[4+ EH] Slush Pool (slushpool.com); Overt AsicBoost; World First Mining Pool

[naypalm]

Looks alright on my end 

[slush]

This weekend (29-30. October) I'm migrating pool from London (linode.com) to France (ovh.com). Although I believe everything will be fine, there will be at least one outage during database migration (probably something around 30 minutes).

[kislam]

Question for Slush:

In describing the logic/calculation of the score-based system, you mentioned this:

====
Matematically said, for every submitted share, pool perform

Code:
score = score + exp(round_time/C)

round_time is count of seconds between "now" and time when round started, C is magic constant which define how fast old shares lose their value. All following calculations are using C=300, but it may be changed in the future to provide the best cheatproof/usability ratio.

And on the end of round, system calculate rewards using formula

Code:
reward = user score / total score * 50
====

Is it really done for EVERY submitted share for EVERY user? that's a lot of calculations, no? or is there some other interval used to calculate the scores? maybe the scores are calculated at the end of each round?

please enlighten....

[slush]

score = score + exp(round_time/C)

This is done for every submitted share immediately.

reward = user score / total score * 50

This is done on the end of round.

Is it really done for EVERY submitted share for EVERY user?

Yes, why not? It's one exponent function per share. With current multigigahertz machines it really isn't an issue :-).

[kislam]

score = score + exp(round_time/C)

This is done for every submitted share immediately.

reward = user score / total score * 50

This is done on the end of round.

Is it really done for EVERY submitted share for EVERY user?

Yes, why not? It's one exponent function per share. With current multigigahertz machines it really isn't an issue :-).

It's one exponent function per share PER USER. if you have 200 users you are doing this calculation 200 times for each share? then adding the new score to the cumulative score for each user and storing that in memory? for all the millions of shares?

not that it would be difficult for the modern pc, but my question is if it is done on an ongoing basis, then why does it take about 10 minutes to publish the rewards after each round?

BTW, thx for the lightning fast response

[kislam]

score = score + exp(round_time/C)

This is done for every submitted share immediately.

reward = user score / total score * 50

This is done on the end of round.

Is it really done for EVERY submitted share for EVERY user?

Yes, why not? It's one exponent function per share. With current multigigahertz machines it really isn't an issue :-).

It's one exponent function per share PER USER. if you have 200 users you are doing this calculation 200 times for each share? then adding the new score to the cumulative score for each user and storing that in memory? for all the millions of shares?

not that it would be difficult for the modern pc, but my question is if it is done on an ongoing basis, then why does it take about 10 minutes to publish the rewards after each round?

BTW, thx for the lightning fast response

sorry, just realized my mistake, a share is not submitted by multiple users and still be valid. so no need to multiply by number of users...

[worldinacoin]

It is great mining for NMC in Slush's Pool but my luck is real bad on BTC for quite sometime.   

[slush]

It is great mining for NMC in Slush's Pool but my luck is real bad on BTC for quite sometime.   

Is your expected round reward significantly lower than it should be? For eliminate effect of score based system, you can take last 50 round rewards and recalculate it with proportional method (your hashrate against pool hashrate, which is pretty constant around 1170 for last few days).

If this fits, then it's pool luck (well, we performed pretty bad last few days). If it don't fit, then we need to investigate more. Usually it's because of some connection troubles, rejected shares or crippled miner software or cards. Actually I don't have a reason to think there's some flaw in pool code, I didn't touch important stuff for many weeks and merged mining cannot affect this.

[DutchBrat]

It's a good thing luck evens out in infinity,

so give it a few more years and we will be alright  

Brat

[slush]

It's one exponent function per share PER USER. if you have 200 users you are doing this calculation 200 times for each share? then adding the new score to the cumulative score for each user and storing that in memory? for all the millions of shares?

As you already realized, score is calculated only for one share submit, not for submit*users :-).

[DutchBrat]

I am experiencing a lot of:

'Pool not providing work fast enough'
'Pool not responding'

Errors

Is this because of the switch to a French provider ?

[slush]

Is this because of the switch to a French provider ?

Not yet.

It's cgminer, right? Whats your ping to api.bitcoin.cz?

Edit: I'm monitoring pool response times once they took longer than 2 seconds; it happen time to time and it's an idicator that something is broken. But now there's no problem with response times, they're far under this threshold (actually most loaded pool backend has server load 0.15, which is excellent), so I don't see any reason why you should see this error.

Edit2: Also is there any ping packetloss?

[DutchBrat]

Pinging api.bitcoin.cz [178.79.183.97] with 32 bytes of data:

Reply from 178.79.183.97: bytes=32 time=6ms TTL=54
Reply from 178.79.183.97: bytes=32 time=6ms TTL=54
Reply from 178.79.183.97: bytes=32 time=6ms TTL=54
Reply from 178.79.183.97: bytes=32 time=6ms TTL=54

Looking good to me

[slush]

Pinging api.bitcoin.cz [178.79.183.97] with 32 bytes of data:

Looking good to me

Yes, it's fine. How often do you see that error? How long it is since last message?

[Tartarus]

Pinging api.bitcoin.cz [178.79.183.97] with 32 bytes of data:

Looking good to me

Yes, it's fine. How often do you see that error? How long it is since last message?

I've been seeing this off-and-on today, also with cgminer, going to api2.  I'm keeping an eye out for it happening again now..

[DutchBrat]

Maybe some 20 mins

My hashrate over all my miners is dropping steadily (this one miner hashes at around 280 MHash but has an avg of 213 MHash now, I restarted it 2 hours ago)

The estimated reward is also dropping from 0.08 to 0.049 at this very moment which is consistent with the drop in my hashrate because of the connection problems

Strange, maybe a hiccup on my own network

I'll let the ping run for a while

[slush]

What's your pool username? (I don't see submits from DutchBrat in last moments) I'll follow it on my side.

[DutchBrat]

Username: dutchbratslaves

I am trying to copy paste the output of cgminer, but that proves more difficult than I thought 

[slush]

Hm, strange. Everything looks ok here, all your requests are responded in less than 5 miliseconds and I also don't see any rejected shares or so, which usually indicate some troubles.

[DutchBrat]

Very strange,

Restarted all my miners, everything was fine, untill the last LongPoll

Now I get the 'Pool 0 not providing work fast enough'  messages again....

It's stabilizing again, so it seems to be related to a new found block

P.S.

You have bigger things to worry about with your upcoming switch, thanks !