[1500 TH] p2pool: Decentralized, DoS-resistant, Hop-Proof pool - page 139.

p3yot33at3r

sr. member

Activity: 266

Merit: 250

Quote from: jtoomim on October 15, 2015, 01:52:08 PM

Quote from: p3yot33at3r on October 14, 2015, 05:20:11 PM

Interesting. Can anyone tell me what nicehash have done to improve the bitmain firmware?

This firmware was not developed by Nicehash. It was developed by smit1237. Nicehash is just distributing it.

There are a lot of changes. The cgminer version isn't the default broken Bitmain one, but instead a cgminer 4.8.0 that still submits stale shares. It might have a different default queue and expiry setting too, but I'm not sure about that. In my testing, it gives about 5% higher hashrate than the one S5s ship with on p2pool.

The smit1237 firmware also allows you to upload custom cgminer binaries to /config/, where they will persist across reboots and will be used instead of the cgminer in (IIRC) /usr/local/bin/.

OK, that sounds more like it. I might give it a go on one of my S5's & have a look-see what the queue & other settings are then, thanks for that info jtoomim

Edit: Just flashed it & queue is set at 8192 Tongue

Searching for the parameters now to change it............

Edit 1: OK, changed the queue setting to 1 for testing. BEWARE: This firmware does not submit stale shares - I noticed the "no-submit-stale" parameter in the /etc/init.d/cgminer.sh file (line 69), so I deleted it as it is my understanding that the default cgminer setting will submit stales - very important for p2pool......I'll monitor it & report back later.

jtoomim

hero member

Activity: 818

Merit: 1006

Quote from: p3yot33at3r on October 14, 2015, 05:20:11 PM

Interesting. Can anyone tell me what nicehash have done to improve the bitmain firmware?

This firmware was not developed by Nicehash. It was developed by smit1237. Nicehash is just distributing it.

There are a lot of changes. The cgminer version isn't the default broken Bitmain one, but instead a cgminer 4.8.0 that still submits stale shares. It might have a different default queue and expiry setting too, but I'm not sure about that. In my testing, it gives about 5% higher hashrate than the one S5s ship with on p2pool.

The smit1237 firmware also allows you to upload custom cgminer binaries to /config/, where they will persist across reboots and will be used instead of the cgminer in (IIRC) /usr/local/bin/.

p3yot33at3r

sr. member

Activity: 266

Merit: 250

Quote from: idonothave on October 15, 2015, 01:34:58 AM

Quote from: p3yot33at3r on October 14, 2015, 05:20:11 PM

Quote from: jtoomim on October 09, 2015, 04:40:48 PM

smit1237's firmware for S5s is here: https://www.nicehash.com/?p=software

Specifically, https://www.nicehash.com/sw/SD-S5-20150415_nicehash_perf_graphs.tar.gz

Interesting. Can anyone tell me what nicehash have done to improve the bitmain firmware?

imho nothing more then having a link on website with extranonce able firmware which created someone other

Right. Won't install that then...I'll stick with my old firmware & ck binary.

idonothave

full member

Activity: 238

Merit: 100

Quote from: p3yot33at3r on October 14, 2015, 05:20:11 PM

Quote from: jtoomim on October 09, 2015, 04:40:48 PM

smit1237's firmware for S5s is here: https://www.nicehash.com/?p=software

Specifically, https://www.nicehash.com/sw/SD-S5-20150415_nicehash_perf_graphs.tar.gz

Interesting. Can anyone tell me what nicehash have done to improve the bitmain firmware?

imho nothing more then having a link on website with extranonce able firmware which created someone other

p3yot33at3r

sr. member

Activity: 266

Merit: 250

Quote from: jtoomim on October 09, 2015, 04:40:48 PM

smit1237's firmware for S5s is here: https://www.nicehash.com/?p=software

Specifically, https://www.nicehash.com/sw/SD-S5-20150415_nicehash_perf_graphs.tar.gz

Interesting. Can anyone tell me what nicehash have done to improve the bitmain firmware?

Meuh6879

legendary

Activity: 1512

Merit: 1012

4 in less than 48 hours ...

THIS IS P2POOOOOOOOOOOOOOOOOOOOOOL !

windpath

legendary

Activity: 1258

Merit: 1027

Quote from: Richy_T on October 13, 2015, 09:54:41 AM

Haha. I guess a bunch of the people who left will be flooding back. Probably for another dry spell.

Has anyone done any calculation as to how much fickle miners benefit those of us who stick with it?

No calculation, but it seems when we hit a streak of bad luck folks stay for 1 more block, then leave just before it changes back to good luck...

Have the data to compair luck vs. # of miners, will run it sometime...

Richy_T

legendary

Activity: 2576

Merit: 2267

1RichyTrEwPYjZSeAYxeiFBNnKC9UjC5k

Haha. I guess a bunch of the people who left will be flooding back. Probably for another dry spell.

Has anyone done any calculation as to how much fickle miners benefit those of us who stick with it?

Meuh6879

legendary

Activity: 1512

Merit: 1012

and 3 ...

p3yot33at3r

sr. member

Activity: 266

Merit: 250

Quote from: p3yot33at3r on October 12, 2015, 06:52:21 AM

Nice to wake up to a block Grin

..make that 2 Grin

idonothave

full member

Activity: 238

Merit: 100

Quote from: p3yot33at3r on October 12, 2015, 06:52:21 AM

Quote from: jtoomim on October 12, 2015, 05:38:48 AM

Quote from: idonothave on October 12, 2015, 05:27:58 AM

someone should hold leadership with this to say, guys, to put this and this is the best way for now... and even why not to vote for major changes and p2pool bitcoin.conf best template form
you as a provider of quite big part of p2pool hashrate would be that

I disagree. That would be centralizing p2pool. We all have to make our own decisions about how to configure and use this system.

Exactly, plus no two setups will be the same - settings that work fine for one, might not work well for another.

@ forrestv - if you're lurking - how is your testing of jtoomim's performance going?

Nice to wake up to a block Grin

disability to cooperate is not an advantage. we are here to learn

p3yot33at3r

sr. member

Activity: 266

Merit: 250

Quote from: jtoomim on October 12, 2015, 05:38:48 AM

Quote from: idonothave on October 12, 2015, 05:27:58 AM

someone should hold leadership with this to say, guys, to put this and this is the best way for now... and even why not to vote for major changes and p2pool bitcoin.conf best template form
you as a provider of quite big part of p2pool hashrate would be that

I disagree. That would be centralizing p2pool. We all have to make our own decisions about how to configure and use this system.

Exactly, plus no two setups will be the same - settings that work fine for one, might not work well for another.

@ forrestv - if you're lurking - how is your testing of jtoomim's performance going?

Nice to wake up to a block Grin

jtoomim

hero member

Activity: 818

Merit: 1006

Quote from: idonothave on October 12, 2015, 05:27:58 AM

someone should hold leadership with this to say, guys, to put this and this is the best way for now... and even why not to vote for major changes and p2pool bitcoin.conf best template form
you as a provider of quite big part of p2pool hashrate would be that

I disagree. That would be centralizing p2pool. We all have to make our own decisions about how to configure and use this system.

idonothave

full member

Activity: 238

Merit: 100

Quote from: jtoomim on October 12, 2015, 04:55:55 AM

I suspect the performance issue is due to the p2pool's known_txns_var (the equivalent of the p2pool mempool) getting large over time, and pypy taking a lot more space to store this for some reason. In any case, the problems that I've seen appear to only manifest themselves after running p2pool for a while, like a few days. Restarting p2pool clears the caches and reduces memory usage, improving performance.

When a node creates a share that includes large low-fee transactions, it forces all other p2pool nodes to download and store those transactions. Setting minrelaytxfee to something reasonably high like 0.0005 and blockprioritysize to something low like 0 may help performance for all p2pool nodes, because your node would force fewer low-fee spam txns onto those nodes.

someone should hold leadership with this to say, guys, to put this and this is the best way for now... and even why not to vote for major changes and p2pool bitcoin.conf best template form
you as a provider of quite big part of p2pool hashrate would be that

jtoomim

hero member

Activity: 818

Merit: 1006

I suspect the performance issue is due to the p2pool's known_txns_var (the equivalent of the p2pool mempool) getting large over time, and pypy taking a lot more space to store this for some reason. In any case, the problems that I've seen appear to only manifest themselves after running p2pool for a while, like a few days. Restarting p2pool clears the caches and reduces memory usage, improving performance.

When a node creates a share that includes large low-fee transactions, it forces all other p2pool nodes to download and store those transactions. Setting minrelaytxfee to something reasonably high like 0.0005 and blockprioritysize to something low like 0 may help performance for all p2pool nodes, because your node would force fewer low-fee spam txns onto those nodes.

idonothave

full member

Activity: 238

Merit: 100

Quote from: p3yot33at3r on October 11, 2015, 11:39:22 AM

Quote from: jtoomim on October 11, 2015, 10:55:47 AM

I have been seeing a lot of performance problems with pypy recently. One of my nodes exceeded 4 GB of RAM usage by pypy/p2pool alone. I recommend using regular python, as it seems to give more reliable performance.

I was wondering myself, but thought it was just me..... Tongue

I have also noticed that and temporarily got back to python but in my opinion the problem was in combination pypy / high GBT latency so I have added to my bitcoin.conf lines

Code:

blockmaxsize=250000 # default: 250000
blockprioritysize=27000 # default: 27000
mintxfee=0.0002 # default: 0.0001
minrelaytxfee=0.0002 # default: 0.0001

then latency went down around 0.6 and I switched back to pypy without any problems again but I think it is not permanent solution and if there is anybody experienced with bitcoin.conf tuning I would welcome to have one recommended template of bitcoin.conf usable here

p3yot33at3r

sr. member

Activity: 266

Merit: 250

Quote from: jtoomim on October 11, 2015, 10:55:47 AM

I have been seeing a lot of performance problems with pypy recently. One of my nodes exceeded 4 GB of RAM usage by pypy/p2pool alone. I recommend using regular python, as it seems to give more reliable performance.

I was wondering myself, but thought it was just me..... Tongue

jtoomim

hero member

Activity: 818

Merit: 1006

I have been seeing a lot of performance problems with pypy recently. One of my nodes exceeded 4 GB of RAM usage by pypy/p2pool alone. I recommend using regular python, as it seems to give more reliable performance.

jtoomim

hero member

Activity: 818

Merit: 1006

Quote from: Richy_T on October 09, 2015, 03:17:34 PM

If that's the one in the link, that was for the S4 (though they may be compatible?). I'll look into it a bit more. I may want to add some extra stuff in in any case.

smit1237's firmware for S5s is here: https://www.nicehash.com/?p=software

Specifically, https://www.nicehash.com/sw/SD-S5-20150415_nicehash_perf_graphs.tar.gz

Richy_T

legendary

Activity: 2576

Merit: 2267

1RichyTrEwPYjZSeAYxeiFBNnKC9UjC5k

Quote from: jtoomim on October 09, 2015, 02:34:50 PM

I think you can also just use the smit1237 firmware. If I remember correctly, if you want to use a custom cgminer with smit1237's FW, you just stick the new cgminer into /config (which is non-volatile) and that becomes the new default.

If that's the one in the link, that was for the S4 (though they may be compatible?). I'll look into it a bit more. I may want to add some extra stuff in in any case.

Topic: [1500 TH] p2pool: Decentralized, DoS-resistant, Hop-Proof pool - page 139. (Read 2591920 times)