My 1/2 PetaByte



  • @kmaxkmax Yeah. me too. i just wouldn't know how to dish it out as an asset. And i've already been super busy. I've put 200 video cards online in the past 90 days. and now this petabyte setup is going to take another 30 days to get running. I'm open for ideas. But i don't want more stuff to manage and take care. I already have a lot on my plate with keeping these 200 cards running for myself and others.



  • @ChuckNorris Wow, that's beauty !! What Hard disks you have chosen for this? we know you would be busy, if you could get some time give us some more info on machine and HDD setup.



  • @ChuckNorris said in My 1/2 PetaByte:

    Still setting up
    Currently plotting

    https://ibb.co/h82hAv

    I'm insanely jealous. Excited for you though. Way cool!



  • Oh dear!!



  • Hello Guys. Sorry, as you can imagine i've been very busy. Here are some numbers and info

    @natsaba They are seagate 8tb crappy old cheap as possible beautiful money making drives 8tb for like 180 each

    I have 4 dedicated burst miner and my burst server plotting 2 drives at a time. I can plot those 2 drives in less than 20 hours. I am currently filling 80tb of drives a day at this point. I'm in full production!! : ) Cough. 80 freaking terabytes a day.
    Each Burstnode will be 250tb.

    Burst1 i tested today with Jminer. It's almost filled up. 200tb so far
    200tb in 100 seconds using jminer.
    Unoptimized 7.4tb single plots
    Optimizing is going to take forever.
    Is says my read speeds are 470mb/s with jminer. That seems a little low to me.
    Gtx 1070 seems to be the best card i've tried to plot with. I get 27500 nounces per drive when plotting two instances. I've tried rx470 8gb nitro+ rx580 msi gamingx. with like 17k nounces doing dual drives.
    Doing 7 drives per usb 3.0. Each 3.0 hub is on it's own memory controller with pcie expansion card

    Can some one answer these questions for me.

    1.Will my read speed of 470mb/s or so go up after the plots are optimized?
    2.Also would an i7 help any? Does jminer use cpu also to count faster? i get heavy cpu usuage. 95-100 percent when jminer counts.

    1. Since i used a stagger size of 10000 for all the drives. Jminer uses a lot of memory. I get out of resource errors with my 16gb of ram. Should i limit with chunks or get more ram?


  • cool are you sure thats massive setup goodjob~



  • @ChuckNorris I strongly recommend you to optimize it... I can't really predict how fast it will be with optimized plots but if each of that 4Tb drives was optimized you would only need to read 1Gb of the plot, and I'm not sure but I think that with the number of scoops you have, you need to read over 1Tb, so expanding RAM should not be able to solve your problem, maybe delay it ;D

    I recommend you to use the CPUs you have to optimize those drives and don't forget you will need the double of space for optimize it while you don't finish...
    There is also a way of generating plots without need to optimize them, if you use the same number of nonces to plot as your stagger, because that way your plot will have just one scoop, making you only need to read a 1/4096 part of your plot...
    Hope it helps you ;D



  • East Germany goes around 0.15-0.20$ per kWh

    @Blago this cheap power i heard at St.Petersburg as well (what is almost EU) :-P



  • @ChuckNorris said in My 1/2 PetaByte:

    stagger 10000

    Your files have a stagger of 10000, that means

    10000 * 64 Bytes == 640 KBytes can be read sequentially, seek to pos + ((640KBytes*4095)), repeat until end.

    You should definitely optimize your file if you want to increase read-speed per disk.
    What that means for your overall performance - i can't say. Do it for one of the miners.

    With a plot size of 7.4TiB you will have a stagger of

    7.4 TiB / 4096 / 64 Bytes == 31037849

    and that gives you a sequential read for

    31037849 * 64 Bytes == 1.986 GBytes

    Your disks will read much faster as A LOT LESS seeks are performed.

    (read ~2 GBytes in 640 KByte chunks: (31037849 * 64 ) / (640000) == 3103 seeks PLUS metadata lookup
    time per seek: 10 ms
    3103 * 10 ms == 31 Seconds just for head movement)



  • @vaxman @gpedro Very helpful guys. I was already going to optimize of course. There was never a thought that i wasn't. I was just saying i just now got my drives buffered. And no. I can't write direct. It's 4 times slower. there is an explanation if you want it.

    It's nice to see all the math tho. Thanks for that. I'm still filling up drives. About 80tb a day at this point. I'll be ready to throw the petabyte online within the next week. Then as it's mining optimize the drives over the course of another month.



  • Wow, that is a beautiful farm you got there. I am currently looking to build a 100TB farm but the external hard drives are quite expensive here in Estonia. About 30$ per TB is the cheapest one and this is with 20% discount i get thanks to my friend.


Log in to reply
 

Looks like your connection to Burst - Efficient HDD Mining was lost, please wait while we try to reconnect.