Topic: [1200 TH] EMC: 0 Fee DGM. Anonymous PPS. US & EU servers. No Registration! - page 78. (Read 499602 times)

now you're over 2/3rds of deepbit's network portion, good job. over btc guild

I don't know what the threshold would be specifically, I haven't done enough testing to really get a good handle on it what it means for what hashrates.  I do know that watching cgminer is kinda weird when you don't see the shares flowing by.

Which server, goxed?

BR0KK: Yes, the hashrate is accurate.  We picked up all the OzCoin miners that left it seems.  Blocks are flowing nicely though.

What's the pain threshold using diff10?  Does the stale issue only get bad when under 1GH?

Hey Inaba, what would the specs of this server? Just curious, nothing else.

Is the payout ok?

I'm getting less then usual (i think)

Does the pool really have 1.95 th?

It will cut your reward per block in half, true, but it will also give you 2x as many blocks to make up for it, on average.  Hopefully the luck is turning around.  It's been pretty crappy this week for sure.

Earning as BTC between now and the reward halving is my main goal. I consider myself a small miner and the variance already is killing me. My vote would be to keep from making any drastic changes that could effect miners earnings or variance of, untill after the reward halving.

The fluctuation in the pool speed thats been spoken of basically cut my earnings per block in half. Some being new miners some being backed up shares and then the time its taking to balance back out is costing me now that blocks are being found quickly after the bad luck lately.

The DGM calculations are not a big deal, it's taking in and processing all the shares in order due to DGM that is the bottleneck.  If I could take them in any order, it would make things a lot easier, but it has to be strictly linear for DGM. 

Hmm... changing everything to 2diff might be an option... that should not affect anyone in any meaningful way and the servers should be ready for it.  Let me give that some thought.

Inaba, could you just change everything to a difficulty of 2 to help with share processing until you find a way to have people run at different difficulties?

I don't live in the Europe so no EU does not effect me, but could you host a PPS only server there so it does not have to do DGM calculations? Or would that be just be silly, I don't know how many people use PPS. 

I'm a little miner @ 200 Mh/s so I'm guessing that would affect me.  BTW I just came back after a-whoring after false pools. 

There's a minimum of 5 servers running at any particular time at the moment.

Yes, there's a chance to add one in the EU.  I'm making changes to make an EU server far more possible than in the past.  Before, I needed a pretty beefy server, and it was wicked expensive.  The changes I'm making will allow for a much more lightweight server, which means I can probably afford to host one in the EU now.

There's some math problems integrating DGM with mixed difficulty shares.  Meni is pretty busy right now, so he's not as available as I had hoped, but right now, all servers are back to running 1diff shares.  Everything is ok as long as everyone is running the same difficulty, but 10diff shares seem to generate more score than expected when compared to 1diff x 10 scoreing and I can't figure out why.  I thought it was a precision issue, but after doing a bunch of experiments (pretty much spent all day yesterday), it doesn't appear to be the case.  It, as of right now, appears to be a formula issue.

I thought I had it cracked yesterday when I found a bug in the code that was still assigning 1diff values to part of the score, but fixing that did not solve the issue sadly. 

I guess I'll go stare at it again and hope for some inspiration. 

As for variance, yeah, it could increase stales for really slow miners, however from my testing so far, it hasn't had much affect at 1 GH/s... not sure about slower than that though.  I really want to get this working for everyone, not just 10diff across the board.

I suspect much of that 700GHps came from Ozcoin patrons who didn't want to pay the 3% fee instigated yesterday. It happens to match the hashrate drop at Ozcoin.

To address your question: the only effect difficulty-10 shares should have is to increase variance. The specific details depend on the pool's DGM parameters. The payout expectation will not change so in the long run it makes no difference. It *should* have a positive effect on pool performance since it will only need to send out 1/10th of the getworks it is currently doing.

My only concern would be the effect on stales. If a miner is working on a 10-difficulty work, will there be a significantly greater chance of stales? Since the miner would be reporting to the pool only 1/10th as often, intuitively it would seem that there would be a greater change of a work unit becoming stale when a new block is announced on the network (instead of miner throwing away, say 5 seconds of work, they could potentially be throwing away 50 seconds of work when that happens).

I know you've been running a test server with a greater-than-1 share difficulty ... what were your results with that?

Yeah, when hashrate spiked on EMC i checked main pools and saw ozcoin seem to have lost around 800 GHs..

https://bitcointalksearch.org/topic/alternatives-to-the-new-ozcoin-3-dgm-fee-99954 is probably where the new hashes came from. I hope they respect this pool enough to give back via a donation.  Kinda upsets me how so many waited till the.last minute to leech off of his 0% till the last freaking second before moving on. His 3% fee was implemented just overnight.

Just out of curiosity, how many servers do you currently run for your pool (if you don't mind sharing)?
Is there a chance to add one in the eu?

Ok... so the hashrate of the pool increased by 700 GH/s overnight.  No idea where that came from and that's why things started getting backed up.  I knew we were up against a hard limit on share processing and was the driving reason behind adding another DC to handle the load, but I didn't expect 700 GH/s overnight.

It's all good now though, I have it covered and I can deploy even another server (each server should be able to handle ~1.5 - 1.7 TH/s in it's current configuration) if need be in a few minutes.

Now let me ask this:

What do you guys think about doing away with difficulty 1 shares entirely and moving to difficulty 10 across the board?  How many people would that impact ultimately?

Inaba, much respect for you being on top of things, I appreciate all you do for us.

It is not correct.  Share processing is almost caught up, it should be accurate within an hour.

So, you're saying the 3.54THash is incorrect?

I thought it was getting attacked or something of the sort. Thanks for clarifying.