[ANN][BURST] Burst | Efficient HDD Mining | New 1.2.3 Fork block 92000

[bathrobehero]

I plotted with the gpu plotter, using 100GB plots:

gpuplotgenerator.exe generate 0 0 "x:\plots" xxx 2457601 409600 4096 64 2048
gpuplotgenerator.exe generate 0 0 "x:\plots" xxx 2867201 409600 4096 64 2048
gpuplotgenerator.exe generate 0 0 "x:\plots" xxx 3276801 409600 4096 64 2048
gpuplotgenerator.exe generate 0 0 "x:\plots" xxx 3686401 409600 4096 64 2048

But with burst-miner I get different nonce amounts:

plot read done. xxx_2457601_409600_4096 299008 nonces
plot read done. xxx_2867201_409600_4096 327680 nonces
plot read done. xxx_3276801_409600_4096 348160 nonces
plot read done. xxx_3686401_409600_4096 319488 nonces


What am I missing here?

bug in miner?

Apparently 

I mean it didn't give me any errors so I'm clueless as to what else it could be.

[bathrobehero]

In real life I am managing a petabyte-sized renderfarm, and storage alone at this scale is not easy.

According to your experience , what would be the typical life span of a HDD mining burst (for example , WD green 3 TB) ? And if it's less than a year , would a HDD that mine burst be able to RMA ?

HDD mining = reading data from HDD = normal using HDD. You can send it to RMA anytime (under warranty).

That's not that simple. If your WD's load/unload cycle goes beyond 300k WD will reject warranty regardless of the duration of your remaining warranty. You can read up on how to prevent that (hint wdidle3).

[burstcoin]

one million dollar can buy 12-15 pb space depending on the infastructure. running costs about 500$ a day.
so the current 7-8 pb networksize would roughly cost 500.000$ and 250$ a day.
on this scale plotting can be done on the storagenodes themselve and maybe a bit speeded up with some gpu plotters within 1-2 weeks after deployment.

Assuming my $100M estimate for bitcoin is correct, that means that a correspondingly secure burst network would have about 1200-1500pb.  So the question is how much energy that much storage would consume.

Personally, I think 1200-1500 pb is on the low side. Consumer prices are ~$33/tb. So $100m could buy a total of 3000pb of space at that price.

As far as energy consumption, I think it's fair to ignore the cost of running the computer that the storage is attached to. Mining that takes place on end-user's desktop computers aren't using any extra energy to keep the computer on (other than the energy of the hard drive itself), and large mining operations will likely be able to be pretty efficient with their operations, with one computer powering many hard drives.

Estimates vary, but it seems that the average power consumption for a single hard drive is about 7-13 watts. They can consume more when under heavy load, but mining will typically be very low load, with only a few reads every 4 minutes (especially if the plot files are optimized with a large stagger size). So, I think it's safe to say burst mining will be on the low end, and only consume 7 watts per drive.

Let's assume the entire 3000pb network is powered by a million 3 TB drives. This makes the entire power consumption for a burst mining network with $100m in capital equal to 7 million watts.

Let's compare this to bitcoin. I'm going to use figures from the bitcoinMining hardware comparison wiki page. For the sake of easy math, let's assume that the entire bitcoin mining network is powered by AntMiner S3's. These cost $382 each, so $100m could theoretically buy 261780 AntMiner S3's. Each S3 consumes 340 watts, which would put the entire network's energy consumption at 89 million watts.

This is all just back-of-the-napkin math, but it shows pretty clearly that burst is significantly more energy efficient per dollar of capital than typical proof-of-work mining. Ultimately, it comes down to the dollars per watt that the main mining instrument consumes. Bitcoin miners (ASICs) tend to consume about 1 watts per dollar, whereas burstcoin mining equipment (hard drives) only consume 0.1 watts per dollar.

Also, I think burst mining can be further optimized for energy consumption due to the nature of the mining algorithm that only requires reading once per block. Imagine an advanced mining setup with 4096 hard drives. The plots could be specially arranged such that the miner would only have to read from one drive per block. This would allow the miner to leave all the hard drives completely unpowered the majority of the time. This setup would consume something like a 1/4096th of the power of a traditional setup. That brings the watts/dollar figure for burst down to only 0.000024414. This more optimized mining method is 40,000 times more energy efficient than typical PoW mining.

You cannot take the host running the miner software out of the equation.
And you cannot spin-down and spin-up a drive 360 times a day, consumer drives are not built for that.

But I agree on the energy-efficiency of hard-disks compared to ASIC mining;
If the plot files are laid out optimally, the disk duty cycle is 10-15 seconds per 240 second-block ~ 6%
Lower-end disks (like WD Green/Red) consume 3,5 Watt idle, 5,5 Watt under load => 23 kWh / annum with 6% duty cycle

Depending on various design decisions regarding the machine hosting the disks, the price to run a TiB for a year is at the very least 65 USD/year incl. energy and investment needed.

And this is the consumer low-end, if you are thinking of running a few racks (read: PiB) and don't want to employ staff to resuscitate failed components daily, this number can easily be in the lower hundreds. If you think of a proper monitoring..but you'd need software ppl for that.

In real life I am managing a petabyte-sized renderfarm, and storage alone at this scale is not easy.
Mining Burst is nice because fails gracefully - when a few disks or processes go down, the rest keeps on mining.

In his described setup, you wouldn't need to spin up each drive 360 times per day. Since only 1/4096th of the data is each block, he's suggesting using 4096 drives and only needing to spin 1 drive up each block. On average each drive would need to be spun up once every 11 days.

I doubt the data could be read fast enough however, so that setup is only theretical.

Putting ~80 scoops per drive in a ~51 drive setup would allow for spinning each drive up once every ~3.5 hours

[vaxman]

In real life I am managing a petabyte-sized renderfarm, and storage alone at this scale is not easy.

According to your experience , what would be the typical life span of a HDD mining burst (for example , WD green 3 TB) ? And if it's less than a year , would a HDD that mine burst be able to RMA ?

If you treat your drive correctly it should last years. Vibration, dust, vibration, steep temperature vectors and vibration are the enemies here.
With such a low duty cycle (6%, see post above) there is no reason for early failure.

It may be worthwhile to invest the 10-20% WD charges for their "Red" series, as these come with a 3 year warranty (in my country, YMMV).
I've never had any RMA denied upon "excessive usage" for SATA disks.

This link has been posted before, recommended reading - as Backblaze is implementing consumer technology at data center scale, and learns a lot along the way (albeit they don't employ a person fit in statistics):
http://www.tweaktown.com/articles/6028/dispelling-backblaze-s-hdd-reliability-myth-the-real-story-covered/index3.html

[vaxman]

In his described setup, you wouldn't need to spin up each drive 360 times per day. Since only 1/4096th of the data is each block, he's suggesting using 4096 drives and only needing to spin 1 drive up each block. On average each drive would need to be spun up once every 11 days.

I doubt the data could be read fast enough however, so that setup is only theretical.

Putting ~80 scoops per drive in a ~51 drive setup would allow for spinning each drive up once every ~3.5 hours

ahh…yes..you have like 4096x 4 TB drives, and you read a 4 TB drive in 240 seconds ? theoretical you said, yes.

[franny2tm]

Hey Burst Fans!

So this Just happened!

https://twitter.com/franny2tm/status/514337634768793600

Get voting! 

If you feel like sending me some love.....

BURST-FCNQ-MV5D-FD99-7B75N

[familiaverde]

Hey Burst Fans!

So this Just happened!

https://twitter.com/franny2tm/status/514337634768793600

Get voting! 

If you feel like sending me some love.....

BURST-FCNQ-MV5D-FD99-7B75N

On my way to vote on cryptsy , today we got 2 good news ... 1 from dev and second from cryptsy ...

Good job franny2tm

[bathrobehero]

So now sometimes I get random nonces:



and sometimes more or less accurate nonce results:



Any idea why this might be? It's driving me crazy.

[oktay50000]

I plotted with the gpu plotter, using 100GB plots:

gpuplotgenerator.exe generate 0 0 "x:\plots" xxx 2457601 409600 4096 64 2048
gpuplotgenerator.exe generate 0 0 "x:\plots" xxx 2867201 409600 4096 64 2048
gpuplotgenerator.exe generate 0 0 "x:\plots" xxx 3276801 409600 4096 64 2048
gpuplotgenerator.exe generate 0 0 "x:\plots" xxx 3686401 409600 4096 64 2048

But with burst-miner I get different nonce amounts:

plot read done. xxx_2457601_409600_4096 299008 nonces
plot read done. xxx_2867201_409600_4096 327680 nonces
plot read done. xxx_3276801_409600_4096 348160 nonces
plot read done. xxx_3686401_409600_4096 319488 nonces


What am I missing here?

bug in miner?

you gpu got 4gb of ram?

[richibichi]

Why we need cryptsy?

Instead of DOGE cryptsy is dead in my opinion.
Which new coin has big volume on this exchange? I think no one unfortunately.

[mtomcdev]

Is any white paper available on POC mining?

[bathrobehero]

I plotted with the gpu plotter, using 100GB plots:

gpuplotgenerator.exe generate 0 0 "x:\plots" xxx 2457601 409600 4096 64 2048
gpuplotgenerator.exe generate 0 0 "x:\plots" xxx 2867201 409600 4096 64 2048
gpuplotgenerator.exe generate 0 0 "x:\plots" xxx 3276801 409600 4096 64 2048
gpuplotgenerator.exe generate 0 0 "x:\plots" xxx 3686401 409600 4096 64 2048

But with burst-miner I get different nonce amounts:

plot read done. xxx_2457601_409600_4096 299008 nonces
plot read done. xxx_2867201_409600_4096 327680 nonces
plot read done. xxx_3276801_409600_4096 348160 nonces
plot read done. xxx_3686401_409600_4096 319488 nonces


What am I missing here?

bug in miner?

you gpu got 4gb of ram?

3GB (GTX 780 Ti), but the plotter only uses 2118 MB.

[yellowduck2]

Why we need cryptsy?

Instead of DOGE cryptsy is dead in my opinion.
Which new coin has big volume on this exchange? I think no one unfortunately.

whats wrong with getting to more exchange ? U mean more exchange will increase coin supply or block reward ?

People who want to sell will eventually sell. Getting to more exchange just speed up the process

[oktay50000]

I plotted with the gpu plotter, using 100GB plots:

gpuplotgenerator.exe generate 0 0 "x:\plots" xxx 2457601 409600 4096 64 2048
gpuplotgenerator.exe generate 0 0 "x:\plots" xxx 2867201 409600 4096 64 2048
gpuplotgenerator.exe generate 0 0 "x:\plots" xxx 3276801 409600 4096 64 2048
gpuplotgenerator.exe generate 0 0 "x:\plots" xxx 3686401 409600 4096 64 2048

But with burst-miner I get different nonce amounts:

plot read done. xxx_2457601_409600_4096 299008 nonces
plot read done. xxx_2867201_409600_4096 327680 nonces
plot read done. xxx_3276801_409600_4096 348160 nonces
plot read done. xxx_3686401_409600_4096 319488 nonces


What am I missing here?

bug in miner?

you gpu got 4gb of ram?

3GB (GTX 780 Ti), but the plotter only uses 2118 MB.

then why use 4096???
plot files messed up
must use:

 2048 for 2gb gpus
3072 for 3gb gpus
4096 for 4gb gpus

[yellowduck2]

In real life I am managing a petabyte-sized renderfarm, and storage alone at this scale is not easy.

According to your experience , what would be the typical life span of a HDD mining burst (for example , WD green 3 TB) ? And if it's less than a year , would a HDD that mine burst be able to RMA ?

If you treat your drive correctly it should last years. Vibration, dust, vibration, steep temperature vectors and vibration are the enemies here.
With such a low duty cycle (6%, see post above) there is no reason for early failure.

It may be worthwhile to invest the 10-20% WD charges for their "Red" series, as these come with a 3 year warranty (in my country, YMMV).
I've never had any RMA denied upon "excessive usage" for SATA disks.

This link has been posted before, recommended reading - as Backblaze is implementing consumer technology at data center scale, and learns a lot along the way (albeit they don't employ a person fit in statistics):
http://www.tweaktown.com/articles/6028/dispelling-backblaze-s-hdd-reliability-myth-the-real-story-covered/index3.html

Thank you for the insight.

Another question.

Would stagger size greatly reduce the life span ?

For example, a 2000 stagger VS a 8000 stagger. Both using same HDD. How long will 2000 stagger last VS 8000 stagger on average.

[jzhoulon]

Why we need cryptsy?

Instead of DOGE cryptsy is dead in my opinion.
Which new coin has big volume on this exchange? I think no one unfortunately.

whats wrong with getting to more exchange ? U mean more exchange will increase coin supply or block reward ?

People who want to sell will eventually sell. Getting to more exchange just speed up the process

still have lots coin in active
Market   Currency   24hr Vol   Price   24hr High   24hr Low
DOGE/BTC    Dogecoin    359 BTC   0.00000075   0.00000076 BTC   0.00000072 BTC
DRK/BTC    DarkCoin    128 BTC   0.00804992   0.00851105 BTC   0.00746422 BTC
LTC/BTC    LiteCoin    82.9 BTC   0.01068209   0.01079251 BTC   0.01050879 BTC
PPC/BTC    Peercoin    59.5 BTC   0.00380000   0.00400093 BTC   0.00370000 BTC
SSV/BTC    SSVCoin    37.0 BTC   0.00001200   0.00001618 BTC   0.00001038 BTC
LTCD/BTC    LitecoinDark    34.8 BTC   0.00001066   0.00001377 BTC   0.00000799 BTC
RDD/BTC    ReddCoin    33.3 BTC   0.00000013   0.00000013 BTC   0.00000012 BTC
XST/BTC    StealthCoin    29.9 BTC   0.00013506   0.00014550 BTC   0.00013501 BTC
UTC/BTC    UltraCoin    29.6 BTC   0.00006429   0.00010909 BTC   0.00005000 BTC
BTCD/BTC    BitcoinDark    28.0 BTC   0.01310139   0.01379592 BTC   0.01182594 BTC
NXT/BTC    Nxt    25.8 BTC   0.00007845   0.00008100 BTC   0.00007701 BTC
BC/BTC    BlackCoin    20.2 BTC   0.00015136   0.00015288 BTC   0.00014023 BTC
SDC/BTC    ShadowCoin    18.9 BTC   0.00025999   0.00029100 BTC   0.00024000 BTC
ZET/BTC    ZetaCoin    18.6 BTC   0.00003595   0.00003690 BTC   0.00002631 BTC
ANC/BTC    AnonCoin    17.9 BTC   0.00143324   0.00169322 BTC   0.00136001 BTC
42/BTC    42Coin    11.4 BTC   6.90000000   7.15000000 BTC   6.61543184 BTC

[bathrobehero]

then why use 4096???
plot files messed up
must use:

 2048 for 2gb gpus
3072 for 3gb gpus
4096 for 4gb gpus

Well, I had no idea the staggersize was meant in MB's, but thanks. But why would is start with 4096? It doesn't start with anything higher.

Check your miningrigrentals messages.

Edit: with 1024 it's still uses 582 MB VRAM.

[callmejack]

In real life I am managing a petabyte-sized renderfarm, and storage alone at this scale is not easy.

According to your experience , what would be the typical life span of a HDD mining burst (for example , WD green 3 TB) ? And if it's less than a year , would a HDD that mine burst be able to RMA ?

If you treat your drive correctly it should last years. Vibration, dust, vibration, steep temperature vectors and vibration are the enemies here.
With such a low duty cycle (6%, see post above) there is no reason for early failure.

It may be worthwhile to invest the 10-20% WD charges for their "Red" series, as these come with a 3 year warranty (in my country, YMMV).
I've never had any RMA denied upon "excessive usage" for SATA disks.

This link has been posted before, recommended reading - as Backblaze is implementing consumer technology at data center scale, and learns a lot along the way (albeit they don't employ a person fit in statistics):
http://www.tweaktown.com/articles/6028/dispelling-backblaze-s-hdd-reliability-myth-the-real-story-covered/index3.html

the backblaze cases are based on rocketraid cards.
a case to attach 45 disks starts with backplanes and controller without mainboard at about 3500$.
to run these 45 drives you also have to add some quite fast cpu and ram. my estimate is at least 5000-6000$ for 45 drives without the hdds...
this makes something about 130 $ only to be able to attach the drives.
http://www.45drives.com/products/order/dw-enclosure.php

i think of putting some of the new supermicro boards with 16 sata ports into a custom made box (not case).
http://www.supermicro.nl/products/motherboard/ATOM/X10/A1SA7-2750F.cfm
the form factor is proprietary and i am not sure which connectors to the case i require.
I have'nt checked further details so far but the board + ram + case and powersupply may end up at $800-1000$ for 16 hdds which is less than half of the backblaze system.
also the intel atom core should be able to operate all 16 hdds on its own.
all together the hardware costs for a tb ends up somewhere around 75$.

[enta2k]

Hey, my "Gaming/Working"-PC runs 24/7 with a single 6TB HDD and this uses a lot of engergy.
Does someone use a Raspberry pi, i have a B+ laying around, but it´s my first rasp i ever had.

Not shure on wich image it would run, is it possible?

Yes, it'll work, only needs java....although it'll take ages to read plots....i don't think it'll even manage to read all of them....and it's usb 2 ....so true answer is no, not with 6tb. You need more RAM and usb3 or eSata

Ah okay, so maybe i untervolt my homepc a little bit.

[Karasur]

WTF???
should find the block but BURST-M78H-JUVQ-VKJG-3H3HL stole it?

and why after 15380 miner (dcct linux) again tried to search previous block?