining devices that would power our mobile wallets. This device could stay at home and mine coins using the standard P2Pool protocol, rather than having a miner in every device. This would probably just get us back to the buy vs mine debate once more though. We could even do without all the physical mining and purchase some virtual mining contracts instead... Moreover, the situation is no different than what one can currently do with the existing mining hardware and since we don't seem to be doing that en masse suggest we wouldn't do it in the future either.

Lastly, we could just ignore the variance and mining profitability altogether and just starting to waste money for the benefit of the Bitcoin network. While this might sound crazy, it might not be that far-fetched of a plan. Chip manufacturers would probably make more money than the chips could ever mine, so they could just fork over some money to pay the mining rewards in a Pay-Per-Share scheme. Mining would still go to secure the network, perhaps in an inefficient way, and we might just end up with the entire Bitcoin mining ecosystem being generally unprofitable to mine in. Since the cost of mining would be distributed between potentially many millions of people, the individual burden might be small in comparison. However, this entire idea would best be suited to an entirely separate post I might do at some other time.

Conclusions

All in all, P2Pool currently is rather incompatible with low-power miners, especially if data bandwidth and profitability is an issue. If there is some way to solve the the underlying problems with how P2Pool operates to address those issues, I would love to hear more, but I doubt we'll see any concrete informations on the subject any time soon.