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

[Aseras]

Do Avalon ASICs work with p2pool out of the box?  If not, does anyone know what settings need to be used?
Thanks!

simple answer, yes. long answer no, you need a special build of p2pool and the firmware we have made to fix delays and absurd long poll wait times because of the avalon driver built in delays.

To those just joining the conversation, no p2pool or cgminer developer had access to an avalon until a user, Aseras, graciously donated his machine time to a cgminer developer this week. All avalon owners owe him a thank you for working out the new firmware and contributions to the cgminer code (that has not been released yet) along with ckolivas.

test firmware is available. PM me if anyone wants it. it has a newer cgminer 2.10.5a build that has some tweaks and bugfixes and seems to be much more stable. it has better longpoll and less delay, making p2pool slightly better but still > 20% DOA. efficiency is up to 95% or so, but so many rejects and DOA on p2pool.

based off of latest gits by xiangfu and ckolivas. It is basically following these directions https://en.bitcoin.it/wiki/Avalon ( which don't work yet BTW debugging with xiangfu to make them idiot proof )

development should continue slowly. trying to work out things.

I'm really working on p2pool, quickly coming up to speed in python right now mostly it's just hacks and workarounds to make it work. I may make a public fork, but considering my recent experiment and crashing the terracoin network for a day, I'll hold off

ckolivas is trying to get code better as much as possible and xiangfu is doing them same plus bugfixes and better avalon drivers internally.

[Richy_T]

Well, in a case of "when you have eliminated the impossible, whatever remains, however improbable, must be the truth", I turned off QOS on my router (Tomato based) and though it's still now downloading shares, the output now gives me a lot more confidence. Odd as though it wasn't in the top tier, it should still have had plenty of bandwidth. Oh well. Hope to be mining in the pool asap. I'll get around to debugging the QOS when I can.

Edit: now mining. Yay!

[lumberjack]

I'm really working on p2pool, quickly coming up to speed in python right now mostly it's just hacks and workarounds to make it work. I may make a public fork, but considering my recent experiment and crashing the terracoin network for a day, I'll hold off

It's looking like you broke it worse than that since the difficulty swung back and forth, but in a way you've now exposed the fact that a rolling daily average for difficulty was not the proper way to fix the underlying issue   EDIT: Actually, it seems to be stuck for 5+ hours right now, other miners abandoned it because difficulty got too high. Maybe a few minutes on with your avalon it to get some blocks moving again would help things.



Unrelated question for anyone - I've not had a chance to peruse the P2Pool source. How hard is it to add other coin networks? With TRC dying/dead a day after I decided to point a miner at it for the first time, I'm curious to see how things like nvc & ppc perform, but the pools I've tried are fairly unstable.

[-ck]

I've spent quite a bit of time hacking on the Avalon(s) the last few days thanks to Aseras and then Xiangfu giving me access to them. I've come to the conclusion that any latency issues that prevent it mining on p2pool should be resolved in the next cgminer release incorporating the changes I've committed to the code. The max latency for restart should be on the order of 100ms now which is fine even for 10s longpolls. Hopefully it will talk to p2pool better via stratum as well.

[Syke]

I've spent quite a bit of time hacking on the Avalon(s) the last few days thanks to Aseras and then Xiangfu giving me access to them. I've come to the conclusion that any latency issues that prevent it mining on p2pool should be resolved in the next cgminer release incorporating the changes I've committed to the code. The max latency for restart should be on the order of 100ms now which is fine even for 10s longpolls. Hopefully it will talk to p2pool better via stratum as well.

Truly impressive. I can't wait to see a bunch of ASICs join p2pool.

[-ck]

Alas the p2pool end is not exactly playing ball with the avalons though, and I'm not touching the p2pool code itself, just trying to make Avalons work well and be prepared to work with p2ool design, assuming p2pool scales and does all the right things. The current p2pool git master does not meet that criteria, and Aseras himself has been playing with that as you can see above.

[Richy_T]

I think I'm mining. My miners look like they are mining. Fans are spinning high and the room is getting warm. Yet the top graph is flat at zero even on the hour view. Is this supposed to be the case?

Local rate on main page also says 0. What is happening to the MH/s that my miners are claiming?

[forrestv]

I think I'm mining. My miners look like they are mining. Fans are spinning high and the room is getting warm. Yet the top graph is flat at zero even on the hour view. Is this supposed to be the case?

Local rate on main page also says 0. What is happening to the MH/s that my miners are claiming?

If your miners have a Bitcoin address username, this will happen. Those graphs and numbers are for the main address, which defaults to one requested from bitcoind and can be set with the -a option on P2Pool. Letting P2Pool manage the payout address is the recommended course.

However, your local rate and current payout should show up under the "Miners" header on the graphs page.

[Richy_T]

I think I'm mining. My miners look like they are mining. Fans are spinning high and the room is getting warm. Yet the top graph is flat at zero even on the hour view. Is this supposed to be the case?

Local rate on main page also says 0. What is happening to the MH/s that my miners are claiming?

If your miners have a Bitcoin address username, this will happen. Those graphs and numbers are for the main address, which defaults to one requested from bitcoind and can be set with the -a option on P2Pool. Letting P2Pool manage the payout address is the recommended course.

However, your local rate and current payout should show up under the "Miners" header on the graphs page.

My miners are using the noob/sauce user & password but I did use the -a option to set the address to my bitcoinspinner wallet.

[forrestv]

My miners are using the noob/sauce user & password but I did use the -a option to set the address to my bitcoinspinner wallet.

Hm. Litecoin mining?

[Smoovious]

My miners are using the noob/sauce user & password but I did use the -a option to set the address to my bitcoinspinner wallet.

bitcoinspinner? do you know if that wallet can receive generated coin? you should confirm that before you mine to that wallet.

-- Smoov

[Richy_T]

My miners are using the noob/sauce user & password but I did use the -a option to set the address to my bitcoinspinner wallet.

bitcoinspinner? do you know if that wallet can receive generated coin? you should confirm that before you mine to that wallet.

-- Smoov

I had no idea that one could not. I will try without setting the address later and see if that helps any.

[PatMan]

Alas the p2pool end is not exactly playing ball with the avalons though, and I'm not touching the p2pool code itself, just trying to make Avalons work well and be prepared to work with p2ool design, assuming p2pool scales and does all the right things. The current p2pool git master does not meet that criteria, and Aseras himself has been playing with that as you can see above.

Appreciate all your hard work ckolivas, fair play to you.  Can I ask you what, in your experienced opinion, do you think needs to be done to the p2pool code/program in order for it to play nicely with avalons/stratum? In laymans terms if possible please.....

[Aseras]

The avalon is not like 1 very fast miner. It's not a Ferrari. It's a freight train. It's a lot of little miners working together. On p2pool with short long polls, it has too much "momentum" going on to stop and start again quickly so you loose a significant amount of work.  99% of work at high difficulties are returned in 15 seconds. Some take as long as 20+ seconds.

Right now most of p2pool work is "weak". I use +16 and +32 difficulty shares. That seems to even out the earnings since a single valid share at that difficulty counts as dozens or more at normal difficulty, and most work is discarded anyways.

The fork I made was to extend LP time to 15 seconds, 20 seconds and 30 seconds to see how each worked. 15 worked almost as well as 30. 20 seemed to be the sweet spot though. 10 still causes considerable lost work to doa or reject shares.

My suggestion currently would be to make a hard fork and create a p2pool-ASIC fork, and keep the current p2pool as p2pool-gpu. They would be completely separate.

Although that's just my assumption as I don't know how hard it would be to integrate them, or how much it would hurt gpu.

I do feel however that any gpu miners days are seriously numbered. And the quickly rising difficulty is going to bury them and necessitate changes to p2pool to accommodate the longer time needed to do much larger difficulty, longer work times. it will pretty much make gpu mining bitcoin unworkable except for the complete die hards, the stupid or botnets within a few months.

[Richy_T]

Still getting 0H/s even with the -a removed. The output from my miner doesn't look right though. I don't know if that's just a p2pool thing though. Any ideas?

07/04/2013 23:31:18, started OpenCL miner on platform 0, device 1 (Cypress)
 07/04/2013 23:31:18, Setting server (noob @ localhost:9332)
localhost:9332 07/04/2013 23:31:18, checking for stratum...
localhost:9332 07/04/2013 23:31:19, diverted to stratum on localhost:9332
localhost:9332 07/04/2013 23:31:19, Setting new difficulty: 0.999984741211
localhost:9332 07/04/2013 23:31:20, authorization failed with noob:sauce@localhost:9332
localhost:9332 07/04/2013 23:31:20, Setting new difficulty: 0.999984741211
localhost:9332 07/04/2013 23:31:23, IO errors - 1, tolerance 2
localhost:9332 07/04/2013 23:31:49, checking 1615820210 <= 0
localhost:9332 07/04/2013 23:31:50, checking 3741571565 <= 0
localhost:9332 07/04/2013 23:32:06, checking 1991599613 <= 0
localhost:9332 07/04/2013 23:32:08, checking 2221794573 <= 0
localhost:9332 07/04/2013 23:32:18, 0:1:Cypress [298.922 MH/s (~0 MH/s)] [Rej: 0/0 (0.00%)]

This is the command line if that helps

./poclbm.exe "http://noob:sauce@localhost:9332" --device=1 --platform=0 --verbose -v -w 128 -r 60

[mdude77]

I do feel however that any gpu miners days are seriously numbered. And the quickly rising difficulty is going to bury them and necessitate changes to p2pool to accommodate the longer time needed to do much larger difficulty, longer work times. it will pretty much make gpu mining bitcoin unworkable except for the complete die hards, the stupid or botnets within a few months.

Actually... people have been saying that for 6 months now.  ASICs are still very hard to come by (and there's no way I'm spending $10k on an avalon).  At $150/BTC, and electricity at $.05, I'm still very profitable with GPUs.  Not reaping it in hand over fist like the lucky folks who got batch 1/2 avalons, but still very profitable.  Probably more so than it's ever been, even back pre-halving days.  The rising value of BTC is what people seem to forget.  Yes difficulty is 2x what it was a few months ago, but so is the value of BTC.

As for forking.. I'm not using p2pool as long as there's < 1TH in it.  The variance is simply too much for me.  If the ASIC fork is unusable by GPUs, then I simply will never use p2pool again.  It's unfortunate, but that's how it is.

M

[lenny_]

As for forking.. I'm not using p2pool as long as there's < 1TH in it.  The variance is simply too much for me.  If the ASIC fork is unusable by GPUs, then I simply will never use p2pool again.  It's unfortunate, but that's how it is.

Instead of switching and helping us you're sitting there and waiting for miracle. Act now and do something, waiting will never help it to raise to that hashrate level. I just switched my miners to p2pool couple of days ago, I will accept variance because I am getting more BTC than anywhere else.

Code:

Local dead on arrival: ~0.9% (0-2%)

Code:

Shares: 1056 (79 orphan, 25 dead) Stale rate: ~9.8% (8-12%) Efficiency: ~114.2% (111-117%)

[mdude77]

As for forking.. I'm not using p2pool as long as there's < 1TH in it.  The variance is simply too much for me.  If the ASIC fork is unusable by GPUs, then I simply will never use p2pool again.  It's unfortunate, but that's how it is.

Instead of switching and helping us you're sitting there and waiting for miracle. Act now and do something, waiting will never help it to raise to that hashrate level. I just switched my miners to p2pool couple of days ago, I will accept variance because I am getting more BTC than anywhere else.

Code:

Local dead on arrival: ~0.9% (0-2%)

Code:

Shares: 1056 (79 orphan, 25 dead) Stale rate: ~9.8% (8-12%) Efficiency: ~114.2% (111-117%)

As I said, the variance is too much for me. 

M

[PatMan]

I do feel however that any gpu miners days are seriously numbered. And the quickly rising difficulty is going to bury them and necessitate changes to p2pool to accommodate the longer time needed to do much larger difficulty, longer work times. it will pretty much make gpu mining bitcoin unworkable except for the complete die hards, the stupid or botnets within a few months.

Actually... people have been saying that for 6 months now.  ASICs are still very hard to come by (and there's no way I'm spending $10k on an avalon).  At $150/BTC, and electricity at $.05, I'm still very profitable with GPUs.  Not reaping it in hand over fist like the lucky folks who got batch 1/2 avalons, but still very profitable.  Probably more so than it's ever been, even back pre-halving days.  The rising value of BTC is what people seem to forget.  Yes difficulty is 2x what it was a few months ago, but so is the value of BTC.

As for forking.. I'm not using p2pool as long as there's < 1TH in it.  The variance is simply too much for me.  If the ASIC fork is unusable by GPUs, then I simply will never use p2pool again.  It's unfortunate, but that's how it is.

M

Agreed. Also, if I DID have an Avalon I would point it at a more compatible pool that would get the most out of it, obviously. Forking p2pool is not a good idea at all, what would happen if Avalon chips decide to pack in after 6 months? Far better for everyone if p2pool was compatible with ALL miners, as is the norm.

lenny, if you consider 3 blocks/payouts  per week the most you've earned than anywhere else you must have been doing something badly wrong before......

[mdude77]

I do feel however that any gpu miners days are seriously numbered. And the quickly rising difficulty is going to bury them and necessitate changes to p2pool to accommodate the longer time needed to do much larger difficulty, longer work times. it will pretty much make gpu mining bitcoin unworkable except for the complete die hards, the stupid or botnets within a few months.

Actually... people have been saying that for 6 months now.  ASICs are still very hard to come by (and there's no way I'm spending $10k on an avalon).  At $150/BTC, and electricity at $.05, I'm still very profitable with GPUs.  Not reaping it in hand over fist like the lucky folks who got batch 1/2 avalons, but still very profitable.  Probably more so than it's ever been, even back pre-halving days.  The rising value of BTC is what people seem to forget.  Yes difficulty is 2x what it was a few months ago, but so is the value of BTC.

As for forking.. I'm not using p2pool as long as there's < 1TH in it.  The variance is simply too much for me.  If the ASIC fork is unusable by GPUs, then I simply will never use p2pool again.  It's unfortunate, but that's how it is.

M

Agreed. Also, if I DID have an Avalon I would point it at a more compatible pool that would get the most out of it, obviously. Forking p2pool is not a good idea at all, what would happen if Avalon chips decide to pack in after 6 months? Far better for everyone if p2pool was compatible with ALL miners, as is the norm.

lenny, if you consider 3 blocks/payouts  per week the most you've earned than anywhere else you must have been doing something badly wrong before......

Unless I'm missing something, forking isn't necessarily a bad thing.  In fact, I'm pretty sure there are already other forks out there.  This is open source after all.  It shouldn't affect the "GPU" pool at all.

M