[1500 TH] p2pool: Decentralized, DoS-resistant, Hop-Proof pool

[Krak]

ASICMiner Blades and Cubes were the ones I was thinking of specifically.  Because they didn't work for shit last time I tried, even through a proxy.

I was not able to get the blades to work very well. I tested it against the latest P2Pool and my blades locally just the other day. The lowest DOA I saw was 25% but it was continuing to increase so I sent them back to a normal pool. If anyone has a solution for this I'd be glad to test it out.

Thanks.

Edit: I should say I have successfully run BFL, USB BLock Erupters, and a Chili on my own P2Pool node.

Did you force difficulty 1 shares on it with +1 after the username?

[1714224744]

1714224744

[1714224744]

1714224744

[TechByPC]

Did you force difficulty 1 shares on it with +1 after the username?

I had forced difficulty, but not at 1. Just tried setting 1 now moving five blades over to it. It seems to be artificially inflating my successful share rate with a bunch of worthless low difficulty shares. My DOA is 25% after running it a bit. It is better in that my observed hash rate is better, so if I were to get a share I suppose it would be better, but the DOA rate is still crazy high compared to my other equipment.

Edit: My DOA is now steady over 32%

[TierNolan]

What does DOA mean exactly?  Does that mean a share was found on the sharechain since you started mining?

[HellDiverUK]

Did you force difficulty 1 shares on it with +1 after the username?

Doesn't make any difference, other than put additional load on the miners and the p2pool node.



So, basically it seems nothing has been done.  Oh well. 

[Carlton Banks]

So, basically it seems nothing has been done.  Oh well.  

No, it means that everything has been done for every other device that has native stratum. Those devices now all work. AM blades are the only available example of ASIC hardware like this.

If it's any consolation, the p2pool hashrate has gone up so much that you'd be fighting to get 1 share in the chain per block found. Because of all the ASICs on the pool.

[int03h]

WRT sizing a p2pool node .. has anyone got any decent guidance on what the  requirements for running the EXE or on linux?

I have tried various scenarios and it seems dead rate is very dependant on available CPU and hash load.

Watching taskmanager it seems to hit the CPU hard every few seconds as it gets new work. If the machine is too small e.g. Amazon EC2 micro it flat lines at 100% for a few seconds and generates orphans and deads like crazy. The more miners connected the more spikes. The bigger the machine the better, it seems, but that is a VERY WIDE definition.

Anyone care to explain how to avoid losing shares - the more technical the better ?

EDIT: My DOA rate hovers well under 10%. Not sure if I am trying to fix something that ain't broke ?

[spooderman]

I would like to start mining here. Honestly though, all I have is 6.5GH to offer. Is there any point (as opposed to continuing on eligius)? And, if someone would be so kind, could you please tell me how the payout works? It's obviously not PPS, and I guess the CPPSRBP that I've grown to love with eligius is also not offered (is this not a possibility here, given that this method prevents the pool from ever paying out more than it can? Or would this jeopardize the pools decentralization?)

Thknas, spodermen

[mitty]

I would like to start mining here. Honestly though, all I have is 6.5GH to offer. Is there any point (as opposed to continuing on eligius)? And, if someone would be so kind, could you please tell me how the payout works? It's obviously not PPS, and I guess the CPPSRBP that I've grown to love with eligius is also not offered (is this not a possibility here, given that this method prevents the pool from ever paying out more than it can? Or would this jeopardize the pools decentralization?)

Thknas, spodermen

It's been a while since I've used p2pool but I believe the payout system is PPLNS (Pay per last N shares).  I don't think decentralized CPPSRBP is currently possible because there'd have to be some way to manage a pool wallet for when there's a surplus from a block reward/pay from it when needed and doing this in a decentralized manner would be very difficult.

The advantage of p2pool is decentralization/not trusting a centralized pool operator and helping the Bitcoin network by running a 24/7 full node.  I'd recommend not using p2pool with a small hashrate because your payouts will be tiny and cost a lot to spend in terms of transaction fees.

[smooth]

I don't think decentralized CPPSRBP is currently possible because there'd have to be some way to manage a pool wallet for when there's a surplus from a block reward/pay from it when needed and doing this in a decentralized manner would be very difficult.

There is no wallet or surplus with CPPSRBP. The Eligius coinbases pay out to miners, not a pool wallet.  There is a share log and every block that gets paid out to the most recent shares adding up to the value of the block. If the round was unlucky then not all shares from that round will be paid. If the round was lucky then some old shares will be paid.  I don't know enough about the internals of p2pool to know whether this is possible.

EDIT: This isn't quite right since Eligius doesn't pay out until your balance reaches the minimum payout. Still the blocks rewards are paid directly to miners (who have accumulated a large enough balance to get paid out), not to a pool wallet.

[spooderman]

I would like to start mining here. Honestly though, all I have is 6.5GH to offer. Is there any point (as opposed to continuing on eligius)? And, if someone would be so kind, could you please tell me how the payout works? It's obviously not PPS, and I guess the CPPSRBP that I've grown to love with eligius is also not offered (is this not a possibility here, given that this method prevents the pool from ever paying out more than it can? Or would this jeopardize the pools decentralization?)

Thknas, spodermen

It's been a while since I've used p2pool but I believe the payout system is PPLNS (Pay per last N shares).  I don't think decentralized CPPSRBP is currently possible because there'd have to be some way to manage a pool wallet for when there's a surplus from a block reward/pay from it when needed and doing this in a decentralized manner would be very difficult.

The advantage of p2pool is decentralization/not trusting a centralized pool operator and helping the Bitcoin network by running a 24/7 full node.  I'd recommend not using p2pool with a small hashrate because your payouts will be tiny and cost a lot to spend in terms of transaction fees.

do p2pool payouts unavoidably incur transaction fees? I don't understand enough about it to know for myself. If I am directly getting the mined coins that shouldn't happen right? A fee to me implies that someone is paying me a tiny amount. Thanks for your response btw. I'll probably stay on eligius, even though I do greatly want to help the network health. Ghash and guild should have 5% between them. Not over half.

[kjj]

It's been a while since I've used p2pool but I believe the payout system is PPLNS (Pay per last N shares).  I don't think decentralized CPPSRBP is currently possible because there'd have to be some way to manage a pool wallet for when there's a surplus from a block reward/pay from it when needed and doing this in a decentralized manner would be very difficult.

The advantage of p2pool is decentralization/not trusting a centralized pool operator and helping the Bitcoin network by running a 24/7 full node.  I'd recommend not using p2pool with a small hashrate because your payouts will be tiny and cost a lot to spend in terms of transaction fees.

do p2pool payouts unavoidably incur transaction fees? I don't understand enough about it to know for myself. If I am directly getting the mined coins that shouldn't happen right? A fee to me implies that someone is paying me a tiny amount. Thanks for your response btw. I'll probably stay on eligius, even though I do greatly want to help the network health. Ghash and guild should have 5% between them. Not over half.

He means when you spend them later.  The payments to you are indeed generation transactions and have no fees.  Current payout for a single active share is something like 0.0045 BTC.  A spend of 1 BTC using them will be like 50kb or something, and will need some fees.

[smooth]

He means when you spend them later.  The payments to you are indeed generation transactions and have no fees.  Current payout for a single active share is something like 0.0045 BTC.  A spend of 1 BTC using them will be like 50kb or something, and will need some fees.

You can avoid these under the current rules if you a really really careful about how you construct the transactions, or at least minimize them given slightly less care. The bitcoin wiki has some examples. The standard client probably won't construct the transactions the right way though.

EDIT: I thought it had examples, but now I can't find the page. The gist of it is that you need to keep the transactions under 10kb and make sure that all outputs are >0.01. In practice this means not having a change output and it might mean having one large input along with the small ones you are aggregating. It might take multiple transactions but as long as each is under 10kb, you're fine. You might then need a fee to spend the entire amount because the outputs of the initial transactions aren't "mature" but the fee will be a reasonable percentage.

[lenny_]

Rules has been changed in last Bitcoin version 0.8.6. Maximum free transaction size is now 3 KB not 10 KB.
Also, all outputs of 0.01 BTC rule also have been removed.

Please read 0.8.6 changelog: http://sourceforge.net/projects/bitcoin/files/Bitcoin/bitcoin-0.8.6/README.txt/download

Code:

- Remove the all-outputs-must-be-greater-than-CENT-to-qualify-as-free rule for relaying
  (see https://gist.github.com/gavinandresen/7670433#086-relaying)

- Lower maximum size for free transaction creation
  (see https://gist.github.com/gavinandresen/7670433#086-wallet)

[smooth]

Rules has been changed in last Bitcoin version 0.8.6. Maximum free transaction size is now 3 KB not 10 KB.
Also, all outputs of 0.01 BTC rule also have been removed.

That being the case all you need to do is collect up as many outputs as you can fit into each 3K transaction (roughly 10 I think) and send them to a new output. Of course this assumes enough of the network has upgraded to 0.8.6, which may or may not be true.

[mitty]

I don't think decentralized CPPSRBP is currently possible because there'd have to be some way to manage a pool wallet for when there's a surplus from a block reward/pay from it when needed and doing this in a decentralized manner would be very difficult.

There is no wallet or surplus with CPPSRBP. The Eligius coinbases pay out to miners, not a pool wallet.  There is a share log and every block that gets paid out to the most recent shares adding up to the value of the block. If the round was unlucky then not all shares from that round will be paid. If the round was lucky then some old shares will be paid.  I don't know enough about the internals of p2pool to know whether this is possible.

EDIT: This isn't quite right since Eligius doesn't pay out until your balance reaches the minimum payout. Still the blocks rewards are paid directly to miners (who have accumulated a large enough balance to get paid out), not to a pool wallet.

Eligius usually pays directly using the coinbase transactions of blocks the pool mines, but from the Eligius CPPSRB page:

When a block is found:

• Go back 25 BTC into the share log, and reward those at PPS price.
• Delete those shares from the log
• If the entire database was paid and there are still funds left, include it in the next block's reward

When a block is orphaned:

• Undelete the shares rewarded for it

I may be wrong but I don't know how BTC saved for the next block's reward could be awarded using the coinbase since the full block reward needs to go somewhere.  I mine on Eligius and every once in a while I get regular payments (i.e., not "mined"/generated payouts) to my address so I just assume it was from coins the pool had to pay to itself due to a lucky streak. (or nice people donating/miners typing their address incorrectly :p )

@spooderman like kjj, smooth, and lenny_ above me said you won't be charged fees when you get paid via p2pool, but mining on p2pool with a low hashrate will lead to a bunch of tiny payments into your wallet which will cost a lot in transaction fees later on when you try to spend the coins if your wallet isn't smart about making small enough transactions which can be sent with no fee.  Normal pools just wait until you have mined a certain amount before sending you a payout which leads to fewer transaction fees when you later spend the coins but I believe p2pool sends you a payout as long as you have found at least one p2pool share during a round which means a bunch of tiny transactions if you usually only find a couple of shares per round.

However you can get around this by trial and error if you're using bitcoin-qt: keep trying to send transactions and decrease the amount until it doesn't ask for a fee.  You can keep paying into an address of an exchange account and then withdraw all the coins from the exchange back into your wallet in one transaction which should make spending your coins much cheaper.
(I hope the exchange wallet has an intelligent way of bundling these tiny inputs into other transactions so they aren't hit with fees!)

[smooth]

• If the entire database was paid and there are still funds left, include it in the next block's reward

This never happens in practice. The database has shares that are months or maybe years old. It is not clear to me why this is other than bad luck early on by the pool (which would build up a backlog of shares in the database). Or maybe it is because p2pool apparently outperforms the centralized pools, so they will always have a bit more than their share of "bad luck."

I mine on Eligius and every once in a while I get regular payments (i.e., not "mined"/generated payouts) to my address so I just assume it was from coins the pool had to pay to itself due to a lucky streak. (or nice people donating/miners typing their address incorrectly :p )

I have no idea what those are. I don't think I've received them (but not positive, I've had payments going to several different wallets/addresses over the years so I may not have noticed). Maybe the operator correcting for some malfunctions?

[gmaxwell]

This never happens in practice.

It's because their payout is higher than the actual expectation once you consider orphans. It ends up being a transfer of value from those with the most weighed hash-power when the pool is unlucky those with the most weighed hashpower when its lucky. Fortunately thats still not hoppable since you don't know in advance which group you'll be in, but it's clearly unoptimal.  I'd tried to talk wizkid into paying 99% of PPS to the regular share queue and have the remaining 1% set aside and moved to a queue which is only paid after all the pay before it was paid out, but he was worried that other pools would claim that it was a fee. (Ideally the fraction would be determined by the true expected reward, but thats hard to estimate and it looks like eligius has a ~1% orphan rate, and indeed P2Pools' appears to be much lower)

I'm glad p2pool uses a pay-per last-n derived method where you don't have to worry about that kind of open loop mismatch.

[TierNolan]

It's because their payout is higher than the actual expectation once you consider orphans.

- pay each share 98% of expected
- pay remaining to last N shares, if queue is empty

I'm glad p2pool uses a pay-per last-n derived method where you don't have to worry about that kind of open loop mismatch.

I think the Eligius one has lower variance?

P2pool relies on honesty to ensure that miners update the bitcoin block they are building on (and the 1% reward for actually finding a block)?

In fact, what is the "cycle" time for an ASIC?

[smooth]

This never happens in practice.

It's because their payout is higher than the actual expectation once you consider orphans.

Okay yes that makes sense. As soon as you get a single orphan (the expectation of) those shares will stay in the log forever.

Thank you for the explanation.

[smooth]

I'd tried to talk wizkid into paying 99% of PPS to the regular share queue and have the remaining 1% set aside and moved to a queue which is only paid after all the pay before it was paid out, but he was worried that other pools would claim that it was a fee.

Why not just dump the oldest 25 BTC worth of shares from the log when you get an orphan? The orphan happened, so 25 BTC worth of shares is worthless and the pool isn't getting that value back ever. That's makes it sort of a PPLN with a very large N, while retaining the low variance of their method.