Setting up a new pool - some pre-launch feedback welcome

[holgerr]

Hi,

I'm in the process of setting up a new Bitcoin mining pool that is designed to offer maximum stability through the use of distributed nodes and a good level of redundancy, as well as a complete separation of the web frontend from the mining backends. The backend part of this is already running quite nicely based on Eloipool and I have a few miners hashing on it just for testing purposes. I'm in no rush to get this launched, so I'd rather do it right than do it quick and dirty. That's why I'd like to solicit some feedback from miners and other pool operators on some of the decisions I have to make.
First off a technical question, though - has anyone written a mysql user database plugin for eloipool? I could write one myself, but if someone would be willing to share...
Okay, the main other question I have is this: what kind of payment model would miners prefer? I can't and won't do PPS and am looking at ways to reward miners fairly and reliably, with as little variance as possible, while keeping the risk to the pool to a minimum and discouraging pool hopping. Slush's method seems pretty fair, but I don't think his algorithm is available for others to use.
Also there's the question of fees - I'd like to keep those to a minimum, but need to charge some to fund the cost of servers and bandwidth. Currently I'm thinking that the pool should keep transaction fees - above that I'm wondering what to do. One thought is to ask for a 1% donation and in exchange to provide some exclusive donors only features - are any pool operators willing to share their experience with the rate of donations received if that's an option on their pool?
Last question: the plan is to have the pool as a Stratum only pool. Considering that most other pools are phasing out Getwork, am I on the right track with that?
Thanks for any useful feedback - any additional suggestions are also appreciated.

Holger

[1714801284]

1714801284

[1714801284]

1714801284

[doublec]

Miners tend to prefer PPS as it's a guaranteed constant income. Pool operators prefer not to have PPS. DGM, PPLNS and CPPSRB seem popular and non-hoppable. DGM has lots of knobs to turn to adjust from high to low variance on the miner or pool side at a cost of being more difficult to understand. I like it though. Miners seem to find it hard to get a handle on so it can scare people off.

With regards to transaction fees, the feedback I've had from miners in my pool is they overwhelmingly want them. My pool pays orphans but keeps fees to cover it but it seems miners prefer it the other way. I'll probably be switching. Donation based pools are fine as long you don't expect any donations. You probably won't get many. You'll want to be able to pay the server costs out of pocket for a while as most funds will be used to keep the pool solvent in the beginning. Even tiny errors can end up costing big and it's hard to recover from an error when pool fees are so low.

Stratum-only seems reasonable. GBT is also a fine choice. GetWork is probably on the way out.

[kinlo]

Your technical question: the latest version of eloipool has module based authentication.  There is no SQL based plugin yet, at least not on all the pools that use eloi that I know of, but it would be appreciated if that is contributed to the code.  eloipool is AGPL so you are obliged to do so if you use it on your pool     It can even be written for you if you want, but I understand that you have the technical know-how.

About the payment model.  What you are asking is impossible.

No variance = pps
low variance = dgm with the correct settings, but that implies the same drawbacks for the poolops as PPS so that's impossible, OR being a big pool.
you could try to get lower variance by using PPLNS with a very high N, but still people would only get paid when a block is found therefore  the variance can be high if several days go by without you finding any blocks.

So basicly, you can't do what you want.

Also note that slush's scoring algorithm is hoppable and should not be used.   I believe slush announced he would switch to dgm at some point, but afaik he didn't complete that yet.

About stratum only: I guess that's the best way to start.  Getwork will cease to exist.  GBT has it's advantages, but also it's disadvantages....  future shall decide how that goes, but for now stratum is just fine.

[holgerr]

Thanks for the answers so far, those gave some useful insights. I'm glad I am in no rush to get things going and can take the time to research my options properly.
Just to clarify, when I talked about donations I wasn't thinking about one off donations to the pool's expenses, but rather an optional percentage fee on the miner's earnings, unlocking some perks (such as mining on port 80 or something similar).
Kinlo, I also thought that the AGPL would require a MySQL module for Eloipool to be made available as source, so I'm wondering if nobody has actually written one or if those that have don't comply with the license. Anyway, if I end up writing one (alternatively I guess a flat file based approach could be more resource friendly...), I'll make sure it gets posted so that people can find it.
Wasn't aware that Slush's algorithm is hoppable - the impression I got from reading about the various methods was that it is specifically designed to prevent that. But it's quite possible that I missed something there. Do you happen to have any details?

[kinlo]

Kinlo, I also thought that the AGPL would require a MySQL module for Eloipool to be made available as source, so I'm wondering if nobody has actually written one or if those that have don't comply with the license. Anyway, if I end up writing one (alternatively I guess a flat file based approach could be more resource friendly...), I'll make sure it gets posted so that people can find it.

Afaik the only pool using authentication with eloipool is Triplemining, and we don't use mysql for that.

Wasn't aware that Slush's algorithm is hoppable - the impression I got from reading about the various methods was that it is specifically designed to prevent that. But it's quite possible that I missed something there. Do you happen to have any details?

Slush's method was indeed specifically designed to prevent hopping.   The Titanic was specifically designed to be used for many, many years on sea.  Unfortunately things don't always play out the way one hoped.  You should read and understand meni's excellent document about poolhopping.

[kevindeangelis]

I recently added a list of pools (it's short right now) to the site I recently set up last week - www.WhereToMine.com  - Can you send me some info about your pool?  I can add it to the list.

Also, if anyone has a good pool that they use that isn't on the list, let me know and I'll add it.

Thanks!

[holgerr]

Thanks kinlo, that article on pool hopping is a fascinating read.