fivebells
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 11, 2014, 10:16:55 PM Last edit: November 11, 2014, 10:30:21 PM by fivebells |
|
I still don't understand how the deadline is calculated.. the infographic in the first post seems to indicate how plots are made. I only vaguely understand how they are later utilized though to do the mining. In this context, the details of the mining calculation themselves aren't that important, it's just that it takes substantial time to look up and process that many plots. Knowing that you almost certainly have the best deadline means that you can get a jump start on that process. Thanks for the thoughts about where to go next. I may take a crack at writing an explanation about ASIC resistance at some point.
|
|
|
|
mig6r
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 11, 2014, 10:35:50 PM |
|
Hello There have been changes on the SG Uray pool? My payments have never been as bad as the last 6 days, even when I had less hdd. I win 3 times less than the number of blocks I find. The diff has not yet increased. I am the only one to notice a decrease in performance?
|
|
|
|
1RuneKing
Newbie
Offline
Activity: 42
Merit: 0
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 11, 2014, 10:44:50 PM |
|
Sooooo... I just looked at the first post and all, so sorry if I missed it... But wouldn't someone with a spare data server layin around with 10+ TB of space have a huge advantage? Or other large data-storage entities, like Google for example?
This is an interesting idea for a coin, but it seems easy for someone with huge storage to manipulate this. Who needs an ASIC when you can buy several ginormous HDDs? LOL
|
|
|
|
pinballdude
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 11, 2014, 10:48:10 PM |
|
Why is there not a need to wait out deadlines? Is it possible to force that and make blocks, and therefore chains containing them, that have not waited out the deadline invalid?
Most of the time, the attacker is going to get the best deadline on a fork, at least initially, particularly with no "Nothing at Stake" advantage, so it would make sense for an attacker to start calculating for the next fork as soon as his own best deadline is calculated. I still don't understand how the deadline is calculated.. the infographic in the first post seems to indicate how plots are made. I only vaguely understand how they are later utilized though to do the mining. I can help with that, conceptually. The nonce will be hashed a lot, with the prior block and your burst address and the 4096 nonces it is in a group with, and the resulting number is pretty random, and unique for your account for that nonce. This number is then divided by basetarget and the resulting number is the deadline. So you need a pretty low number to begin with, even though the basetarget is in the low millions. There might be some rounding or discarding of bits, but above is the gist of it, as i understand it. So the more nonces (the larger plots) the higher the chance that you can find a low number. Only the lowest of all nonces on all computers wins a given block, so your chance of getting a given block is practically your_plot_size / total_plot_size_of_all_who_mine_right_now Of course if you join a pool, you join forces with lots other ppl and the pool will win blocks often, and share them using some sort of reward scheme (differs between pools how this is done)
|
|
|
|
uray
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 11, 2014, 10:49:34 PM |
|
...
Never really tested it out or tried to calculate it, but I've always guessed that under uray's system, more smaller addresses work better. I figure that since it only takes one deadline per id per block, for any given scenario, if part of plots it is using were plotted to another id, you'd get credit for an additional deadline in addition to the same one as earler, leading to more segmentation always yielding more credit. This is why my pools take have target deadlines, and take all deadlines under it, although people generally don't seem to like deadline limits that well.
you might be correct, and also note that reward does not linear to the best deadline submitted on my system, and also reward distribution is "skewed", thats why it has different curve between US, EU and SG each of them tuned to favor different type of miners capacity. on SG pool you might be get less payout if u are using 1000 accounts each of 1GB capacity instead of 1 account of 1 TB capacity, but on US pool its yet to be checked does thousands of small account result to better payout, while on EU pool you might not get payout at all by using 1000 accounts each with 1GB size i dont see this as disadvantage for miners, because it give them options to choose pool, each pool can have unlimited different characteristic instead of just PPLNS, PPS or Prop like on conventional pool Yes I've seen the curves. I'm not trying to say that more but lower quality deadlines would yield more, but that your best deadline on average would still be of the same quality regardless of how much you segment things out. Your best deadline on average for a 1TB plot file to one account should be on average the same as your single best deadline for 1TB made of 10 plots to different ids of 100GB each, and on normal solo mining both scenarios should be able to mine the same amount of blocks. The difference only comes into play when using a pool that takes 1 best deadline for each id, since the single best deadline for each scenario should be of the same quality, but the segmented setup allows 9 extras to be tacked on along with it. yeah I understand your point, to make it simple : (accountId+NonceNum) == ActualNonce, so statistically for (1 + 1000) which is single account with large nonce would be equal to (1000+1) which is 1000 accounts with small nonce, so it would be better to choose (1000+1) since 1000 account with bad deadline (which no problem because of no deadline limit) would be added up into cumulative reward. but the actual calculation does not end up there, there are some "distribution window" which is when pool found a block, the reward is distributed into past-blocks allocation and future-blocks allocation, what I am trying to do with this is, to get some "averaging" so that if you have large variance on deadline you still get low reward because on average your scores still bad, but if your variance is good (which is low) the reward is high. now if we try to use 1000 accounts with small plots, all those 1000 accounts will have high variance on deadlines, and what i am to achieve with curve is, if we summed up all of those 1000 accounts reward it would be less or equal than 1 account with low variance i know its complicated and i can't really proof it if it works, i am open for suggestion if anyone have better idea, the point i want to achieve is : if possible we shouldn't need deadline limit, because we are using pool to reduce variance, no matter how small your plot, you want to get stable earning, even if its very small, if we have deadline limit the payout will be irregular because in most of the blocks your deadline will be higher than limit and it does not counted and then if we does not use deadline limit, miners can just send all nonce they have and pool would be dead to verify all of those nonces, so we do need deadline limit (which what dev-pool did) or we just say it each account can only send one of their best nonce (which what my pool did) and then we have a problem again, what happen if miners would just create a lot of account with small plot for each of those account, so each of the account can send its best nonce, by doing this they can summed up their total earning. up to now, my solution is to use averaging, you are rewarded by your average best nonces up to 100 blocks, by doing this miner would only need to send one of their best nonce per block which will reduce computation on pool and (should) reduce traffic, and also very small capacity miners still be counted up every block for any deadline they have ![](https://ip.bitcointalk.org/?u=https%3A%2F%2Fi.imgur.com%2F288HjVM.jpg&t=665&c=Zdkyj47t91_Low) illustration of what i am trying to achieve, total reward of 5 accounts still be lower than single account because single account would have better average best deadline
|
|
|
|
pinballdude
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 11, 2014, 10:57:46 PM |
|
Sooooo... I just looked at the first post and all, so sorry if I missed it... But wouldn't someone with a spare data server layin around with 10+ TB of space have a huge advantage? Or other large data-storage entities, like Google for example?
This is an interesting idea for a coin, but it seems easy for someone with huge storage to manipulate this. Who needs an ASIC when you can buy several ginormous HDDs? LOL
I've got nearly 50TB mining, of which i have bought 40TB just for that purpose. I get 5 to 7 blocks or so on average on a good day, so there is plenty of blocks to go around, given there is a new one every 4 minutes. I only get half an hour of the day lol If you don't have many terabytes, just use a pool, with a pool you won't need that much to start getting paid - i don't know what is too little though. You don't want it all to end up in fees (1 burst for every transfer) When I started out, i solomined my first block with just spare disk space and a lot of patience, long before i bought the 40TB it has taken a lot of time to build a reliable farm, and the disks went online slowly as they got plotted and as i figured how to do stuff, still, between 17/09/2014 and now, i have mined 99 blocks, and gotten me 900K+ burst for the trouble. This has already paid several of the disks on paper at least (i haven't cashed any burst in yet, only spent some for giving away , donating and fountains) Beware that block reward go down 5% each month, so over the lifetime of a harddisk, the payout in burst will go down steeply. Price might go up, though, negating that effect. I will try to hold on to most of my burst until we're finished mining. If mining seems to be too much of a burden, just go buy burst at an exchange. Unless we run out of developers, this ought to be a pretty good candidate for one of the winner altcoins.
|
|
|
|
pinballdude
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 11, 2014, 11:08:14 PM |
|
illustration of what i am trying to achieve, total reward of 5 accounts still be lower than single account because single account would have better average best deadline
With some versions of gpu mining , the miner and pool negotiate (or rather, the pool instructs the miner) what is considered a deadline worthy of consideration. So a really slow miner will be allowed to send small deadlines, while a really fast miner will have a higher threshold reg. what to send. This should result in a reduced traffic, but still enough samples to figure what the power of the miner is. Given the custom deadline limit sent to the miner, and the nonces returned, the pool should be able to figre an estimate as to how much plot file has been plotted, and thus, what share of found blocks to allocate. for instance of one miner negotiates a minimum of 1000 and another a minimum of 10000 then you could "just" calculate each of the sub 1000 nonces you get with the same weight as the number of sub 10000 nonces that you would've gotten if you had allowed them to be transferred. Not sure i make myself clear, but i hope this might be of inspiration for perhaps a new miner protocol down the line - you should be able to create miner and pool that is pretty much totally fair, esp if you calculate the estimated plot size over a reasonable number of blocks. If things are dynamic enough the pool could even hike the proof-of-work difficulty to reduce traffic, and just compensate in the calculations. So you could have the pool run at a relatively stable traffic burden. You would of course have to code both miner and pool part of such a negotiation protocol, but pehaps you can be very much inspired by scrypt pool source.
|
|
|
|
1RuneKing
Newbie
Offline
Activity: 42
Merit: 0
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 11, 2014, 11:09:54 PM |
|
Sooooo... I just looked at the first post and all, so sorry if I missed it... But wouldn't someone with a spare data server layin around with 10+ TB of space have a huge advantage? Or other large data-storage entities, like Google for example?
This is an interesting idea for a coin, but it seems easy for someone with huge storage to manipulate this. Who needs an ASIC when you can buy several ginormous HDDs? LOL
I've got nearly 50TB mining, of which i have bought 40TB just for that purpose. I get 5 to 7 blocks or so on average on a good day, so there is plenty of blocks to go around, given there is a new one every 4 minutes. I only get half an hour of the day lol If you don't have many terabytes, just use a pool, with a pool you won't need that much to start getting paid - i don't know what is too little though. You don't want it all to end up in fees (1 burst for every transfer) When I started out, i solomined my first block with just spare disk space and a lot of patience, long before i bought the 40TB it has taken a lot of time to build a reliable farm, and the disks went online slowly as they got plotted and as i figured how to do stuff, still, between 17/09/2014 and now, i have mined 99 blocks, and gotten me 900K+ burst for the trouble. This has already paid several of the disks on paper at least (i haven't cashed any burst in yet, only spent some for giving away , donating and fountains) Beware that block reward go down 5% each month, so over the lifetime of a harddisk, the payout in burst will go down steeply. Price might go up, though, negating that effect. I will try to hold on to most of my burst until we're finished mining. If mining seems to be too much of a burden, just go buy burst at an exchange. Unless we run out of developers, this ought to be a pretty good candidate for one of the winner altcoins. Wow... Impressive... I may get into this, but at the current moment I don't have much storage space to spare.... Now it says HDD mining... Ik SSDs are similar to HDDs, but they have a finite number of writes before they fail. How well do they hold up to this? Are the plots constantly being written and rewritten? Do the devs have any plans to adapt this to be of use? Because right now all I see is diska space being used up, but not being used. I may get a team together, or inform the right people (I know a dev whod be interested) and see if we can adapt this to be used as a decentralized storage server. That is unless this dev team is already planning on going that way, if that's the case I'll step off.
|
|
|
|
xizmax
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 11, 2014, 11:15:54 PM |
|
Sooooo... I just looked at the first post and all, so sorry if I missed it... But wouldn't someone with a spare data server layin around with 10+ TB of space have a huge advantage? Or other large data-storage entities, like Google for example?
This is an interesting idea for a coin, but it seems easy for someone with huge storage to manipulate this. Who needs an ASIC when you can buy several ginormous HDDs? LOL
I've got nearly 50TB mining, of which i have bought 40TB just for that purpose. I get 5 to 7 blocks or so on average on a good day, so there is plenty of blocks to go around, given there is a new one every 4 minutes. I only get half an hour of the day lol If you don't have many terabytes, just use a pool, with a pool you won't need that much to start getting paid - i don't know what is too little though. You don't want it all to end up in fees (1 burst for every transfer) When I started out, i solomined my first block with just spare disk space and a lot of patience, long before i bought the 40TB it has taken a lot of time to build a reliable farm, and the disks went online slowly as they got plotted and as i figured how to do stuff, still, between 17/09/2014 and now, i have mined 99 blocks, and gotten me 900K+ burst for the trouble. This has already paid several of the disks on paper at least (i haven't cashed any burst in yet, only spent some for giving away , donating and fountains) Beware that block reward go down 5% each month, so over the lifetime of a harddisk, the payout in burst will go down steeply. Price might go up, though, negating that effect. I will try to hold on to most of my burst until we're finished mining. If mining seems to be too much of a burden, just go buy burst at an exchange. Unless we run out of developers, this ought to be a pretty good candidate for one of the winner altcoins. Wow... Impressive... I may get into this, but at the current moment I don't have much storage space to spare.... Now it says HDD mining... Ik SSDs are similar to HDDs, but they have a finite number of writes before they fail. How well do they hold up to this? Are the plots constantly being written and rewritten? Do the devs have any plans to adapt this to be of use? Because right now all I see is diska space being used up, but not being used. I may get a team together, or inform the right people (I know a dev whod be interested) and see if we can adapt this to be used as a decentralized storage server. That is unless this dev team is already planning on going that way, if that's the case I'll step off. The plots are written during the plotting process, afterwards they are only being read so there would be no wear&tear of the SSD. However, since SSDs speed doesnt really matter for burst minig it is better to use regular HDDs. Distributed storage is on the roadmap, but far in the future. I suppose you could talk to the dev about it ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif)
|
|
|
|
brizio71
Sr. Member
![*](https://bitcointalk.org/Themes/custom1/images/star.gif) ![*](https://bitcointalk.org/Themes/custom1/images/star.gif) ![*](https://bitcointalk.org/Themes/custom1/images/star.gif)
Offline
Activity: 456
Merit: 250
Blockchain Just Entered The Real World
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 11, 2014, 11:17:56 PM |
|
I'm trying to understand why with 11 TB of plot files generate just 3000-3500 BURST per day that it seems low according the BURST calculator that show me the average of 6500 bursts.
Actually I have 5 Disks connected with USB 3.0 for a total of 8 TB and 2 nfs NAS share for a total of 3TB, I have a total of 63 plot files of various dimensions between 25GB to 1TB.
All this files are on 1 miner istance on a Linux server. The server is an Intel Celeron CPU with 4GB of RAM.
Is it better to create 1 miner instance for each physical disk ?
I have to change the hardware for a faster CPU and more RAM ?
Thanks
Fabrizio
|
|
|
|
1RuneKing
Newbie
Offline
Activity: 42
Merit: 0
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 11, 2014, 11:19:27 PM |
|
Sooooo... I just looked at the first post and all, so sorry if I missed it... But wouldn't someone with a spare data server layin around with 10+ TB of space have a huge advantage? Or other large data-storage entities, like Google for example?
This is an interesting idea for a coin, but it seems easy for someone with huge storage to manipulate this. Who needs an ASIC when you can buy several ginormous HDDs? LOL
I've got nearly 50TB mining, of which i have bought 40TB just for that purpose. I get 5 to 7 blocks or so on average on a good day, so there is plenty of blocks to go around, given there is a new one every 4 minutes. I only get half an hour of the day lol If you don't have many terabytes, just use a pool, with a pool you won't need that much to start getting paid - i don't know what is too little though. You don't want it all to end up in fees (1 burst for every transfer) When I started out, i solomined my first block with just spare disk space and a lot of patience, long before i bought the 40TB it has taken a lot of time to build a reliable farm, and the disks went online slowly as they got plotted and as i figured how to do stuff, still, between 17/09/2014 and now, i have mined 99 blocks, and gotten me 900K+ burst for the trouble. This has already paid several of the disks on paper at least (i haven't cashed any burst in yet, only spent some for giving away , donating and fountains) Beware that block reward go down 5% each month, so over the lifetime of a harddisk, the payout in burst will go down steeply. Price might go up, though, negating that effect. I will try to hold on to most of my burst until we're finished mining. If mining seems to be too much of a burden, just go buy burst at an exchange. Unless we run out of developers, this ought to be a pretty good candidate for one of the winner altcoins. Wow... Impressive... I may get into this, but at the current moment I don't have much storage space to spare.... Now it says HDD mining... Ik SSDs are similar to HDDs, but they have a finite number of writes before they fail. How well do they hold up to this? Are the plots constantly being written and rewritten? Do the devs have any plans to adapt this to be of use? Because right now all I see is diska space being used up, but not being used. I may get a team together, or inform the right people (I know a dev whod be interested) and see if we can adapt this to be used as a decentralized storage server. That is unless this dev team is already planning on going that way, if that's the case I'll step off. The plots are written during the plotting process, afterwards they are only being read so there would be no wear&tear of the SSD. However, since SSDs speed doesnt really matter for burst minig it is better to use regular HDDs. Distributed storage is on the roadmap, but far in the future. I suppose you could talk to the dev about it ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif) I was just wondering because the only thing I have access to atm is an SSD, and i didnt want to use that if there was a risk for it do be damaged while mining this. In a few weeks I'll have a TB HDD to work with though ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif)
|
|
|
|
|
pinballdude
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 11, 2014, 11:23:28 PM |
|
Wow... Impressive... I may get into this, but at the current moment I don't have much storage space to spare....
Now it says HDD mining... Ik SSDs are similar to HDDs, but they have a finite number of writes before they fail. How well do they hold up to this? Are the plots constantly being written and rewritten?
Do the devs have any plans to adapt this to be of use? Because right now all I see is diska space being used up, but not being used.
I may get a team together, or inform the right people (I know a dev whod be interested) and see if we can adapt this to be used as a decentralized storage server. That is unless this dev team is already planning on going that way, if that's the case I'll step off.
The way BUSRT works, the blocks you get do not depend on the speed of your harddisk,but on the amount of space you have laid out. As long as you can read at a reasonable rate (usb 3.0 disks are fine for 4TB with my experiments, i even have some disks connected to USB not 3.0, but seems like a fast version of 2.0) So... a SSD will not get you more blocks, but it is a lot more expensive than an ordinary HDD. I think storage is a little bit in the cards reg. BURST, but i also think that it is not the top priority at the main dev, so if someone got together and made some groundwork, i guess, that would be fine). There are also some political issues with storage, but the discussion is probably worth having, check out burstforum there might be a suitable place there for such a discussion. Personally i would prefer a very secret kind of storage that made it safe to rent out harddisk space for storage bc you would have no idea, and nobody would ever know, what was stored on the space you rented out. Also noone should be able to know where their stuff was stored. So people storing and reading should pay a few burst for doing so, and people putting up harddisk space should then simply get those coins for doing so. You could then chose to use your disks for burst mining, or storage rental according to needs and prices. Perhaps just a simple sector based system with a gazonkillion sectors, then we could build file systems and shit on top of that.
|
|
|
|
pinballdude
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 11, 2014, 11:27:21 PM |
|
I'm trying to understand why with 11 TB of plot files generate just 3000-3500 BURST per day that it seems low according the BURST calculator that show me the average of 6500 bursts.
Actually I have 5 Disks connected with USB 3.0 for a total of 8 TB and 2 nfs NAS share for a total of 3TB, I have a total of 63 plot files of various dimensions between 25GB to 1TB.
All this files are on 1 miner istance on a Linux server. The server is an Intel Celeron CPU with 4GB of RAM.
Is it better to create 1 miner instance for each physical disk ?
I have to change the hardware for a faster CPU and more RAM ?
Thanks
Fabrizio
I have had bad results due to several changing factors over time - network traffic was a bit unstable and that was not good for blocks mined (i think it might've been a utorrent client i had been experimenting with, got it to run real fast, but lots other applications didn't like it) - if the computer time drifts away from internet time, the number of blocks seem to go lower dramatically until time is fixed up again - I am not entirely sure, but it seems like restarting the wallet also sometimes gets the blocks flowing again - if you have some plots on real slowly connected hdd's you might not finish reading plots until the block is actually found by someone else. on top of that i had issues specific to windows, but you will not have those issues.
|
|
|
|
xizmax
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 11, 2014, 11:29:04 PM |
|
Bought 10 ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif)
|
|
|
|
pinballdude
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 11, 2014, 11:34:44 PM |
|
I was just wondering because the only thing I have access to atm is an SSD, and i didnt want to use that if there was a risk for it do be damaged while mining this. In a few weeks I'll have a TB HDD to work with though ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif) The disk will of course be used.. once when the plot file is created in the first place. After that, every time there is a block, 1/4096 of the nonces (each sized 256Kb) that the file is made up of, is read. So after 4096 blocks you will have read in the entire file one time (on average, what nonce is read next is random) there are 1 block every 4 minutes, so in principle you are reading all the byes in the the plot file once every 4 * 4096 minutes Wether that is wear and tear beyond your personal comfort zone is.. up to you to decide.
|
|
|
|
Irontiga
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 11, 2014, 11:36:13 PM |
|
I was just wondering because the only thing I have access to atm is an SSD, and i didnt want to use that if there was a risk for it do be damaged while mining this. In a few weeks I'll have a TB HDD to work with though ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif) The disk will of course be used.. once when the plot file is created in the first place. After that, every time there is a block, 1/4096 of the 256MB nonces that the file is made up of, is read. So after 4096 blocks you will have read in the entire file one time (on average, what nonce is read next is random) there are 1 block every 4 minutes, so in principle you are reading all the byes in the the plot file once every 4 * 4096 minutes Wether that is wear and tear beyond your personal comfort zone is.. up to you to decide. Use ur ssd to gpu plot, it'll be a great boost in speed ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif)
|
|
|
|
Klarisskin
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 11, 2014, 11:37:31 PM |
|
I have 3 small farms: 4Tb, 4Tb, and 5Tb. So far pool mining delivers unimpressive results. Should I try solo mine or its futile with these farms sizes? How can I revert to solo mine?
|
|
|
|
|
xizmax
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 11, 2014, 11:42:48 PM |
|
Not enough of it, mczarnek was talking about dev support a page back. This asset has a portion of the earnings marked as dev kickback ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif)
|
|
|
|
|