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

[centove]

Looks okay here:
http://ask.gxsnmp.org:9332/static/graphs.html?Week

not here though:

http://nogleg.com:9332/static/graphs.html?Week

cool Bitcoind times.

which bitcoind version ?
I got only an average from 0.10 sec

Don't know if this was directed at me or not...

gregm@db:~$ bitcoind getinfo
Botcoind
{
    "version" : 80100,
    "protocolversion" : 70001,
    "walletversion" : 60000,
...
}

However.... It's kinda a beefy machine....

processor   : 15
vendor_id   : GenuineIntel
cpu family   : 6
model      : 26
model name   : Intel(R) Xeon(R) CPU           L5520  @ 2.27GHz

and 32 Gigs of memory....

[Smoovious]

A 2600 share will count for twice as much as a 1300 share, the only difference would be that it'd increase deviation, but should average out eventually.

That doesn't seem quite right. I just put in a 21,700 share, and it seems to be worth just as much as any other share I've put in.

The amount the share is worth is based on the target diff you were aiming for, not what diff the share ended up being.

So if you were targetting diff 1,000 shares, and you got 2 shares, at 1,074 and 7,771 then both of those shares would be worth a diff 1,000 share.

If you were targetting diff 2,000 shares, then the 1,074 wouldn't have counted as a share, you would have still got a share for the 7,771 one, which would be worth a diff 2,000 share, which should give you the same share of the payout.

(my example is probably over-simplified, but you get the idea)

-- Smoov

[Searinox]

Running fine for 3 1/2 hours but memory usage graph still empty.

[KALRONG]

That one is just 50Mh/s as Im testing to see if I get everything up and running before switching the whole rig (I know is just a few minutes but dont want to waste even a second hehe), so what do you mean that the shares submitted by cgminers are not the same as shares accepted by p2pool? shouldn't I receive a rejected message or see something on p2pool?

Thanks for your reply

bitpop:

How the hell? bitcoind as server and miner on it

yes, it uses accepted shares to calculate the hash rate it shows on the graph

so if you submit a 2 difficulty share, then you should see some hash rate on the chart

but you wont 'earn' anything out of the p2pool pot.  for that you have to hit the rolling difficulty, which attempts to be as close to 6 new works per (ed: minute) as possible

we could also really use a code change to be able to set the custom difficulty to at least 5000 , instead of the 2000 it is currently.  i know you can go in and modify the source yourself, but most people wouldnt/cant do that

Thanks for the nice explanation

[Mogsington]

The amount the share is worth is based on the target diff you were aiming for, not what diff the share ended up being.

Brilliant! Got it now thanks!
Uhm.
So what does the +d bit do at the end?
(No really! I'd like to be fairly sure I knew!)

[zvs]

Looks okay here:
http://ask.gxsnmp.org:9332/static/graphs.html?Week

not here though:

http://nogleg.com:9332/static/graphs.html?Week

cool Bitcoind times.

which bitcoind version ?
I got only an average from 0.10 sec

& gyverlb


i update once every day or two, to the latest git version.  though, the main thing as far as the getblocktemplate latency goes is how many transactions you have stored, i.e. bitcoind getmininginfo.  the more that builds up, the slower it goes.   mine is at 360 right now but clears out after blocks..... because I "cheated" a bit and set my relay fee to 50000 (the same as the creation fee, in everything *BUT* 8.1.99)... once the release is made public w/ the 10000 relay & creation fee, then i won't mess w/ it anymore... but as long as all the official releases have the creation fee at 50000, i don't see anything wrong with setting my relay fee to 50000 as well.  that cuts out a lot of the transactions that end up clogging your pooledtx.  

also, max blocksize has a lot to do with it.  i'm at 5k max blocksize, if I went up to 50k, I'd be getting 2 or 3s latency..

It's all on a ramdrive, too... but that has less to do with it then you think, probably.... example (and i had nothing to do w/ his bitcoin address):

http://rehtid.com:9332/static/graphs.html?Week

i sent him a custom copy of bitcoind on tues, now he runs around a constant 3.5-4ms latency.  this is on some cheap $40 OVH machine.  8GB RAM and some semi-junk CPU.   this was accomplished essentially by just raising the tx fee to 100000.  at that rate, you'll only get tx'es that'll clear out on the next block.   he's also set to 5000 max blocksize.   though, you'd start to notice the ramdrive a lot more once you start raising the blocksize.  the problem I have with it is that it also increases the # of orphans..

i think it'd be much better if the block solver got all the TX fees instead of however the reward system works now (and, no, this isn't favoring someone that's 60ghash, because like the share difficulty, in the long run it would all even out).

it does bother me when people have the maxblocksize set to <1000, though.... because occasionally you will get that transaction with the 5 BTC fee and it only takes a few KB to get it in a block...  lenny solved one of those just about a week or two ago (block was worth 33, I believe)

in any case, it isn't a huge deal right now since ~7 blocks are being solved every hour.. but I don't like limiting my blocksize due to the orphans...  there needs to be the counterbalance (re: getting the tx fees).

with my avg 6-7ghash, i've solved 3 p2p blocks in the last 3-4 weeks (yes, amazing luck (ed: also solved 1 block on bitminter, for about 1.5BTC worth of mining... and got an 8M difficulty share earlier today, lol))...   but everyone would distribute my normal reward and the blocks would be .3 or .4 btc more if i got to keep all the transaction fees =p

[gyverlb]

Looks okay here:
http://ask.gxsnmp.org:9332/static/graphs.html?Week

not here though:

http://nogleg.com:9332/static/graphs.html?Week

cool Bitcoind times.

which bitcoind version ?
I got only an average from 0.10 sec

& gyverlb

Sorry it seems I mixed up your graphs with centove's graphs, painkillers seems to have a measurable effect on my ability to read...

[zvs]

Looks okay here:
http://ask.gxsnmp.org:9332/static/graphs.html?Week

not here though:

http://nogleg.com:9332/static/graphs.html?Week

cool Bitcoind times.

which bitcoind version ?
I got only an average from 0.10 sec

& gyverlb

Sorry it seems I mixed up your graphs with centove's graphs, painkillers seems to have a measurable effect on my ability to read...

I just thought you were being sarcastic, w/ the quoted Subo msg 

[mdude77]

i think it'd be much better if the block solver got all the TX fees instead of however the reward system works now (and, no, this isn't favoring someone that's 60ghash, because like the share difficulty, in the long run it would all even out).

I'd like to see the reward system change as well.  The current one doesn't give much incentive.  I like your idea, all the TX fees go to the solver + his portion of the block.

M

[Mogumodz]

Is there an easy way to keep graph stats when upgrading/using to a newer version ?

I'm using the windows binary.

[Rassah]

i think it'd be much better if the block solver got all the TX fees instead of however the reward system works now (and, no, this isn't favoring someone that's 60ghash, because like the share difficulty, in the long run it would all even out).

Sounds like something someone with a 60ghash miner would say 

[iongchun]

Is there an easy way to keep graph stats when upgrading/using to a newer version ?

I'm using the windows binary.

Just copy the "data" folder to the directory of a newer version.
Or you can keep the "data" folder somewhere outside the p2pool directory,
and add a "--datadir" parameter to your batch file, for example:

Code:

c:/p2pool/p2pool_win32_11.4/run_p2pool.exe --datadir c:/p2pool/data  ...

[Mogumodz]

Is there an easy way to keep graph stats when upgrading/using to a newer version ?

I'm using the windows binary.

Just copy the "data" folder to the directory of a newer version.
Or you can keep the "data" folder somewhere outside the p2pool directory,
and add a "--datadir" parameter to your batch file, for example:

Code:

c:/p2pool/p2pool_win32_11.4/run_p2pool.exe --datadir c:/p2pool/data  ...

Thanks a lot.

[daemondazz]

How is the hash rate on the network calculated? I have two miners which combined are averating 1.08Gh/s but p2pool.info only shows ~800MH/s ?

Edit: the 1.08GH/s is from the mean value on my nodes daily graph, not from the miners themselves.

[Smoovious]

i think it'd be much better if the block solver got all the TX fees instead of however the reward system works now (and, no, this isn't favoring someone that's 60ghash, because like the share difficulty, in the long run it would all even out).

Sounds like something someone with a 60ghash miner would say 

Well, unless it has changed, the block solver already gets a bonus, so if the solver is going to get all the TX fees, then the other bonus shouldn't be done anymore.

Pick one or the other, not both.

-- Smoov

[zvs]

i think it'd be much better if the block solver got all the TX fees instead of however the reward system works now (and, no, this isn't favoring someone that's 60ghash, because like the share difficulty, in the long run it would all even out).

Sounds like something someone with a 60ghash miner would say  

    except I'm at 5-7ghash, not 60.

Well, unless it has changed, the block solver already gets a bonus, so if the solver is going to get all the TX fees, then the other bonus shouldn't be done anymore.

Pick one or the other, not both.

-- Smoov

Yeah, so there's really no reason not to just change it over to the TX'es.     Add them to your blocks, take the risk of having a few more orphans, but get the whole TX amt if you solve it

[kano]

...
Yeah, so there's really no reason not to just change it over to the TX'es.     Add them to your blocks, take the risk of having a few more orphans, but get the whole TX amt if you solve it

Except that is the exact opposite of the BTC design.
BTC design is to halve every 4 years coz transaction fees will 'supposedly' cover this over time.
So doing it that way means to head in the direction of giving all the reward to the block finder ... i.e. the opposite of being a pool.

[iongchun]

Well, unless it has changed, the block solver already gets a bonus, so if the solver is going to get all the TX fees, then the other bonus shouldn't be done anymore.

Pick one or the other, not both.

-- Smoov

Yeah, so there's really no reason not to just change it over to the TX'es.     Add them to your blocks, take the risk of having a few more orphans, but get the whole TX amt if you solve it

All p2pool nodes help propagating the transactions contributed to the blocks you mine,
so I guess they all deserve a portion of fees
And it should not be measured in hashrates...

[maqifrnswa]

The amount the share is worth is based on the target diff you were aiming for, not what diff the share ended up being.

Brilliant! Got it now thanks!
Uhm.
So what does the +d bit do at the end?
(No really! I'd like to be fairly sure I knew!)

+d changes the difficulty of work that is served to your miner. e.g., if you only want your miner to return 16 difficulty shares to your p2pool node, add +16 to user name. If you want your miner to tell the p2pool node to only send 5000 difficulty blocks back to the p2pool network, then add /5000 to user name. If you want both, append /5000+16 to the name.

[davidkassa]

...
Yeah, so there's really no reason not to just change it over to the TX'es.     Add them to your blocks, take the risk of having a few more orphans, but get the whole TX amt if you solve it

Except that is the exact opposite of the BTC design.
BTC design is to halve every 4 years coz transaction fees will 'supposedly' cover this over time.
So doing it that way means to head in the direction of giving all the reward to the block finder ... i.e. the opposite of being a pool.

this