Topic: [ANN][BURST] Burst | Efficient HDD Mining | New 1.2.3 Fork block 92000 - page 601. (Read 2171065 times)

Ok, so hdd are still currently the best techonology.. if you can turn it off or on for extended periods of time, maybe energy efficiency isn't a big deal.

It's an interesting algorithm, couple thing I'm trying to figure out if someone could help me out:

How reliably can you be sure when it'll be your turn to mine?  I guess multiple people could potentially mine this block but it's just the person with the highest number wins or something like that?

Also, I mostly understand how you create the plots from reading the introduction.. how do you then verify that you pulled out the correct one and that you have rights to author the block to the rest of the network in a way that they can verify it.. seems to me like the rest of the network would almost have to do a lot of heavy duty hashing to copy the way you generated that number in the first place and arrive at the same number in order to verify it.  Must be some way around that?

the bluray storage for now is only useful for really offline stored data. this may change as soon as blueray disks become much cheaper than hdds. the main effort for facebook to be able to invest into this technology is that they have to keep really much data for legal reasons which is randomly accessed once if ever. i looked into alternative storage methods for burst almost three month back. for each block only 1/4096 is read from your plots but which part this is depends on the previous block. if you think of putting 100tb into some sort of cold storage you have to have almost instant access to "random" 25 gb within the blocktime.
for the case of blueray this means you require to read half of a disk within 4 minutes or even faster. the speed is limited to a maximum of 36 mb/s (8x) on the outer sectors and physically limited due to the facts that the medium rotates.
current holographic storage approaches look promising but there is no company releasing any products to public. eg. inphase announced to release a 300gb storage media for 180$ in 2007 and to release them in 2009. this means if the technology works it would be the perfect base to store your plots.
however, i just want to point out that hdds may not stay the only way to store your plotfiles in near future.

Soon will add this option and re-reading problem files

Yes, but Nxt has plans for a system that would enable protection against even 90% of the forging stake attacking the network.  Actually it's a fairly simple and ingenious tweak to the algorithm.. if you don't receive a block within 15 seconds (maybe to be bumped to a minute), then you don't count it's "Base Target" toward the total cumulative score of the chain when choosing the longest chain to follow.

It is true that blockchain trimming would be much trickier.. and probably involves using something along the lines of service providers, where others host the chain for you and you check multiple chains, or every few months or every year or however often you create a 'new genesis block' that includes all the current balances.  But you needing to use data internal to the blockchain means 'lite' clients are probably out of the question, unless you trust others to provide the blockchain up to this point.  Which may be doable, so, they are using economic clustering in order to provide extra protection against attack on the network.  So you can have many big businesses all choose the 'correct' chain they are running on and provide.


But back on topic:

I wasn't trying to insult Burst.. I was thinking about sticking a few BTC into it and I'm just researching my potential investment.

A couple problems with my calculations:
Turns out my calculation was one using an 'energy efficient' hard drive.. which is less reliable.  So you'd pretty much want to use one that works at 10W instead under full load.

Also, I started thinking that maybe you could frequently idle your hard drive but it is believed that's much worse on it's life than continually running it.. so you would want to keep it under full load.. at which point you actually use more energy than the Bitcoin network.  Which is a shame.. I really thought this was a cool idea and might address that problem and that it was basically the whole point of choosing proof of storage over proof of work.  I will admit that reusablity and the fact that this can be made into something that allows everyday people to use that piece of hard drive space that they don't use and get paid some tiny amount for it, is appealing.

It's possible that new and upcoming hard drives will be more efficient and fix this issue but the biggest problem is that their price is going down so much faster than their energy consumption, so this problem will likely become even worse.  And not only that but I suspect that Bitcoin miners will also continue to get more efficient with time.

What are other advantages to using Proof of Capacity over Proof of Work besides energy? 

Ones I see are:
It's pretty ASIC resistant, because if you can make better harddrives.. then you can use them in PCs everywhere.
Related to which, you don't need fancy hardware and if you get a hard drive specifically for this issue.
Hard drives are reusable for other purposes, if they don't die first.

I really do like this coin, just trying to look at it from all angles.


EDIT: WAIT.. I might see what you are saying.. I misunderstood the blu-ray thing since that specific article doesn't mention energy efficiency, but after looking around other places.. I see that those are significantly more energy efficient.  Also, could someone explain how the algorithm works?  Is it called Plotting? Scooping?  How can you only read the data every 512 blocks?  I'm assuming 1 minute blocks since based on Nxt?

I am sorry about the delay,  but I want to ask a couple questions.

First, did the original plot file get automatically deleted when it was done optimizing?
If not,  then it did not complete properly.

Second, did you use any memory optimizers with the delay function?
If you did not use the memory optimizer, did you just the delay function?
Memory optimizers can delete important cache or the delay function may have allowed Windows to delete something important.

Third, how many scoops did you see processed at once?
It has to be 2, 4, 8, 16, 32, 64, ..., 2048, or 4096.

Lastly, is every plot file you optimize coming out corrupted? I have had success with the windows optimizer so I want to figure out what factors I am not accounting for.  I tried to avoid making any changes to dcct's optimization algorithm.

Let me know if any of my suggestions correct your issue. I am also curious if version 1.1 from my previous post works better.

You do realize there were pure-PoS schemes before NXT, right?

But since we're flaming here and all, here goes equally vague statement:

In Pure-PoS schemes, consensus can be deadlocked to dangerously small number of holders.
Human nature is not always rational, some safeguards are necessary (such as PoW, or
something else of that nature).

I wont explain why is that, just head over to Peercoin thread for detailed criticism of PoS and why it needs a PoW backup.


Also, https://burstforum.com/index.php?threads/what-about-optical-media.261/

yeah, that's what I found, the windows re-plotter produces corrupted file.  it was incorrect size given stagger size etc...
localhost is fine.  I tested that with other files to make sure.

Sent deadlines != confirmed deadlines. Daemon is not synced or plot-file is corrupted.

oh pluuuze, muh superior pos sales pitch

The future of burst is not in HDD's, it'll be in robots and blurays, and that'll be replaced.

If you want to make the calculation, base it off gpu's, cause burst has no asic's yet, and when it does they won't be BAD. They will be new, better storage devices.

http://www.wired.com/2014/02/facebook-robots/

Also, please note that, if you want to mine effeciently, you could do only like 8 scoops/drive, which means that the drive is off for a majority of the time.

8 scoops would mean that drive only comes on once every 512 blocks.

NOTE: storage is expensive, so perhaps the robots will use waaayyy less power, but the home user with an on computer would always have the upper hand, as their computer is on in any case, and the hdd doesn't use extra power.

BURST miner, v1.141020
Programming: dcct (Linux) & Blago (Windows)
CPU support:  AES  SSE4  AVX  AVX2
Node: localhost (ip: 127.0.0.1)
Node: localhost (ip: 127.0.0.1)
Using plots:
Y:\plots\       files: 1         size: 1858 Gb
TOTAL: 1858 Gb

--- 21:14:39 ---    New block 29540, basetarget 2998332    ------------
*** Chance to find a block: 0.02970%
21:14:39 [xxxxxxxxxxxxxxxxxxx]  found deadline 790396957
21:14:39 [xxxxxxxxxxxxxxxxxxx]  found deadline 392940853
21:14:39 [xxxxxxxxxxxxxxxxxxx] sent DL: 392940853       4547d 22h 14m 13s
21:14:39 [xxxxxxxxxxxxxxxxxxx]  found deadline 165320872
21:14:39 [xxxxxxxxxxxxxxxxxxx]  found deadline 143340366
21:14:39 [xxxxxxxxxxxxxxxxxxx] sent DL: 143340366       1659d 0h 46m 6s
21:14:39 [xxxxxxxxxxxxxxxxxxx]  found deadline 27461602
21:14:39 [xxxxxxxxxxxxxxxxxxx] sent DL: 27461602        317d 20h 13m 22s
21:14:39 Confirmed DL: 5446372073854    63036713d 19h 37m 34s
21:14:39 [xxxxxxxxxxxxxxxxxxx]  165320872 > 27461602  discarded
21:14:39 Confirmed DL: 6091147647021    70499394d 1h 30m 21s
21:14:39 [xxxxxxxxxxxxxxxxxxx]  790396957 > 27461602  discarded
21:14:39 Confirmed DL: 6134291670940    70998746d 4h 35m 40s
21:14:39 [xxxxxxxxxxxxxxxxxxx]  found deadline 2275525
21:14:39 [xxxxxxxxxxxxxxxxxxx] sent DL: 2275525 26d 8h 5m 25s
21:14:40 Confirmed DL: 5975898902375    69165496d 13h 19m 35s
21:14:40 [xxxxxxxxxxxxxxxxxxx]  found deadline 1561449
21:14:40 [xxxxxxxxxxxxxxxxxxx] sent DL: 1561449 18d 1h 44m 9s
21:14:40 Confirmed DL: 4101564822843    47471815d 1h 54m 3s
21:14:42 [xxxxxxxxxxxxxxxxxxx]  found deadline 915567
21:14:42 [xxxxxxxxxxxxxxxxxxx] sent DL: 915567  10d 14h 19m 27s
21:14:42 Confirmed DL: 5635954431523    65230954d 1h 38m 43s
21:14:48 [xxxxxxxxxxxxxxxxxxx]  found deadline 59908
21:14:48 [xxxxxxxxxxxxxxxxxxx] sent DL: 59908   0d 16h 38m 28s
21:14:48 Confirmed DL: 3470582522731    40168779d 4h 45m 31s
21:14:53 Thread "Y:\plots\" done! [~14 sec] (1 files)
[100%] 1858 GB. deadline 3470582522731s  sdl:7/0(0) cdl:7(0) ss:8(0) rs:8(0)
...

Someone want to check my calculations?

I'm calculating that assuming Bitcoin gets 280,000,000 GH/s, then assuming it uses the most energy and cost efficient miner I could find, then it costs $2 per GH and uses 0.5W/GH.

This means that the cost to buy the equipment to power the network is $560,000,000.
Also, the Bitcoin network uses 0.5 W per GH.  In other words 280,000,000 * 0.5 = 140,000,000 W


Now I went with the most energy efficient drive after quickly looking around, I found this: http://www.tomshardware.com/reviews/4tb-3tb-hdd,3183-15.html

The HITACHI Deskstar 5K4000:  http://www.amazon.com/HITACHI-0F14697-Deskstar-5K4000-INTERNAL/dp/B00B6TMG7O

This costs you $130 for 4TB.  So $560,000,000 will buy you ~4,300,000 4TB drives.

4,300,000 drives at 6W per drive gets you 25,800,000 W

Or in other words, ~4 times more energy efficient than Bitcoin..  which is nice but considering how much cheaper it is to forge for Nxt than mine for Bitcoin.. maybe not as much potential as I initially was thinking.  Especially if new Bitcoin mining equipment truly runs at 0.19 W instead of 0.5 W as used for the above calculation.. at which it's maybe half as energy efficient as Bitcoin.

See here: http://blogs.wsj.com/digits/2014/08/15/german-startup-says-its-new-chip-halves-bitcoin-mining-energy/

So this would use half as much energy as a similarly sized Bitcoin network.

Here is my much longer analysis of the Nxt network: https://docs.google.com/document/d/1J8uhdshu9epGRrQHBaloGc4itdvuAHZDAUtNDjOhz-8/edit?usp=sharing

It's still a cool idea and hard drives are more reusable though..  idk.  I'll probably still buy a little but it's not as exciting as I was initially thinking unless he can figure out a way to use that data to store something instead of simply mine with it.


Btw, FakeAccount, you may want to read this article on RAM optimizers:
http://www.howtogeek.com/171424/why-memory-optimizers-and-ram-boosters-are-worse-than-useless/

Plot Optimizer v1.5 - j6jq fork

I have made some bug fixes since version 1.1 and added a delay feature if your available memory fluctuates with mining.  Read more at the Burst forum located at: https://burstforum.com/index.php?threads/plot-optimizer-v1-5-j6jq.268.

Download Link (includes source code):
http://bitshare.com/?f=otatujkx

So, what you really mean is "I wouldn't do RAID 0/JBOD".
A Synology's RAID5, for example, doesn't even have downtime when a disk fails. (just slowness after the bad disk is replaced and the volume rebuilds)

there is something wrong with payments @http://pool2.burstcoin.io/

both their schedule and amounts have gone drastically down 

there is something wrong with payments @http://pool2.burstcoin.io/

both their schedule and amounts have gone drastically down 

The account named burstpool2 has a balance of 165.5712568 BURST.

hi koko..
can you explain how to work "bonus time"??

thank you