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

[Crystallas]

Although the BFL SC devices supposedly support the ZPX command that was added for MiniRig FPGA devices to allow it to work on shorter nonce ranges (and therefore wake up for new work much more frequently and therefore have much lower DOA %), neither cgminer nor bfgminer (I tried them both) use this command with the BFL SC devices and instead use the ZNX queue command.

Check the BFL forums. We asked them why they didn't implement the command because we tried it and it doesn't work on BFL SC devices.

They did not respond. I think BFL are too busy drowning in fail to respond to petty questions like these.

BFL_StevenM did say something in the chat about reviewing some FPGA wake script for the ASICs yesterday. What exactly, and why nobody has gotten back to you or others, I dunno.

[Amph]

2-3% doa with 0.165s at default setting on 0.8.2 version
for the moment it seem that income are lower...

[PrintMule]

Although the BFL SC devices supposedly support the ZPX command that was added for MiniRig FPGA devices to allow it to work on shorter nonce ranges (and therefore wake up for new work much more frequently and therefore have much lower DOA %), neither cgminer nor bfgminer (I tried them both) use this command with the BFL SC devices and instead use the ZNX queue command.

Check the BFL forums. We asked them why they didn't implement the command because we tried it and it doesn't work on BFL SC devices.

They did not respond. I think BFL are too busy drowning in fail to respond to petty questions like these.

More like drowning in stolen money, lol

[gyverlb]

Although the BFL SC devices supposedly support the ZPX command that was added for MiniRig FPGA devices to allow it to work on shorter nonce ranges (and therefore wake up for new work much more frequently and therefore have much lower DOA %), neither cgminer nor bfgminer (I tried them both) use this command with the BFL SC devices and instead use the ZNX queue command.

Check the BFL forums. We asked them why they didn't implement the command because we tried it and it doesn't work on BFL SC devices.

They did not respond. I think BFL are too busy drowning in fail to respond to petty questions like these.

I'll update the guide mentioning these problems.
As long as a device doesn't support being fed new work to switch to with nearly no latency or work on short nonce ranges with at most 0.1s spent on them P2Pool payouts will start to be noticeably decreased (>1% lost income).
IIRC BFL did know about this in the early days, there were discussions of the ZPX command or an equivalent last summer for their ASICs. Guess they simply don't care about such details.

[bitpop]

1. Fuck bfl, suck bird dick

2. I'm using lancelots and I remember setting cg miner to submit stales. Am I punished by p2pool for submitting stales? These are the rejected ones .1s after stratum.

[twmz]

Although the BFL SC devices supposedly support the ZPX command that was added for MiniRig FPGA devices to allow it to work on shorter nonce ranges (and therefore wake up for new work much more frequently and therefore have much lower DOA %), neither cgminer nor bfgminer (I tried them both) use this command with the BFL SC devices and instead use the ZNX queue command.

Check the BFL forums. We asked them why they didn't implement the command because we tried it and it doesn't work on BFL SC devices.

They did not respond. I think BFL are too busy drowning in fail to respond to petty questions like these.

Ah, ok.  I just read their (apparently incorrect) documentation on the protocol for SC devices.  My bad.  

Given this, nothing can be done to make these devices work with p2pool effectively short of changes by ButterflyLabs.

[gyverlb]

1. Fuck bfl, suck bird dick

2. I'm using lancelots and I remember setting cg miner to submit stales. Am I punished by p2pool for submitting stales? These are the rejected ones .1s after stratum.

P2Pool actually asks for stales as they can be valid blocks. There's no reason it would punish you :-)

[bitpop]

Ok so I had it set right. Yeah it would be dumb punishing something you could just hide. So where do stales and doa come from? A valid can just be late?

1. Fuck bfl, suck bird dick

2. I'm using lancelots and I remember setting cg miner to submit stales. Am I punished by p2pool for submitting stales? These are the rejected ones .1s after stratum.

P2Pool actually asks for stales as they can be valid blocks. There's no reason it would punish you :-)

[gyverlb]

Ok so I had it set right. Yeah it would be dumb punishing something you could just hide. So where do stales and doa come from? A valid can just be late?

stales (including doa and orphan shares) are when you submit a share that isn't included in P2Pool's sharechain. DOA are because your miner was working on obsolete dat, orphans are because some other node managed to publish a share at roughly the same time than yours and the P2Pool network chose his share instead of yours.
Both type of stales don't reduce your income unless you have more than the average miner (as both can generate blocks which will have their income distributed to everyone).

There's a bit more info in the guide in my signature and on the wiki.

[bitpop]

Thanks. I'm not even sure submit-stale does anything in cgminer any more, maybe it is the p2pool telling cgminer to submit stales. either way, glad its recommended.

Ok so I had it set right. Yeah it would be dumb punishing something you could just hide. So where do stales and doa come from? A valid can just be late?

stales (including doa and orphan shares) are when you submit a share that isn't included in P2Pool's sharechain. DOA are because your miner was working on obsolete dat, orphans are because some other node managed to publish a share at roughly the same time than yours and the P2Pool network chose his share instead of yours.
Both type of stales don't reduce your income unless you have more than the average miner (as both can generate blocks which will have their income distributed to everyone).

There's a bit more info in the guide in my signature and on the wiki.

[-ck]

Thanks. I'm not even sure submit-stale does anything in cgminer any more, maybe it is the p2pool telling cgminer to submit stales. either way, glad its recommended.

cgminer respects the pool message telling it to submit stales.

[zvs]

yah, it 'makes' it submit stales, since p2pool stales arent network stales

and orphans are much worse than DOA, in most cases DOA is just a matter of your latency to wherever you are mining at, orphans means wherever you are mining at isnt propagating it's shares fast enough (or occasionally you'll get screwed when someone that was rly slow happens to get two in a row)

i use more resources using my home computer for essentially the same % of lost shares as I get mining at nogleg.  they're just orphans instead of DOA.   but if you're some person that wants to mine at a remote pool, you certainly wouldnt want to mine at my home pool getting all the orphans, since then you'd still get just as many orphans + the DOAs associated with your latency

[xgtele]

Graphs on http://p2pool.info/ are not working  

UPDATE: fixed

[lenny_]

Is there anyone willing to make BE Blades working with p2pool?
Just to remind Last forrestv patch to stratum mining proxy make it working now nicely with p2pool, (c)gminer or any other now can connect to stratum proxy and to p2pool. But BE Blades still not working that way

[gyverlb]

Is there anyone willing to make BE Blades working with p2pool?
Just to remind Last forrestv patch to stratum mining proxy make it working now nicely with p2pool, (c)gminer or any other now can connect to stratum proxy and to p2pool. But BE Blades still not working that way

I wouldn't hold my breath: although they have mining software on board, AFAIK there's no way to update it if the problem lies on their end (which it seems to be if other software can use the stratum proxy with p2pool fine). You would be better off asking friedcat for help than p2pool devs.

Last time someone pointed his fingers to p2pool as a source of problems with an ASIC device, the problem was the ASIC device itself (BFL SC with no support for interrupting work or short nonce ranges). There are devices that simply won't work or work properly with short intervals between LPs or stratum work restarts.

[bitpop]

This may be a good place to ask
 Looking for a developer to patch bytecoin https://bitcointalk.org/index.php?topic=223904.0

[davidkassa]

Is there anyone out there running p2pool on Amazon AWS for themselves and a few friends? How is performance/efficiency? What instance size do you use and what is the monthly cost? I was starting to test this out on the free tier as a proof-of-concept, but the free tier can't even keep bitcoind up with the memory usage.

[Smoovious]

I tried running my jalapeno on a segregated p2pool network (2 nodes only, both nodes on computers I control, no outside connections permitted).

The problem I've been observing using the jalapeno with p2pool is definitely revolving around how the jalapeno returns work. It doesn't look like valid work is returned as soon as it was found, but sending back multiple valids at the same time.

The effect is, the first valid has a good chance of being a valid share, but the rest of the valids sent after it are always rejected.

While they may still be a valid block, I think the same problem arises there too. First valid solves the block, the rest of the valids are rejected as they won't build on the block the first valid solved. Granted, this won't happen often, but it will happen.

The problem when it comes to p2pool is, the miner using the jalapeno, might not get accurate credit for the amount of work they're doing, just because the jalapeno won't abort the work it is doing to start new work, so those miners that are using equipment that will do that, will always have an edge just because they will be more responsive to change.

As the jalapeno keeps mining for more results after it already found a valid result, and not yet sending it along, it is just wasting power at that point.

Just my observations this past week.

Protocol-wise, it seems to behave, it just doesn't appear to handle valid results efficiently.

All that said... after almost 2 years of being stuck at ~160MH/s, now mining at ~5.2GH/s? I don't quite believe I'm watching my own stats yet, thinking I'm being punked by a feed of someone else's screenshots. This is taking some getting used to. XD

-- Smoov

[tiktoc]

Is there anyone out there running p2pool on Amazon AWS for themselves and a few friends? How is performance/efficiency? What instance size do you use and what is the monthly cost? I was starting to test this out on the free tier as a proof-of-concept, but the free tier can't even keep bitcoind up with the memory usage.

Yeah I have done it with
Memory optimized    m2.xlarge    64-bit    2    6.5    17.1    1 x 420    -    Moderate

using 10gig as a ram drive  for bitcoind.

But it was getting close to/was using all the memory running two bitcoinds.

So was contemplating Memory optimized    m2.2xlarge    64-bit    4    13    34.2    1 x 850

If you dont need rock solid uptime, you can do it on the cheap using spot instances using your own ami and a ebi volume for persistent data for p2pool and first bitcoind data.

Without volume charge your looking at around $25 to $30 for the 17gig memory one and double that for the 34gig memory one a month for usa. Other pricing areas you'll have to check.

With the second bitcoind  in ram drive the getwork latencies were pretty low  and was about 115% efficiency.

Was getting jumps now and then but, might have been the instance running out of memory, didnt look into it much more.

[bitpop]

Mine below in sig is running on aws
try min, its fee free