bitHopper: Python Pool Hopper Proxy - page 90.

ed64

newbie

Activity: 42

Merit: 0

Quote from: Clipse on August 03, 2011, 11:32:54 PM

re: penalty option, this is where the dynamic penalty option of ryo worked so well.

Ha, I guess I don't trust anything I don't understand... I found for non-prop pools or delayed pools, dynamic penalty would push those quite low and I felt like I was mining unprofitable pools quite often. Oh well, in anycase, hopefully I'll have time to port it over, but not sure how well I can test it.

I think I'm secretly hoping Ryouiki or someone else will take the plunge, I just set the stage by extracting out most of the "scheduling" code Wink

djex

full member

Activity: 196

Merit: 100

Ed, does your SliceScheduler option work? Or is it only AltSliceScheduler that works for now? If not which is best to use?

I'm currently using altslice at 900, does that sound about right?

Clipse

hero member

Activity: 504

Merit: 502

Quote from: ed64 on August 03, 2011, 11:28:57 PM

Quote from: Clipse on August 03, 2011, 11:21:46 PM

Ed, I see you defined mine_slush at 0.4*difficulty, is that correct afaik it should be something between 0.10 - 0.25 if I remember correctly.

I heard something like 15%? I think with the *4 factor (copied from the default scheduler), it ends up around 11%, since it's 1/4 of the default threshold of 0.435*difficulty.

Will add in "penalty" option to user.cfg and then people can muck as they please...

Ok reason Im asking, I am using your fork atm with the altslice 1200 etc. etc. Wink

and im noticing something that I am not sure is the right approach unless its a bug.
pools mining to higher than 43.5% , mine_slush went to 38% ? not sure why this is happening but right now mtred is going still and its at 45%, is this the way you setup your fork?

EDIT: ok I notice it is finishing the timeslice allocated(even if pool is above 43.5%) before moving onto other lower percentage servers, just finding out if this is the intended setup by design with altslicing.

re: penalty option, this is where the dynamic penalty option of ryo worked so well.

ed64

newbie

Activity: 42

Merit: 0

Quote from: Clipse on August 03, 2011, 11:21:46 PM

Ed, I see you defined mine_slush at 0.4*difficulty, is that correct afaik it should be something between 0.10 - 0.25 if I remember correctly.

I heard something like 15%? I think with the *4 factor (copied from the default scheduler), it ends up around 11%, since it's 1/4 of the default threshold of 0.435*difficulty.

Will add in "penalty" option to user.cfg and then people can muck as they please...

ed64

newbie

Activity: 42

Merit: 0

Could be sheer luck? At 1 hour delay, wouldn't they have found a block by then (on average)? Almost like jumping in blind...??

Ryo has a simulation where time slicing pays off:
https://github.com/ryouiki/bitHopper/pull/3#issuecomment-1699605

Alas, that's the whole point of the scheduler class, to be able to add in as many as we like and to each their own!

Quote from: Clipse on August 03, 2011, 11:12:43 PM

Quote from: joulesbeef on August 03, 2011, 11:01:12 PM

is it really worth hopping btcguild? with the hour delay?.. how are you working out when to hop? are you just saying like 23% takes an hr and adjusting what you think the shares should be in reality bashed on their total mh?

With ryo's fork using dynamic penality, deepbit(backup) and btcguild(based on poolspeed it actually worked out positive jumping in even with the hour delayed stats). Im not going to discuss the efficiency level before someone spams btcg owner with this post. Lets just say its +EV so far.

joulesbeef

sr. member

Activity: 476

Merit: 250

moOo

I have problems with them too.. they are a new pool, they have only found 2 blocks, last one took over 400 hours. I think they just have issues.

ed64

newbie

Activity: 42

Merit: 0

Ugh, who likes the new christmas tree colour scheme?

Time to use some complimentary colour schemes:
http://www.colorschemer.com/online.html
http://kuler.adobe.com/

The graphs look better now though... not sure how useful they are...

Clipse

hero member

Activity: 504

Merit: 502

Quote from: simonk83 on August 03, 2011, 11:17:42 PM

Quote from: simonk83 on August 02, 2011, 11:12:35 PM

EDIT2: Ok, found the culprit (I assume):

2011-08-03 13:22:40: Listener for "Hopper2": [03/08/2011 13:22:40] Connected to server
2011-08-03 13:27:06: Listener for "Hopper2": [03/08/2011 13:27:06] Warning: work queue empty, miner is idle
2011-08-03 13:27:10: Listener for "Hopper": [03/08/2011 13:27:10] Warning: work queue empty, miner is idle
2011-08-03 13:34:17: Listener for "Hopper": [03/08/2011 13:34:17] Warning: work queue empty, miner is idle
2011-08-03 13:34:22: Listener for "Hopper2": [03/08/2011 13:34:22] Warning: work queue empty, miner is idle
2011-08-03 13:34:59: Listener for "Hopper": [03/08/2011 13:34:59] LP: New work pushed
2011-08-03 13:34:59: Listener for "Hopper2": [03/08/2011 13:34:59] LP: New work pushed
2011-08-03 13:35:14: Listener for "Hopper": [03/08/2011 13:35:14] Warning: work queue empty, miner is idle

Right, with these dropouts I've been having, it's definitely a problem with bloody's. It only ever happens on that server. Not sure why it only happens to me and not others, but there you go. I'm in Australia so I guess distance is a factor, but then it would be with the majority of the other pools as well and they're fine

Ive been having issues with bloody aswell lately before that it had no issues however miner idles I received atleast ~2 an hour no matter what server.

Clipse

hero member

Activity: 504

Merit: 502

Ed, I see you defined mine_slush at 0.4*difficulty, is that correct afaik it should be something between 0.10 - 0.25 if I remember correctly.

simonk83

hero member

Activity: 798

Merit: 1000

Quote from: simonk83 on August 02, 2011, 11:12:35 PM

EDIT2: Ok, found the culprit (I assume):

2011-08-03 13:22:40: Listener for "Hopper2": [03/08/2011 13:22:40] Connected to server
2011-08-03 13:27:06: Listener for "Hopper2": [03/08/2011 13:27:06] Warning: work queue empty, miner is idle
2011-08-03 13:27:10: Listener for "Hopper": [03/08/2011 13:27:10] Warning: work queue empty, miner is idle
2011-08-03 13:34:17: Listener for "Hopper": [03/08/2011 13:34:17] Warning: work queue empty, miner is idle
2011-08-03 13:34:22: Listener for "Hopper2": [03/08/2011 13:34:22] Warning: work queue empty, miner is idle
2011-08-03 13:34:59: Listener for "Hopper": [03/08/2011 13:34:59] LP: New work pushed
2011-08-03 13:34:59: Listener for "Hopper2": [03/08/2011 13:34:59] LP: New work pushed
2011-08-03 13:35:14: Listener for "Hopper": [03/08/2011 13:35:14] Warning: work queue empty, miner is idle

Right, with these dropouts I've been having, it's definitely a problem with bloody's. It only ever happens on that server. Not sure why it only happens to me and not others, but there you go. I'm in Australia so I guess distance is a factor, but then it would be with the majority of the other pools as well and they're fine

Clipse

hero member

Activity: 504

Merit: 502

Quote from: joulesbeef on August 03, 2011, 11:01:12 PM

is it really worth hopping btcguild? with the hour delay?.. how are you working out when to hop? are you just saying like 23% takes an hr and adjusting what you think the shares should be in reality bashed on their total mh?

With ryo's fork using dynamic penality, deepbit(backup) and btcguild(based on poolspeed it actually worked out positive jumping in even with the hour delayed stats). Im not going to discuss the efficiency level before someone spams btcg owner with this post. Lets just say its +EV so far.

simonk83

hero member

Activity: 798

Merit: 1000

Quote from: ed64 on August 03, 2011, 11:06:14 PM

Also remove "role: mine" from pools.cfg if you have having issues disabling bloodys....

Ah ha, what's that doing in there

Thanks

Clipse

hero member

Activity: 504

Merit: 502

Quote from: ed64 on August 03, 2011, 11:06:14 PM

Sorry for the funky slice merge, the latest AltSliceScheduler works well, tested it for a few hours, looks good to me, hopped everything it was supposed to and switched to eligius.st when pools became unhoppable.

For the round time scheduler, I'm looking to add in Ryouiki's algorithm from his fork, as yet another way of hopping. His hopping may be more invasive, so I've left it for last.

In the meantime, I like his idea of adding a "penalty" option for each pool. Rather than have "mine" and "mine_slush" we add in penalty: 1.5 for say ozco.in if they delay stats or something like that.

In Ryouiki's UI, he has the ability to manually change the penalty. He's also enhanced his hopping algorithm with a feedback loop to adjust penalties on the fly...

Also remove "role: mine" from pools.cfg if you have having issues disabling bloodys....

Yeh, I love the ryo dynamic penalty system, waiting for that to get added to c00w's version or your fork

ed64

newbie

Activity: 42

Merit: 0

Sorry for the funky slice merge, the latest AltSliceScheduler works well, tested it for a few hours, looks good to me, hopped everything it was supposed to and switched to eligius.st when pools became unhoppable.

For the round time scheduler, I'm looking to add in Ryouiki's algorithm from his fork, as yet another way of hopping. His hopping may be more invasive, so I've left it for last.

In the meantime, I like his idea of adding a "penalty" option for each pool. Rather than have "mine" and "mine_slush" we add in penalty: 1.5 for say ozco.in if they delay stats or something like that.

In Ryouiki's UI, he has the ability to manually change the penalty. He's also enhanced his hopping algorithm with a feedback loop to adjust penalties on the fly...

Also remove "role: mine" from pools.cfg if you have having issues disabling bloodys....

joulesbeef

sr. member

Activity: 476

Merit: 250

moOo

is it really worth hopping btcguild? with the hour delay?.. how are you working out when to hop? are you just saying like 23% takes an hr and adjusting what you think the shares should be in reality bashed on their total mh?

and the btcpool24 guy says they found a block but his payment going to take time cause he is a crappy coder or something like that.. they are also changing the proportional equation which in my boat means something other than prop

i'm just going to leave them disabled until my account has a ballance

ed64

newbie

Activity: 42

Merit: 0

Thanks, try my latest pull, the one i posted was flaky, couldn't get it to slice predictably, maybe you can look at it? I tried to simplify...

As for jumping pools, with the getwork store now, it doesn't seem to be much of an issue, even when i switched pools every 30s or so when using a small slice size for testing... I did add in a min slice size option to prevent too frequent hopping though.

Quote from: flower1024 on August 03, 2011, 07:08:33 PM

Quote from: ed64 on August 03, 2011, 06:56:54 PM

Quote from: paraipan on August 03, 2011, 06:50:05 PM

Quote from: djex on August 03, 2011, 06:37:54 PM

So for some reason it likes to get stuck on BTCPool24.

yep, something is not right, 40593 seconds from beginning of the round (11 hours) at 30Gh on average (have seen 50Gh+), has only 288422 shares...

Try my AltSliceScheduler: https://github.com/echiu64/bitHopper
Use options like

Code:

--scheduler AltSliceScheduler --altslicesize=1200

Default is 600 "slice" or "pie" size, in that the 600 will be divided amongst the pools, weighted by their share count.

For really fast switch set it to 60. Takes about a minute or so to wait for all the stats to initialize and have the slice redistribution get set. Not sure if there is a better way to trigger this earlier.

Still using the previous stats page for the alt slice scheduler, so should be functional...

You can see the time slice count on the stats page as well and will decrement over time. The next largest slice will be selected. Once all are exhausted, slices are recalculated...

Credit goes to flowers for the original implementation

really nice! thank you very much for that.

just a little suggestion:

if "reslice" occurs don't take the pool with the biggest slice. instead check if the current pool got another slice and if so stay there -> reducing hopping a little and worked well for me (i changed it in a later version, so you might not have catched it up).

i'll switch to your version now. do you want me to make the suggestion as a pull request? its really not much

simonk83

hero member

Activity: 798

Merit: 1000

Quote from: c00w on August 03, 2011, 10:30:13 PM

Slice scheduler now works.
And I just realized that macboys port looks really glarish to me.
Does everyone like that color scheme better or worse?

Right, now we seem to be working

Quick thought though, would be it worth disabling the slicing when on Slush, as you're only on there till 10% anyway which tends to be over pretty quickly.

muyoso

member

Activity: 84

Merit: 10

Quote from: Clipse on August 03, 2011, 10:27:18 PM

On a sidenote, looks like btcpool24 run off with the 2nd block or he is just to lazy to respond in the forum. Json says blocks found:2 yet the site disabled mining. Another gripe I have with btcpool24, this recent block they seem to misreport ~15% of shares less than actually submitted / accepted, wtf .

Just post this:

http://www.youtube.com/watch?v=6Pv9oRxPOIc

Clipse

hero member

Activity: 504

Merit: 502

For those in the know-how, how to get btcguild(default mining) and deepbit(backup mining) working the way it did before the pool / user config splits.

I used this in the old pool.py and it worked out pretty good for me but have no clue how to get the same structure in the pools.cfg to make it work the same again:

btcguild - Current btcguild code isnt fixed to read the correct shares location.

Code:

    def btcg_sharesResponse(self, response):
        parsed = response.replace('\n', '').replace('\r','').replace('\t','')
        outputTime = re.search('([:0-9]*)', parsed)
        outputSpeed = re.search('([ 0-9\.]*)GH/s', parsed)
        if outputTime != None and outputSpeed != None:
            server = self.servers['btcg']
            extractTime = outputTime.group(1).split(':')
            roundTime = int(extractTime[0])*60*60 + int(extractTime[1])*60 + int(extractTime[2])
            server['ghash'] = float(outputSpeed.group(1).replace(' ',''))
            server['duration'] = roundTime
            round_shares = int( server['ghash'] * roundTime / 4)
            self.UpdateShares('btcg',round_shares, True)
        else:
            self.bitHopper.log_msg('regex fail : btcg')

deepbit

Code:

    def deepbit_sharesResponse(self, response):
        round_shares = self.bitHopper.difficulty.get_btc_difficulty()*0.431
        self.UpdateShares('deepbit',round_shares)

Clipse

hero member

Activity: 504

Merit: 502

Quote from: c00w on August 03, 2011, 10:30:13 PM

Slice scheduler now works.
And I just realized that macboys port looks really glarish to me.
Does everyone like that color scheme better or worse?

The idea originally sounded nice but after trying it for few mins I wanted to puke Tongue

Topic: bitHopper: Python Pool Hopper Proxy - page 90. (Read 355678 times)