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

[twmz]

It seems p2pool.info miss block 169625 http://blockexplorer.com/block/00000000000006ca7bb5be0dbd0e2b81ea3b179f17218d6a2ca3116c4678ff78 (2012-03-04 16:03:22)

That wasn't found by the main p2pool but was found by a small fork of miners that didn't update.  Blocks found by the un-updated splinter group are now intentionally excluded.  Those miners are not part of the reported pool hashrate and so their blocks shouldn't be credited against the main pool's hashrate for the purposes of things like round length and luck statistics.  I suppose I could still display "fork blocks" in some distintive way, though.

[ChanceCoats123]

Bit of an unlucky day today, boys. I thought we were starting out hot with two blocks early on, but the last block is just dragging.

[O_Shovah]

Short note:

Bitcoin-qt, p2pool and Modular python bitcoin miner sucessfully running on 600 mhz ARM board.

I expect to get it running on rasberry pi as soon as i get my hands on one

[Elmojo]

Short note:
Bitcoin-qt, p2pool and Modular python bitcoin miner sucessfully running on 600 mhz ARM board.
I expect to get it running on rasberry pi as soon as i get my hands on one

That's awesome....but why?
I don't mean that in a sarcastic way, I'm actually asking what makes it awesome?
Without a GPU, wouldn't hashing on a Rasberry Pi be in the 1-2 Mh/s range?
Please tell me I'm missing something, those Pi's look SWEET and I need an excuse to get one!

[rjk]

Short note:
Bitcoin-qt, p2pool and Modular python bitcoin miner sucessfully running on 600 mhz ARM board.
I expect to get it running on rasberry pi as soon as i get my hands on one

That's awesome....but why?
I don't mean that in a sarcastic way, I'm actually asking what makes it awesome?
Without a GPU, wouldn't hashing on a Rasberry Pi be in the 1-2 Mh/s range?
Please tell me I'm missing something, those Pi's look SWEET and I need an excuse to get one!

It isn't hashing, it is simply acting as a server to hold p2pool and FPGA control software. It is awesome because it removes the need for a host computer to do the same task, and instead runs on a low power embedded device.

[DeathAndTaxes]

Short note:
Bitcoin-qt, p2pool and Modular python bitcoin miner sucessfully running on 600 mhz ARM board.
I expect to get it running on rasberry pi as soon as i get my hands on one

That's awesome....but why?
I don't mean that in a sarcastic way, I'm actually asking what makes it awesome?
Without a GPU, wouldn't hashing on a Rasberry Pi be in the 1-2 Mh/s range?
Please tell me I'm missing something, those Pi's look SWEET and I need an excuse to get one!

It isn't hashing, it is simply acting as a server to hold p2pool and FPGA control software. It is awesome because it removes the need for a host computer to do the same task, and instead runs on a low power embedded device.

Hell I don't have an FPGA but if someone can get p2pool and bitcoind running on one I would buy one (and donate the developer).  Pretty cool to have p2pool server for a 10GH/s farm running on 5 watts.

[Elmojo]

It isn't hashing, it is simply acting as a server to hold p2pool and FPGA control software. It is awesome because it removes the need for a host computer to do the same task, and instead runs on a low power embedded device.

Ah, I figured it may have been something like that.
I missed the part where this thread is about P2Pool! lol
That would be awesome, running P2Pool on a Pi, controlling a cluster of FPGAs or even full GPU machines, like D&T said.
I can't wait to see where this goes!

On another note, I'm trying to get my 7970 hashing in the pool, but I can't seem to get it configured correctly with diablominer.
It works fine with cgminer, but I get better hashrates with diablo.
Anyone got this combo (7970+diablominer+P2Pool) working?
I'd love a few pointers in the right direction.

[DeathAndTaxes]

Anyone else notice the trend of rising hashing power and falling users?  The avg hashing power per user is rising.

Higher network hashing power = lower block variance.
Higher network hashing power = higher share variance.

The above combo is a win-win for large miners but a win-lose for smaller ones.  Taken to the extreme at 600 GH/s avg block time is roughly 3 hour (current difficulty).  However at 600 GH/s share difficulty is ~1400 and for a 500 MH/s miner that means a 3.3 hour share time.  Further increases in network hashing power mean increased not reduced effective variance.

Of course there are potential solutions I am just pontificating.  

[rjk]

Anyone else notice the trend of rising hashing power and falling users?  The avg hashing power per user is rising.

Higher network hashing power = lower block variance.
Higher network hashing power = higher share variance.

The above combo is a win-win for large miners but a win-lose for smaller ones.  Taken to the extreme at 600 GH/s avg block time is roughly 3 hour (current difficulty).  However at 600 GH/s share difficulty is ~1400 and for a 500 MH/s miner that means a 3.3 hour share time.  Further increases in network hashing power mean increased not reduced effective variance.

Of course there are potential solutions I am just pontificating.  

Should we have a "CPU And Crappy GPU Miners" p2pool, and a "Big Ass Hard Core Rig Monster" p2pool?

[jamesg]

Anyone else notice the trend of rising hashing power and falling users?  The avg hashing power per user is rising.

Higher network hashing power = lower block variance.
Higher network hashing power = higher share variance.

The above combo is a win-win for large miners but a win-lose for smaller ones.  Taken to the extreme at 600 GH/s avg block time is roughly 3 hour (current difficulty).  However at 600 GH/s share difficulty is ~1400 and for a 500 MH/s miner that means a 3.3 hour share time.  Further increases in network hashing power mean increased not reduced effective variance.

Of course there are potential solutions I am just pontificating.  

Should we have a "CPU And Crappy GPU Miners" p2pool, and a "Big Ass Hard Core Rig Monster" p2pool?

Why can't p2pool stay the way it is with small miners migrating to p2pool subpools that send out 1 difficulty shares so small miners can still get their cut?

I would be willing to setup / run a small p2pool pool (i coin this a p2p pool, you heard it hear first) without a fee but would need some open source software to run since my time is limited.

[DeathAndTaxes]

Why can't p2pool stay the way it is with small miners migrating to p2pool subpools that send out 1 difficulty shares so small miners can still get their cut?

I have indicated that a backbone & subpool model is likely the best and the most lasting however it is the most complex.  In the short term p2pool front ends or proxys (conventional pools using p2pool as backend) which accept diff 1 work buys us some time.  forrest also indicated he will consider splitting p2pool at around 500 GH/s into two instances.


As far as I can tell three major things need to change (and be extensively tested) to allow a multilayered p2pool network:

1) Increase LP interval from 10 sec to something like 30 sec or 60 sec.  This is necessary otherwise sub pools LP interval will be too small. Remember a sub pool is dependent on the main network thus it's effective LP it whenever an LP occurs on the subpool OR main network.  Yes increasing LP target interval will increase difficulty so it shouldn't be done until sub pools are ready.

2) Allow p2pool shares to be assigned to multiple addresses.  Currently when you find a share the network records a single address (your payment address) in the reward split.  Sub pools will need to split 1 share across multiple miners (based on their current share chain work distribution).

3) Modify p2pool code to work as a sub pool. It means building a "subpool" level share chain, and reward splits and being aware of backbone p2pool status (blocks, LPs, share split, etc). The subpool is "invisible" to backbone p2pool accept when subpool submits shares.

[stevegee58]

Anyone else notice the trend of rising hashing power and falling users?  The avg hashing power per user is rising.

Higher network hashing power = lower block variance.
Higher network hashing power = higher share variance.

The above combo is a win-win for large miners but a win-lose for smaller ones.  Taken to the extreme at 600 GH/s avg block time is roughly 3 hour (current difficulty).  However at 600 GH/s share difficulty is ~1400 and for a 500 MH/s miner that means a 3.3 hour share time.  Further increases in network hashing power mean increased not reduced effective variance.

Of course there are potential solutions I am just pontificating.  

So in other words little guys like me doing 200 MH/s should probably just stick with a traditional pool until these p2pool issues are settled.

[DeathAndTaxes]

So in other words little guys like me doing 200 MH/s should probably just stick with a traditional pool until these p2pool issues are settled.

Well p2pool may not grow any larger so it may not matter.  Still at 200 MH/s p2pool growing larger doesn't help your variance.  In the long run you expected return remains the same however in short run higher hashing power results in more variance.

If/when p2pool grows larger it becomes more of a concern.  If you want to use a conventional pool try p2pool hybrid pool (it provides a 1 difficulty shares as a front end to p2pool so you can still support p2pool indirectly).

[stevegee58]

So in other words little guys like me doing 200 MH/s should probably just stick with a traditional pool until these p2pool issues are settled.

Well p2pool may not grow any larger so it may not matter.  Still at 200 MH/s p2pool growing larger doesn't help your variance.  In the long run you expected return remains the same however in short run higher hashing power results in more variance.

If/when p2pool grows larger it becomes more of a concern.  If you want to use a conventional pool try p2pool hybrid pool (it provides a 1 difficulty shares as a front end to p2pool so you can still support p2pool indirectly).

I'm willing to "support" p2pool if it benefits me, but I really don't see any advantage for myself at this point.  There are 0% fee pools that give me a steady payout.  I'm agnostic on the arguments about whether p2pool is "better" than normal pools so I'll do whatever gets me the most BTC on average.

Security isn't an issue for me since if the pool's web site is hacked I'll only lose less than 1 day's worth of mined BTC.

[O_Shovah]

I am running a x6500 FPGA board and a icraus FPGA board with my ARM board as host.
I hope this week a bitforce box from butterflylabs will join them

The ARM plattform is using ~4W out the wall at max performance, so slightly less in its current operation mode.
I will do simmilar systems on panda board and rasberry pi as soon as i get them.
These systems feature full operability as Host computers (ethernet,HDMI,USB OTG,...)

I will also create a thread with HowTo's and downloadble images for these systems soon.
I just quite busy at work at the moment so i have to ask your patience for some time.

[Gabi]

Anyone else notice the trend of rising hashing power and falling users?  The avg hashing power per user is rising.

Higher network hashing power = lower block variance.
Higher network hashing power = higher share variance.

The above combo is a win-win for large miners but a win-lose for smaller ones.  Taken to the extreme at 600 GH/s avg block time is roughly 3 hour (current difficulty).  However at 600 GH/s share difficulty is ~1400 and for a 500 MH/s miner that means a 3.3 hour share time.  Further increases in network hashing power mean increased not reduced effective variance.

Of course there are potential solutions I am just pontificating.  

So in other words little guys like me doing 200 MH/s should probably just stick with a traditional pool until these p2pool issues are settled.

340mhash/s here and i have no problems so far... sure i just find some shares but since each p2pool share right now is worth like 620 normal shares... i'm fine

[forrestv]

The transition ~40 hours ago went better than expected. Thanks to all for upgrading.

Now that we're using the new implementation, miners can volunteer to raise their share difficulty by adding something like "/1300" to the end of their miners' usernames. The 1300 is the difficulty of your own shares, and can be changed but must be higher than P2Pool's difficulty (currently 650) to have any effect. I urge anyone whose variance is dominated by P2Pool's block finding to try this (which really means anyone who gets more than a few shares per hour). This has the effect of lowering P2Pool's difficulty for the benefit of small miners, which may let P2Pool grow further.

Last, a side note (mainly to DeathAndTaxes): I just pushed a commit that will add another option ("+1") that lets you choose your pseudoshare difficulty, so you can fix it to some value.

[Elmojo]

Sorry to (semi)re-post this, but I'd really like to jump into the pool!
I'm trying to get my 7970 hashing in the pool, but I can't seem to get it configured correctly with diablominer.
It works fine with cgminer, but I get better hashrates with diablo.
Anyone got this combo (diablominer+P2Pool) working?
I can get diablo running, but it just sits at 0.0Mh/s, even though the P2Pool app seems to be working fine.
Is there a log file or something I can post that will help track down the issue?
I followed the setup instructions in the first few posts, but they don't seem to address diablo specifically, and I think I'm just missing some minor (but important) bit of string to get it hashing.
Any help appreciated.

[forrestv]

I can get diablo running, but it just sits at 0.0Mh/s, even though the P2Pool app seems to be working fine.

It sounds like you might not have the port set .. are you running DiabloMiner with "-r 9332"? If you are, can you pastebin the command that you're running and its output? Check if P2Pool has any error messages and run DiaboMiner with the debug flag too (I think it's just -d).

[Diapolo]

The transition ~40 hours ago went better than expected. Thanks to all for upgrading.

Now that we're using the new implementation, miners can volunteer to raise their share difficulty by adding something like "/1300" to the end of their miners' usernames. The 1300 is the difficulty of your own shares, and can be changed but must be higher than P2Pool's difficulty (currently 650) to have any effect. I urge anyone whose variance is dominated by P2Pool's block finding to try this (which really means anyone who gets more than a few shares per hour). This has the effect of lowering P2Pool's difficulty for the benefit of small miners, which may let P2Pool grow further.

Last, a side note (mainly to DeathAndTaxes): I just pushed a commit that will add another option ("+1") that lets you choose your pseudoshare difficulty, so you can fix it to some value.

By "adding to the username" you mean the username specified in bitcoin.conf (rpcuser=)?

Dia