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

[DeathAndTaxes]

Yeah I have it set to /1000+1 also.  With 12GH/s I don't notice the higher share variance.  Hopefully that gives some breathing room for smaller miners.  I am not sure if that is a stat that can be tracked; % of hash power using higher share diff or maybe more useful is "diff per GH stat".  i.e right now network is 749 diff for 348  = 2.15 diff/GH.  Lets see if we can get that below 2!

If it can be tracked and charted we can encourage more people to help out.

[echris1]

Yeah I have it set to /1000+1 also.  With 12GH/s I don't realize notice the higher share variance.  Hopefully that gives some breathing room for smaller miners.  I am not sure if that is a stat that can be tracked.  % of hash power using higher share diff or maybe more useful is "diff per GH stat".  i.e right now network is 749 diff for 348  = 2.15 diff/GH.  Lets see if we can get that below 2!

If it can be tracked and charted we can encourage more people to help out.

Is there a recommended setting for a certain hashrate?

I have a total of 2.5GH/s, but some of the individual cards are only 200Mh/s.  If I understand it correctly, that shouldn't really matter since they are all looking together, but some clarification here would be nice. 

Also, does P2Pool then keep track of shares in decimal points, or am I just off on how that part works as well?

[Isokivi]

Yeah I have it set to /1000+1 also.  With 12GH/s I don't realize notice the higher share variance.  Hopefully that gives some breathing room for smaller miners.  I am not sure if that is a stat that can be tracked.  % of hash power using higher share diff or maybe more useful is "diff per GH stat".  i.e right now network is 749 diff for 348  = 2.15 diff/GH.  Lets see if we can get that below 2!

If it can be tracked and charted we can encourage more people to help out.

Im down with this, what would your recomended settings be for 1.7GS/s. Also I wouldnt mind a detailed howto on the setup, I browsed ~5 pages backwards in this thread and cant seem to find one.

[Ente]

Lets see.. Right now, we can expect a found block (whole p2pool network finds a bitcoin block) every 5 hours.
To not have any significant impact on payout, you should constantly find several shares in every round. So if you set the difficulty so that you find around 5 shares in 5 hours, you wont have much additional variance by the amount of shares you find..
Which would mean: 5 shares every 5 hours, or one share every 60 mins, which calculates to:

http://www.alloscomp.com/bitcoin/old_calculator.php

<1Gh/s --> default difficulty
1Gh/s --> difficulty 1000
2Gh/s --> difficulty 2000
etc

Which would give you a share every 71 mins.
Which would give you 4 to 5 shares every block.

Easy rule of thumb, one thousand for every gigahash.
For us who actually understand the system with variance and all, this seems reasonable?

You could of course set it smaller or to default. But the actual variance you then see wouldnt go down a lot, since then the variance of the block-finding is a lot larger than the variance by share-finding.
And, also, it all is better comparable and nicer with a constant 1000 or 2000 difficulty instead of the default variable and ever changing difficulty ;-)

Ente

[Ente]

Im down with this, what would your recomended settings be for 1.7GS/s. Also I wouldnt mind a detailed howto on the setup, I browsed ~5 pages backwards in this thread and cant seem to find one.

Its all much easier than expected!
You dont have to change anything on your p2pool, you just change the miner's name:

Minername/sharedifficulty+localdifficulty

So for example:

Miner/1500+1

You set this on all miners. You can give different difficulty targets, but I dont see too much sense in this..

Miner01/1500+1
Miner02/1500+1
Miner03/1500+1

Maybe this is worth stating on the first page, and/or hinting in the help/wiki/etc?
Lets drop dat diff!

Ente

[DeathAndTaxes]

Lets see.. Right now, we can expect a found block (whole p2pool network finds a bitcoin block) every 5 hours.
To not have any significant impact on payout, you should constantly find several shares in every round. So if you set the difficulty so that you find around 5 shares in 5 hours, you wont have much additional variance by the amount of shares you find..
Which would mean: 5 shares every 5 hours, or one share every 60 mins, which calculates to:

http://www.alloscomp.com/bitcoin/old_calculator.php

<1Gh/s --> default difficulty
1Gh/s --> difficulty 1000
2Gh/s --> difficulty 2000
etc

Which would give you a share every 71 mins.
Which would give you 4 to 5 shares every block.

Easy rule of thumb, one thousand for every gigahash.
For us who actually understand the system with variance and all, this seems reasonable?

You could of course set it smaller or to default. But the actual variance you then see wouldnt go down a lot, since then the variance of the block-finding is a lot larger than the variance by share-finding.
And, also, it all is better comparable and nicer with a constant 1000 or 2000 difficulty instead of the default variable and ever changing difficulty ;-)

Ente

1 share per hour is a steep target.  I wouldn't want to do that myself nor would I expect anyone else to do so.

Maybe something more like
Goal 5 shares per hour.  Difficulty Target:  170 * GH/s
Goal 4 shares per hour.  Difficulty Target:  210 * GH/s
Goal 3 shares per hour.  Difficulty Target:  280 * GH/s
Goal 2 shares per hour.  Difficulty Target:  420 * GH/s
Goal 1 shares per hour.  Difficulty Target:  840 * GH/s

>5 shares per hour should be pretty low variance (most of your variances come from block time so helping network get larger is beneficial)
1 share per hour is going to be pretty hard variance.

Miners can pick a target that gives them a variance they can stomach.  Everyone's treshold is different.  Remember avg can be deceptive.  1/6th to 6x the average isn't that rare in the short term (say over 100 shares).  So shooting for goal of 1 share per hour means you will have some hours with 6 shares and sometimes go 6 hours (oops missed an entire block = payout) without a single share.  That can be frustrating for some miners. (Total brain fart.  p2pool is PPLNS so you don't need a share during a block to be paid for the block just 1 valid share in the share chain)



I will ask Meni for some better analysis (nobody knows this stuff better than him) but hopefully this provides some insight.

I was running only 1000 as my target but looking at the numbers I am going to try running at a goal of 4 shares per hour.  At 12GH/s that gives me a target difficulty of 210*12 = 2500.

[echris1]

Hmmm, I seem to have pretty long stretches of bad luck sometimes, and I only get three shares an hour as it is.

So, I change all of my miners to the new difficulty I want to use, even the slower ones? 

I keep thinking that in some way my 200Mhs desktop miner will be affected by this much more than my 1.1Gh/s rig.  Someone reassure me I'm wrong and I'd be willing to bump it up a bit to something like 2 an hour.

[DeathAndTaxes]

It depends on what you care about.

If you care about a single GPU always getting an even # of shares per hour then don't raise difficulty.  If all you care about it the TOTAL revenue per block then raising diff on all miners is the only thing that matters.

Honestly w/ 2.29 GH/s I am not sure I would raise difficulty and if I did it would be only a small amount.  Maybe to 1000?  Now 29 GH/s sure.

[Isokivi]

Im down with this, what would your recomended settings be for 1.7GS/s. Also I wouldnt mind a detailed howto on the setup, I browsed ~5 pages backwards in this thread and cant seem to find one.

Its all much easier than expected!
You dont have to change anything on your p2pool, you just change the miner's name:

Minername/sharedifficulty+localdifficulty

So for example:

Miner/1500+1

You set this on all miners. You can give different difficulty targets, but I dont see too much sense in this..

Miner01/1500+1
Miner02/1500+1
Miner03/1500+1

Maybe this is worth stating on the first page, and/or hinting in the help/wiki/etc?
Lets drop dat diff!

Ente

It would appear BAMT:s pools.conf dosent like this for some reason, propably the / or + sign is confusing it, could someone test and confirm ?

[echris1]

It depends on what you care about.

If you care about a single GPU always getting an even # of shares per hour then don't raise difficulty.  If all you care about it the TOTAL revenue per block then raising diff on all miners is the only thing that matters.

Honestly w/ 2.29 GH/s I am not sure I would raise difficulty and if I did it would be only a small amount.  Maybe to 1000?  Now 29 GH/s sure.

Ha, sometimes I forget I'm on the small end of things =)

Total revenue is all that matters of course, but I'll probably stick with default for now.

[DeathAndTaxes]

Ha, sometimes I forget I'm on the small end of things =)
Total revenue is all that matters of course, but I'll probably stick with default for now.

Another way to look at it is the avg miner on p2pool is ~1.5GH/s. 

1.5GH/s @ 750 diff ~= 2.3 GH/s @ 1150 dif.  If you went > 1150 you would have more variance than the avg miner.

I really think higher difficulty is most useful for miners 4GH/s+.

[Ente]

Sure, 1 share per hour is on the very low end. But with a block every 5 hours, it should already even out after a day?
(I am too lazy to write "on average after a long time" behind every number)

Anyway, with your numbers people might get a feeling for these things and decide for themselves what diff they feel comfortable with.

Remember avg can be deceptive.  1/6th to 6x the average isn't that rare.  So shooting for goal of 1 share per hour means you will have some hours with 6 shares and sometimes go 6 hours (oops missed an entire block = payout) without a single share.  That can be frustrating for some miners.

Uhm, even if you dont find a single share in one block, you get almost your full payout. Its the shares from last 24 hours which are counted. Maybe I just misunderstood your words ;-)

Ente

[Ente]

P2Pool release 0.10.3 signed tag: 0.10.3 UPGRADE REQUIRED before April 1st (BIP 16 date), March 27th for Litecoin

* New graph implementation (works on Windows!) with per-miner graphs. Go to http://127.0.0.1:9332/static/graphs.html once installed to view.

RELEASE NOTES: You have to go to http://127.0.0.1:9332/static/graphs.html , not http://127.0.0.1:9332/static to see graphs.

I am playing around with usernames (to change difficulty). Now the whole username (including the postfix /1000+1) is shown in the statistics, and whenever I change the postfix a new miner shows up in the stats.
WHen I delete the rrd stuff, the deprecated rrd graphs are cleaned.
How can I clean/reset the new graphs? I deleted most files in /data/bitcoin (or similar), no help. I had to reinstall p2pool, which starts the whole sharechain-load+verify again.

I would love to see the postfix ignored in the graphs altogether?

In the meantime, where is the data stored?

Ente

[DeathAndTaxes]

DUH your right.  Not sure what I was thinking.  Still with one block per hour your avg is only 24 which is likely too small of a sample to avoid significant swings due to variance.

[NothinG]

Anyone have a guess at when we might hit our next block?

[forrestv]

P2Pool release 0.10.3 signed tag: 0.10.3 UPGRADE REQUIRED before April 1st (BIP 16 date), March 27th for Litecoin

* New graph implementation (works on Windows!) with per-miner graphs. Go to http://127.0.0.1:9332/static/graphs.html once installed to view.

RELEASE NOTES: You have to go to http://127.0.0.1:9332/static/graphs.html , not http://127.0.0.1:9332/static to see graphs.

I am playing around with usernames (to change difficulty). Now the whole username (including the postfix /1000+1) is shown in the statistics, and whenever I change the postfix a new miner shows up in the stats.
WHen I delete the rrd stuff, the deprecated rrd graphs are cleaned.
How can I clean/reset the new graphs? I deleted most files in /data/bitcoin (or similar), no help. I had to reinstall p2pool, which starts the whole sharechain-load+verify again.

I would love to see the postfix ignored in the graphs altogether?

In the meantime, where is the data stored?

Ente

In the graph_db file. There isn't a way to just remove one username, but I'll change it to make it only consider the first part of the username.

[cabin]

DUH your right.  Not sure what I was thinking.  Still with one block per hour your avg is only 24 which is likely too small of a sample to avoid significant swings due to variance.

I believe each share is paid on average 3 times, so that helps smooth things out. Not today though!

[twmz]

Anyone have a guess at when we might hit our next block?

Right.......  NOW

[DeathAndTaxes]

Need clarification on how p2pool builds a block header.

Does this sound right?
1) p2ool uses RPC call "getmemorypool" to get block components from bitcoind
2) p2pool generates coinbase transaction from current sharechain
3) p2pool generates merkle tree (combining tx from bitcoind w/ coinbase from sharechain).
4) p2pool issues block headers to miners via getwork call (prior block, timestamp, merkle tree root, difficulty)

Right?  Wrong?  Clarifications?

[DeathAndTaxes]

Need clarification on how p2pool builds a block header.

Does this sound right?
1) p2ool uses RPC call "getmemorypool" to get block components from bitcoind
2) p2pool generates coinbase transaction from current sharechain
3) p2pool generates merkle tree (combining tx from bitcoind w/ coinbase from sharechain).
4) p2pool issues block headers to miners via getwork call (prior block, timestamp, merkle tree root, difficulty)

Right?  Wrong?  Clarifications?

Nobody?  

OK how about 1 BTC bounty
a) verify and/or correct the above execution flow above.
b) provide filename and line # of code (p2pool github) where p2pool pulls works from bitcoind.
c) provide fielname and line # of code (p2pool github) where p2pool modifies work from bitcoind to include coinbase reward split.