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

[TiborB]

Here's some reading for you both:

http://organofcorti.blogspot.com/2012/04/41-slushs-pool.html
http://organofcorti.blogspot.com/2012/08/42-slushs-score-method-and-miner.html
http://organofcorti.blogspot.com/2012/09/43-slushs-score-method-and-miner.html

Probably shouldn't comment since I haven't had a chance to read those yet, but I did mention above about how I thought the system was outdated.
Judging by the hash rates in the top of those they aren't very recent.
If im wrong, please dont hang me! I'll apologise for my ignorance once I have had time to look at things properly.

4.2 and 4.3 are the important ones - 4.1 is just background. They are from last year, but the score method is the same now as it was then.

Very good read indeed. I wonder how many of the forum members will take the time to read & understand (!) it. I see a rather wide spectrum of skills and attitude and many reoccurring questions that makes me assume the majority will just skip anything longer than a few paragraphs. Anyway, congratulations, I really recognize the full worth of these quality articles.

Cheers,
   T

[ewitte]

Understand yes - freak out when network interruption at the end of a block YES

[not.you]

Here's some reading for you both:

http://organofcorti.blogspot.com/2012/04/41-slushs-pool.html
http://organofcorti.blogspot.com/2012/08/42-slushs-score-method-and-miner.html
http://organofcorti.blogspot.com/2012/09/43-slushs-score-method-and-miner.html

Probably shouldn't comment since I haven't had a chance to read those yet, but I did mention above about how I thought the system was outdated.
Judging by the hash rates in the top of those they aren't very recent.
If im wrong, please dont hang me! I'll apologise for my ignorance once I have had time to look at things properly.

4.2 and 4.3 are the important ones - 4.1 is just background. They are from last year, but the score method is the same now as it was then.

Very good read indeed. I wonder how many of the forum members will take the time to read & understand (!) it. I see a rather wide spectrum of skills and attitude and many reoccurring questions that makes me assume the majority will just skip anything longer than a few paragraphs. Anyway, congratulations, I really recognize the full worth of these quality articles.

Cheers,
   T

Yes I found those the other day and thought they were excellent as well.  (When is the next one coming out?)

I kept seeing blocks that had under-calculated (by a lot) and it was bugging me until I found those.

The TL;DR version is that in order to prevent pool hopping Slush's scoring mechanism rewards shares found later in the round more than shares found earlier in the round.  This mechanism as implemented also introduces variance in the pay out.  Sometimes you get more than expected for a block and sometimes you get less as compared to a purely proportional payout system.  To check, look at your highest and lowest rewards over time and you should find some lower and some higher than the value calculated by this formula: (Block Value * 0.98 * your shares) / total shares

If you are a full-time miner however, the long term effect of the variance will disappear and your long term reward will look almost identical to what it would if that formula above were actually being used.  The end result is that pool hopping from Slush's pool is only profitable to a very, very tiny degree and full time miners are not losing much reward to pool hoppers. 

Slush could lower that pool hopping profitability even more but it would raise the payout variance even more which would make some blocks look even worse.

[organofcorti]

Here's some reading for you both:

http://organofcorti.blogspot.com/2012/04/41-slushs-pool.html
http://organofcorti.blogspot.com/2012/08/42-slushs-score-method-and-miner.html
http://organofcorti.blogspot.com/2012/09/43-slushs-score-method-and-miner.html

Probably shouldn't comment since I haven't had a chance to read those yet, but I did mention above about how I thought the system was outdated.
Judging by the hash rates in the top of those they aren't very recent.
If im wrong, please dont hang me! I'll apologise for my ignorance once I have had time to look at things properly.

4.2 and 4.3 are the important ones - 4.1 is just background. They are from last year, but the score method is the same now as it was then.

Very good read indeed. I wonder how many of the forum members will take the time to read & understand (!) it. I see a rather wide spectrum of skills and attitude and many reoccurring questions that makes me assume the majority will just skip anything longer than a few paragraphs. Anyway, congratulations, I really recognize the full worth of these quality articles.

Cheers,
   T

Yes I found those the other day and thought they were excellent as well.  (When is the next one coming out?)

I kept seeing blocks that had under-calculated (by a lot) and it was bugging me until I found those.

The TL;DR version is that in order to prevent pool hopping Slush's scoring mechanism rewards shares found later in the round more than shares found earlier in the round.  This mechanism as implemented also introduces variance in the pay out.  Sometimes you get more than expected for a block and sometimes you get less as compared to a purely proportional payout system.  To check, look at your highest and lowest rewards over time and you should find some lower and some higher than the value calculated by this formula: (Block Value * 0.98 * your shares) / total shares

If you are a full-time miner however, the long term effect of the variance will disappear and your long term reward will look almost identical to what it would if that formula above were actually being used.  The end result is that pool hopping from Slush's pool is only profitable to a very, very tiny degree and full time miners are not losing much reward to pool hoppers. 

Slush could lower that pool hopping profitability even more but it would raise the payout variance even more which would make some blocks look even worse.

Well, TiborB,  there's you answer - not.you read them and understood them

Good job understanding and explaining in simple terms what the effects of Slush's score method is. It is exploitable, although the possible cost to miners is minimal, and no-one is pool hopping anymore anyway. Slush's score was why this pool was my favourite for so long. The basic idea is simple, as not.you has explained, but it had hidden complexities that were waiting to be discovered. It's a great proving ground for your mathematical abilities, and without Meni Rosenfeld's help I would have fallen in battle.

As for when the next one is coming out ... um, I kind of lost track of time. Since there's so much interest in the pool atm, I'll get started on it.

[agento2]

Depends on how long the outage was. Score is an exp function so your earning degrade to 0. I'm not sure how long it takes. I have been thinking listing the shares on the stats page isn't the best metric. Have score and total score would be better since those are what your earnings are actually calculated on.

I know why the functions are there, but its effects do seem a little harsh. Maybe it could be reduced a little? Otherwise those with massive rigs could switch do the opposite of the normal problem, run them on a PPS pool, then they notice its been 45 minutes here switch.

sadly your argument is the opposite of what it's meant to be, you just described the very actions that the score system are trying to deter. And if people were doing that he would have to make the functions even harsher.

[agento2]

Surely if the rounds been going on one pool for 45 minutes, Then no matter how long the round continues, thats 45 minutes where you are being paid on a PPS. You then earn a few shares from the pool with considerable value. If the round ends within a couple minutes, its not much but you're miners haven't lost much.
If its a few hours, then for the 45 minutes you have been making coins while everyone else here has shares which value is now void.

Now whether 45 minutes is the sweet spot I don't know, and without looking into the maths properly / how much you get per share at a PPS i can't be certain you would earn more, but It seems probable.

Will work it out later. Hopefully you are right and the numbers will prove me wrong!
I mine on a fairly low power system. 40Mhash. If my last share is 30s from the end of the round I seem to get twice as much as if its 2 minutes!

Ok, I am surprised that the scoring system works that way....

It means even if I am an ASIC miner if I have an outage of 90minutes due to a storm then I lose all my shares value. Everyone else profited from my shares. So down time is my enemy.

How does the score system work? (I am sorry for my ignorance!)

----------------------

The thing that came to mind though is the following after being surprised by this part of the score system:

If I run really high difficulty shares...let say 512...and I therefore submit larger shares but less frequently...is this the reason why I earn alot less?

I noticed if I submit very low difficulty shares with a very high frequency...my score seems to stay very low (closer to 100 than 1 billion. Yet, if I submit shares at 512 difficulty every minute, my score goes into the hundreds of millions over time. I assume from watching this that my delay in submitting very high difficulty shares is actually hurting me in my score. Is this correct? If so, then this deeply discourages using high difficulty...right? I can see this turning into an ASIC-festival of people submitting diff1 shares. With BFL coming out and delivering (a minute amount) of ASICs...I assume this is going to be a bandwidth problem in the future, correct?

I eventually decided to run at diff-1 because the super short rounds didn't give me much of any chance as an ASIC to submit the higher difficulty shares.

I find myself scratching my head as what is the right thing to do. It seems to discourage any lapse in submitting shares, even if only a minute between submissions.

I also understood that a low score in 4 digits is better than a high score in the billions. Correct?

The score should be high i'm pretty sure. But as was said at some point in the recent past periodically resets just do to large numbers (computers can only count so high easily). And the score system sadly has no way of knowing if you have an outage or you just went to another pool for a while. Stratum was slush's answer to bandwidth issues. Your view that everyone else profited from your shares is a bit off too though I'm pretty sure. I haven't read enough on how pools actually work but I would guess that the older the share the less useful it is.

[ewitte]

Depends on how long the outage was. Score is an exp function so your earning degrade to 0. I'm not sure how long it takes. I have been thinking listing the shares on the stats page isn't the best metric. Have score and total score would be better since those are what your earnings are actually calculated on.

I know why the functions are there, but its effects do seem a little harsh. Maybe it could be reduced a little? Otherwise those with massive rigs could switch do the opposite of the normal problem, run them on a PPS pool, then they notice its been 45 minutes here switch.

sadly your argument is the opposite of what it's meant to be, you just described the very actions that the score system are trying to deter. And if people were doing that he would have to make the functions even harsher.

That is the way it is though if your on a 4h block you can get 99% payout just being on the last hour.  There has to be some way of determining how long you were connected and calculate off that?

[Lucko]

Ok, I am surprised that the scoring system works that way....

It means even if I am an ASIC miner if I have an outage of 90minutes due to a storm then I lose all my shares value. Everyone else profited from my shares. So down time is my enemy.

How does the score system work? (I am sorry for my ignorance!)

There are some resets of the scores also so after a reset you don't get any... 90 minutes is more then enough for reset to accrue... Probably more then once. So even 1000 ASIC wouldn't help... That why I was asking for pool to have a maintenance mode and that way I'm using backup PPS pool in time of maintenance... And also why backup pool is PPS. So I get something out... If the backup pool would use Slush method for short outage I would get close to 0. But then again I could be lucky... For some time I used PPLNS method but I got out about 10% of what I lost so PPS is probably much better... But then again I might be wrong...

[organofcorti]

Ok, I am surprised that the scoring system works that way....

It means even if I am an ASIC miner if I have an outage of 90minutes due to a storm then I lose all my shares value. Everyone else profited from my shares. So down time is my enemy.

How does the score system work? (I am sorry for my ignorance!)

There are some resets of the scores also so after a reset you don't get any... 90 minutes is more then enough for reset to accrue... Probably more then once. So even 1000 ASIC wouldn't help... That way I was asking for pool to have a maintenance mode and that way I'm using backup PPS pool in time of maintenance... And also why backup pool is PPS. So I get something out... If the backup pool would use Slush method for short outage I would get close to 0. But then again I could be lucky... For some time I used PPLNS method but I got out about 10% of what I lost so PPS is probably much better... But then again I might be wrong...

Not how the "reset" works. It's more of a "renormalisation".

[Epoch]

I suspect you are being hasty in blaming stratum.  My CGMiner instances - 5 of them, have been running on slush's pool continuously for over a month with stratum without any downtime.

I merely described my personal experience; YMMV. The vast difference in possible hardware configurations between users, and the different effect that has on code execution in the software, makes this impossible to generalize.

I can say that, for me, mining on Slush using both stratum and getwork has been quite stable. I can also say that mining on some other pools (bitminter is one example) using stratum has not, while getwork has.

For example (again, not to generalize my own experience) my stratum instance crashed 3 times in the past 3 days; my 2 getwork instances have not.

[digital]

I have been using stratum, without any stratum related crashes, since it was introduced.  And I was a beta tester.

But I also have only used Slush.  If I'm not mistaken, Slush created stratum, so I would imagine he has a better handle on how to use it than some other pools may have.

[Lucko]

Ok, I am surprised that the scoring system works that way....

It means even if I am an ASIC miner if I have an outage of 90minutes due to a storm then I lose all my shares value. Everyone else profited from my shares. So down time is my enemy.

How does the score system work? (I am sorry for my ignorance!)

There are some resets of the scores also so after a reset you don't get any... 90 minutes is more then enough for reset to accrue... Probably more then once. So even 1000 ASIC wouldn't help... That way I was asking for pool to have a maintenance mode and that way I'm using backup PPS pool in time of maintenance... And also why backup pool is PPS. So I get something out... If the backup pool would use Slush method for short outage I would get close to 0. But then again I could be lucky... For some time I used PPLNS method but I got out about 10% of what I lost so PPS is probably much better... But then again I might be wrong...

Not how the "reset" works. It's more of a "renormalisation".

OK then why my scores goes to 0? If it was renormalisation it would be some small number... And since I have a lot of problems with my miner right now I see score 0 toooooo many times right now...

[digital]

Ok, I am surprised that the scoring system works that way....

It means even if I am an ASIC miner if I have an outage of 90minutes due to a storm then I lose all my shares value. Everyone else profited from my shares. So down time is my enemy.

How does the score system work? (I am sorry for my ignorance!)

There are some resets of the scores also so after a reset you don't get any... 90 minutes is more then enough for reset to accrue... Probably more then once. So even 1000 ASIC wouldn't help... That way I was asking for pool to have a maintenance mode and that way I'm using backup PPS pool in time of maintenance... And also why backup pool is PPS. So I get something out... If the backup pool would use Slush method for short outage I would get close to 0. But then again I could be lucky... For some time I used PPLNS method but I got out about 10% of what I lost so PPS is probably much better... But then again I might be wrong...

Not how the "reset" works. It's more of a "renormalisation".

OK then way my scores goes to 0? If it was renormalisation it would be some small number... And since I have a lot of problems with my miner right now I see score 0 toooooo many times right now...

If you are having problems with your miner, you should probably get those resolved first before complaining about what slush is doing.  Seriously, when my miners are stable, which is almost all the time, my payouts are constant and regular.  I've never had to worry about the score normalizing.

[genpayne]

I know why the functions are there, but its effects do seem a little harsh. Maybe it could be reduced a little? Otherwise those with massive rigs could switch do the opposite of the normal problem, run them on a PPS pool, then they notice its been 45 minutes here switch.

sadly your argument is the opposite of what it's meant to be, you just described the very actions that the score system are trying to deter. And if people were doing that he would have to make the functions even harsher.

The function is there to stop people from hopping after mining at the start. I am suggesting the opposite.
For example, the 6:43 block this morning. If you noticed that at 6 hours you can be reasonably confident that its not going to be too much longer. It could be, but statistically it is unlikely.
So you were mining on either a PPS or even another pool that’s just started a block. Statistically you would expect our block to finish before the average block length on your current pool. You hop over to us and on average are better off.

Now, to do this manually would be tricky since you would have to monitor multiple pools etc. But to script it would hardly be tricky.

Without looking at the maths its impossible to work out whether the gains would be worth the effort
I mentioned before, I am in the middle of exams at the moment and have little time to work on this.
Or to look properly at the algorithms in place (although I did do last week so am working from those memories).

Next week I will take a proper look. I might be completely wrong! But from the limited information I have it was something I considered feasible. Nobody (As far as I can see) has provided a quick explanation as to why not. organofcorti presented some links to documents which may well do this, but I currently have not got the time to read them properly. Hence I can't comment on them.
If there is a quick explanation please give it? If not I will take a proper look next week. If I still feel this is possible I will probably attempt to prove it by having two identical clusters of machines, one just mining slush and the other attempting what I have suggested.

Then I can report back with some actual figures to back this up, or admit to myself It doesn’t work out on paper as it does in my head.
If anyone does have the time now feel free to take a look!

That is the way it is though if your on a 4h block you can get 99% payout just being on the last hour.  There has to be some way of determining how long you were connected and calculate off that?

You can get an even payout using pay per share! But the overall average payout is less because there is risk for the operator.

*Anyone having stratum problems with a client, just use getwork and the proxy (set your mining address to localhost or 127.0.0.1*

[Lucko]

If you are having problems with your miner, you should probably get those resolved first before complaining about what slush is doing.  Seriously, when my miners are stable, which is almost all the time, my payouts are constant and regular.  I've never had to worry about the score normalizing.

I'm not one that is complaining... Read posts... I just told my experiences... And I'm resolving it right now. Idiot employee drooped it because he felled the need to move it...

[digital]

a whole bunch of stuff...

I think the point is...

You may be able to make a little more by mining @ a pps pool during the beginning of slush's rounds, then coming on at the back end of a round on slush to get your reward for that block.

BUT, you won't be taking extra profits from everyone else @ slush.  You will still be getting your fair share.  As opposed to mining only right at the beginning of rounds, when you find a very short round, you actually make more than you normally would, and take a percentage of the share from the other miners on slush.

I could be way off base here, but that's how I interpreted it...

[digital]

Idiot employee drooped it because he felled the need to move it...

Well, you know what they say.

If you want a job done right, you gotta do it yourself!!! lol

[Meska]

Here is the response (in French) about the hack of OVH : http://pastebin.com/c08STp9E

Quick Translate (I'm French and a big customer of OVH) :

Hi, to change your OVH password, you need to go to the backoffice. An email is sent to the email adresse with an unique URL.
This URL is formed by 21 chars using 3 differents RAND algorythm, which every algorythim create 7 chars (7x3 = 21). After this, the client receive an email with this unique URL

[snip]

This is a procedure used since 7 years who didn't change.

[snip]

The 26th april, we (OVH) saw this problem. 2 of the 3 functions who generate the RAND number where "not really rand" (quick translate). So this "unique" URL can be bruteforced (only 1/3 of the string was RAND).
This problem has been discovered the 26th of April, at 11:03:14 (Maybe French hours, GMT +1?) and has been fixed at 12:54:13.
A patch have been developped to use a corrected RAND functions.
Some change has been made to use 2 reals new RAND functions/algo.

[snip]

After, we made a big search in our database to find if other clients have been compromised, during the last 3 years. (OVH is allowed to keep 10 years of logs, according to CNIL Agreement (Wikipedia CNIL : http://en.wikipedia.org/wiki/CNIL )). Now we (OVH) are going back to 10 years to be sure...

[snip]

(personal note) : More about the hack & Bitcoin cumminity

We found 3 ID who has been hacked, in this 3 cases, the "bitcoin" cummunity was targeted who was using OVH services. The hacker found the exploit the 23th of April, at 22h, testing during an hour. At 23h, the exploit method was working and the first OVH ID has been hacked. The day after, 2 others Bitcoin OVH ID has been hack.

We have been in contact with this clients, but the quality of the exchanges didn't help us to find the problem/security hole. One of our internal dev finally found the hole and fixed it. We certainly must find some lessons to learn in the way to speak with our customers in this kind of problems.

[snip] (Line 72 on pastebin)

It took a long time to communicate about this, because we seen it was a really "small" effect (not sure about my translation in this) but "only 3 clients", and we wanted to check completly if there was only 3 clients affected or more, before speaking about this issue...

[snip] Since Line 86, more "commercial" speaking, I'll not translate it.

Please note, I'm not working at OVH, I'm just a bitcoin user, and an OVH customer working on dedicated servers.

Edit: my translation can be "really bad", I tried to do my best as quickly as I can to share the information, this email was from the OVH Mailing List called "hosting".

[Slave2school]

That was a perfectly good translation, thanks!  Better than I'm sure google would give.  Nice to have a few inside details, but I'm still glad slush has moved on. Hopefully it reduces his workload a bit on the new servers.

[genpayne]

I think the point is...

You may be able to make a little more by mining @ a pps pool during the beginning of slush's rounds, then coming on at the back end of a round on slush to get your reward for that block.

BUT, you won't be taking extra profits from everyone else @ slush.  You will still be getting your fair share.  As opposed to mining only right at the beginning of rounds, when you find a very short round, you actually make more than you normally would, and take a percentage of the share from the other miners on slush.

I could be way off base here, but that's how I interpreted it...

So you are saying a miner working for an hour before a block is found is not actually doing anything towards finding the block?
You are contributing little work but getting a big share from the block. The block has the same value no matter how many miners are involved, so those who have been mining continually with slush have to lose out!