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

[mahrens917]

I just created a new p2pool node scanner at nodes.p2pool.co.  The other ones were either too buggy and even at the times they were not I still could never figure out which p2pool to mine.  How much does latency matter over "efficiency if no miners"?  

So I created the P2Pool.co Node Scanner.  I think it is the best P2Pool node scanner out there.

It includes a scoring algorithm to highlight the best p2pool near your location, and it factors every datapoint and combines it into a p2pool score.

The latency numbers are much more realistic than other node scanners (compare our latency numbers to actual pings that you run yourself and compare them to the other p2pool node scanners).

If you would like us to add your node or if you have any ideas to make this scanner better than PM me.

I would also like your feedback on the scoring algorithm.  At it stands now this is how the scoring works:

• The score starts at 100
• 16 * the fee is subtracted (fees are bad)
• 4 * the donation fee is subtracted (donations are less bad)
• 16 * the latency (ms) divided by 100 is subtracted (latency caries much weight with p2pool)
• 8 * the getwork latency (ms) divided by 100 is subtracted (GWL is important as well)
• Efficiency over 100 divided by 2 is added (under 100 is subtracted) (efficiency over 100% is good and less than 100% is bad)
• Same for Efficiency If Miner Perfect (Nodes should not be punished if they just have a bad miner)
• For 0 or 1 users, 8 is subtracted from the score (Low user p2pools have efficiencies rate that are unrealistically high)
• Hash rate (TH/s) is divided by four and than added (people will put up more hashes the more they like the pool)
• Uptime over 2 days adds a point and 0 days subtracts a point
• An unknown version of the p2pool software subtracts 10 points
• A warning (like bitcoind not running) gives a score of 0

[1713574924]

1713574924

[windpath]

I just created a new p2pool node scanner.  The other ones were either too buggy and even at the times they were not I still could never figure out which p2pool to mine.  How much does latency matter over "efficiency if no miners"? 

So I created the P2Pool.co Node Scanner.  I think it is the best P2Pool node scanner out there.

...

Where can we access it?

[Sempruls]

I just created a new p2pool node scanner.  The other ones were either too buggy and even at the times they were not I still could never figure out which p2pool to mine.  How much does latency matter over "efficiency if no miners"? 

So I created the P2Pool.co Node Scanner.  I think it is the best P2Pool node scanner out there.

...

Where can we access it?

yeah, i try to find it too but cant locate it, need info

[mahrens917]

Wow, sorry about that.  You can find it at nodes.p2pool.co

[IYFTech]

Wow, sorry about that.  You can find it at nodes.p2pool.co

Well look at that - top UK p2pool node........

I like your scoring method 

[windpath]

Wow, sorry about that.  You can find it at nodes.p2pool.co

Looks really good.

Have not evaluated the scoring, but have three suggestions that would make this the go-to resource for choosing a p2pool node:

1. Try to include all active nodes automatically - perhaps by looking at each discovered nodes complete peer list for missing nodes?

2. Latency scoring is basically irrelevant unless you are running from the visitors computer (or close to it) to the actual node, when I looked into this a while ago the most suitable solution I found was https://www.ookla.com/ Netguage, however the price of the service was prohibitive for me, perhaps you will come up with a better solution.

3. If you solve the latency and missing node issue, a valuable service to node operators would be to provide an export of the 10 best/closest online nodes to add to the command line when starting p2pool.

Great work!

[mahrens917]

Wow, sorry about that.  You can find it at nodes.p2pool.co

Looks really good.

Have not evaluated the scoring, but have three suggestions that would make this the go-to resource for choosing a p2pool node:

1. Try to include all active nodes automatically - perhaps by looking at each discovered nodes complete peer list for missing nodes?

2. Latency scoring is basically irrelevant unless you are running from the visitors computer (or close to it) to the actual node, when I looked into this a while ago the most suitable solution I found was https://www.ookla.com/ Netguage, however the price of the service was prohibitive for me, perhaps you will come up with a better solution.

3. If you solve the latency and missing node issue, a valuable service to node operators would be to provide an export of the 10 best/closest online nodes to add to the command line when starting p2pool.

Great work!

Thanks for the suggestions and glad you like nodes.p2pool.co.

1. Yes, automating collecting the node data will be in phase 2.  Getting this version out took me longer than I would have liked though so I could not get everything out all at once.

2. nodes.p2pool.co IS running from the visitors computer.  Other node finders grab some data from the node and calculate the time to get that data.  Unless you run that test many times it is a highly variable and poor latency test.  What nodes.p2pool.co does is it calculates the physical distance between your computer and each p2pool node.  (I use maxmind.com to locate the p2pool server from its IP address.)  From there it computes latency which is pretty accurate.  It calculates the speed of light between the two distances on the globe and multiples it by three since no connection is a straight line and it needs to travel through routers and since fiber slows the signal down a bit.  It also adds 20ms for the computers to process the signals.  Try it, it should be a pretty accurate method.  With all that said I will still check out Ookla - it is probably even more accurate.

3. Yes, I was thinking the same.  However, does p2pool actually figure this out on its own over time?  I actually don't know but looking at the logs I know it drops unresponsive peers and pulls in new ones.

[IYFTech]

Whoever is mining on my node using address 1PuEtn8MzGbF58r1KLXwAnyvtq5g2FLZsw, your DOA rate is rather high. PM me & we'll see if we can improve it 

[mdude77]

The number 1 issue is fixed with latest firmware, which runs cgminer 4.6.1.  When the work restart request comes in, I usually see the flush work within 1 second, then the new block task usually arrives in another second.  Often there are new accepted shares submitted within 2 seconds of the work restart request.  The new cgminer 4.6.1 has brought new life to the S2.  I used to point the S2's to a different pool, while all my other miners used P2Pool.  It's so nice to see them doing the full 1000 GH/s with P2Pool.  I have 2x S2's and 2x C1's, and all 4 of them run neck-and-neck.

Are you sure this is fixed?  I upgraded my S2 to the newest firmware, and yes I get 1 TH/s now, but I still get 10-15% rejects because of the 30 second work restarts that the Ant can't respond to quick enough.

M

[PatMan]

The number 1 issue is fixed with latest firmware, which runs cgminer 4.6.1.  When the work restart request comes in, I usually see the flush work within 1 second, then the new block task usually arrives in another second.  Often there are new accepted shares submitted within 2 seconds of the work restart request.  The new cgminer 4.6.1 has brought new life to the S2.  I used to point the S2's to a different pool, while all my other miners used P2Pool.  It's so nice to see them doing the full 1000 GH/s with P2Pool.  I have 2x S2's and 2x C1's, and all 4 of them run neck-and-neck.

Are you sure this is fixed?  I upgraded my S2 to the newest firmware, and yes I get 1 TH/s now, but I still get 10-15% rejects because of the 30 second work restarts that the Ant can't respond to quick enough.

M

I'm getting ~3% using the -queue 0 setting mdude - have you tried that?

[mdude77]

The number 1 issue is fixed with latest firmware, which runs cgminer 4.6.1.  When the work restart request comes in, I usually see the flush work within 1 second, then the new block task usually arrives in another second.  Often there are new accepted shares submitted within 2 seconds of the work restart request.  The new cgminer 4.6.1 has brought new life to the S2.  I used to point the S2's to a different pool, while all my other miners used P2Pool.  It's so nice to see them doing the full 1000 GH/s with P2Pool.  I have 2x S2's and 2x C1's, and all 4 of them run neck-and-neck.

Are you sure this is fixed?  I upgraded my S2 to the newest firmware, and yes I get 1 TH/s now, but I still get 10-15% rejects because of the 30 second work restarts that the Ant can't respond to quick enough.

M

I'm getting ~3% using the -queue 0 setting mdude - have you tried that?

It's leveled out to 6.7%.  I haven't adjusted the queue yet.  What file do I need to change for that?  I'm still a linux n00b.

M

[PatMan]

vi /etc/init.d/cgminer.sh

Press “i” (enter edit mode)
Scroll down to line 60 in the code, starting with :

PARAMS="--bitmain-dev....

change the queue parameters at the end of that string to:

--queue 0

press "esc" (exit edit mode), then type:

:wq      & hit enter.

Then click save & apply in the web UI miner configuration tab  

EDIT: You'll have to do this after every reboot unfortunately, as it will reset to defaults. Just don't reboot it......

[mdude77]

vi /etc/init.d/cgminer.sh

Press “i” (enter edit mode)
Scroll down to line 60 in the code, starting with :

PARAMS="--bitmain-dev....

change the queue parameters at the end of that string to:

--queue 0

press "esc" (exit edit mode), then type:

:wq      & hit enter.

Then click save & apply in the web UI miner configuration tab  

EDIT: You'll have to do this after every reboot unfortunately, as it will reset to defaults. Just don't reboot it......

tyvm.  Do you set your worker difficulty, or just live with the p2pool varying one?

So far it setting the queue to 0 hasn't made a smidgen of difference on the rejects.  Trying 1 now.

M

[mdude77]

EDIT: You'll have to do this after every reboot unfortunately, as it will reset to defaults. Just don't reboot it......

Seriously??  Wow.

Isn't this easier?

Code:

/usr/bin/cgminer-api "setconfig|queue,0"

M

[PatMan]

EDIT: You'll have to do this after every reboot unfortunately, as it will reset to defaults. Just don't reboot it......

Seriously??  Wow.

Isn't this easier?

Code:

/usr/bin/cgminer-api "setconfig|queue,0"

M

Dunno, never tried it.....if it works - let me know 

[mdude77]

EDIT: You'll have to do this after every reboot unfortunately, as it will reset to defaults. Just don't reboot it......

Seriously??  Wow.

Isn't this easier?

Code:

/usr/bin/cgminer-api "setconfig|queue,0"

M

Dunno, never tried it.....if it works - let me know 

It works on S2s.  Haven't figured out how to get privileged API commands to work on S3s yet.

M

[mahrens917]

I have introduced the concept of "Miner Power" on my p2pools located at us-east.p2pool.co and europe.p2pool.co.  "Miner Power" is the effective rate of a miner on a particular node compared to other p2pool nodes. It is calculated by taking a miner's DOA % rate and comparing it the DOA rate of all miners on the global p2pool network. It is then multiplied by the efficiency of that particular p2pool node. Is it correct that that is the rate of increase of payout of the miner on that p2pool node compared to other p2pool nodes?  If so shouldn't "efficiency if miner perfect" be used over just efficiency (where other miners can take the efficiency rate down)?

For those interested in the code:

Code:

doa = local_stats.miner_dead_hash_rates[address] || 0;
doa_prop = (parseFloat(doa) / parseFloat(hashrate)) * 100;
doa_global = (global_stats.pool_hash_rate - global_stats.pool_nonstale_hash_rate) / global_stats.pool_hash_rate * 100;
miner_power = 100 * ((100 - doa_prop) / (100 - doa_global)) * local_stats.efficiency;

[naplam]

I think your formulas are a bit arbitrary. For instance, getwork latency is given way too much importance in the pool formula. And everything is linear.. and so on (latencies are mostly ok  within a range before the problems they cause skyrocket).

[MissouriMiner]

EDIT: You'll have to do this after every reboot unfortunately, as it will reset to defaults. Just don't reboot it......

Seriously??  Wow.

Isn't this easier?

Code:

/usr/bin/cgminer-api "setconfig|queue,0"

M

Dunno, never tried it.....if it works - let me know  

It works on S2s.  Haven't figured out how to get privileged API commands to work on S3s yet.

M

The S3's are even easier because it will save over a power cycle, unlike the S2/S4/C1.   Just change the PARAMS variable in /etc/init.d/cgminer.  That is where you add --api-allow and whatever other cgminer params you want.

Like so:

PARAMS="$AOPTIONS $POOL1 $POOL2 $POOL3 $_pb $_ow $_bec --api-listen --api-allow W:0/0 --bitmain-checkn2diff --bitmain-hwerror --queue 0"

[phillipsjk]

P2Pool looks busy lately. I noticed the P2Pool stale rate is now at 20%

Code:

2014-11-28 13:29:18.898450 P2Pool: 17324 shares in chain (17328 verified/17328 total) Peers: 22 (16 incoming)
2014-11-28 13:29:18.898605  Local: 0H/s in last 0.0 seconds Local dead on arrival: ??? Expected time to share: ???
2014-11-28 13:29:18.898680  Shares: 0 (0 orphan, 0 dead) Stale rate: ??? Efficiency: ??? Current payout: 0.0000 BTC
2014-11-28 13:29:18.898768  Pool: 2548TH/s Stale rate: 21.6% Expected time to block: 18.9 hours

Nothing to panic about, as with a 30 second share chain interval, that works out to a block propagation lag of only 6 seconds.

My node is using ~40% of one CPU core for P2Pool at the moment (need a new machine with moar CPU and disk space).