make:
mkdir /mnt/usb-hdd
chmod 777 /mnt/usb-hdd
sudo ntfs-3g -o big_writes /dev/sdc2 /mnt/usb-hdd
See /mnt/usb-hdd
1 Like
Thank you very much sir.
two hard drives running 150 MB/sec now.
After they are filled full I will plug into Windows array to see whether they read OK. I will report back.
Update.
-
File copy with ntfs -3g, filled NTFS hard drive plugged back into windows system. No error no problem at all.
-
ntfs -3g under ubuntu, one hard drive write at maxiuxm speed I think, 150 -200 MB/sec. Constent writing.
-
ntfs -3g under ubuntu, two hard drive write, speed total is about 200 MN/sec. That is 100 MB/sec each hard drive. Writing is on and off (hard drive light indicates).
-
ntfs -3g under ubuntu, three hard drive write, everything slows down. Very slow, total speed lower than 100 MB/sec.
My source disk is a 4TBx12 RAID 0. Total 48TB, I can see read wirte speed is above 1000 MB/sec (iotop says).
My plotting speed is 5.8 minutes each. Plot copying out canāt match plot creation.
Hmmmā¦ no idea if this is an update or a problem message.
With your setup you should probably be able to fill 2 hdd in parallel.
Test it and check in about 30-60min if your raid is running full.
About your raid and 1000mbits.
Stupid test.
When plotting cuda writes and reads from the out.
- 2x read to move the plots.
Of course your raid goes to its knees.
If you run this setup cs. 30min like this you will see the actual plot time. Probably not more 6min.
Because the read/write rate for cuda decreases dramatically.
Get a nvme. Seagate 520 is cheap and lasts forever (gpu plotting).
1 Like
Thanks for the fast reply.
I did a 200 plots CUDA plot sequence. Speed is consistent. 5.8 minute a plot regardless whether writing out to n hard drives or not. 5.8 miunute a plot stable output to the 48TB RAID 0.
I have not used ssd because I havenāt figured out how to use the PlotPLow.
Not sure it is because the external SAS array I use. It is a very old 6GB EMC STL3.
With your instuction it is already the fast speed copying out I can get. I ordered a 8088 to 4 hard drive cable. Let me see whether it helps.
Confused
What is it about plow that you donāt understand?
There are only minimal changes to be made.
Have you tried it at all?
If soā¦ what error message?
P.s. looks not expensive at firstā¦ but you plot only once. To invest a lot of money I see as unnecessary.
Thanks for the reply.
I will do the PLOW but I need to figure out the write out speed firstly. Linux write to NTFS has been problem for me. Maybe I should go all Linux on the whole chia thing. Plot dump would be no problem. I have a 1.2PB farm. Convert to all Linuxā¦ I am thinking.
Bullshit. You are going about it all wrong.
Always mount 3 hddās and check the first 3 hours whether your raid runs full.
Then let them plot full.
In attempt 2 you take only 2 hdds and check again.
It should not matter to you whether 3 hdds are written at the same time or 2.
And the way you present your skills with Linux here = you want to switch completely?
Well then have fun. But there are also more painless ways to drive against the tree 
Addendum.
I am out. Your test are for the āassā anyway.
Plow pushes the first plot with your approx. 150-200mbits. About 6min with full speed.
Then the next plot is ready and plow pushes it at the same time.
Yesā¦ your speed is reduced.
But eventually the first one is done and you push again with 150-200.
But you obviously know better.
Go ahead and learn from the same mistakes we (probably) all have made.
Thanks for the reply.
I will learn PLOW this weekend.
No idea what lord-icon is on or on about.
You do not what to be writing more than one plot to a drive at a time, you can of course move more than one plot at a time, as long as they are to different drives.
I move three plots at a time to three drives, across 10Gbps network. I could move more than that, but I need to sort a networking issue on the server - I think its related to Hyper-V virtual switch which is single threaded.
2 Likes
10Gb network. Three write to three hard drive. How fast is overall? It could solve my problem. Thanks sir.
1 Like
Iām moving Bladebit C7 plots which are 78GB in Windows, or 83.8GB on Linux, generally it takes about 8 minutes to move that plot, but when drives start to get full data transfer rates slow down, so the longest Iāve seen is 15 minutes according to my logs.
So thatās three plots moved in 8 minutes, if Iām moving three at a time.
In theory you should be able to move 5 plots at the same time to five different drives across 10Gbps network, but it will depend on the drive speed.
10G networking can be had pretty cheaply nowadays.
2 Likes
So, I assume you do Linux to Windows tranfer. Do you use SMB protocol? How to define 10Gbe card to be the sending hardware in Linux? Thanks!
I built my 1.2PB farm over some weekends in two years. My file transfer protocol is mostly NikeNet, copy to hard drive, walk to another machine and plug in.
The 10Gbe card was already installed in the PC, installed Linux Mint, and the card just worked.
I use SMB, Samba if I remember correctly, only recently setup Linux, I was running Windows 11 for my plotter, but Linux is so much quicker.
1 Like
Solved my network bottle neck by removing the virtual switch.
PS. Write speeds are slow because these are full drives.
1 Like
I had also tried your tool (@Ronski).
Ran with me rather badly. Have only VGA on n server board. And that can only max 1280px. But your tool is not designed for this. Everything wrapped and some information is not visible.
Next problem: Farming and plotting is very bad.
I did that for about 1/2 year and never won anything. The reason is: The load is too high. CHIA is all about quick responses. If then the client also has to fight with CPU/HDD/network load you will never(rarely) win anything.
From your picture you can already see that the CPU has a quite high partial load.
This is certainly not conducive to CHIa.
Besides: Why put a lot of money in hardware that is no longer needed in 2-3 months (500TB farm)!
I always have 2x18TB plotting. Takes about 2 days with 6min creation.
Now you can whine around and say: In 2 days I could already earn a little something = true. BUT: The risk for the farmer is high. Especially since there should run as little software as possible anyway. Who knows what security gaps you get into the system with additional software XY.
Addendum: The plotter can also hold an instance of CHIA. Say: Harvester 2.
The plotter then delivers the finished plots to the farmer. At some point they are full and then move manually to the client / farmer. 0% loss and FREE
P.s. I have readjusted it timesā¦ so that here times calm comes in.
Picture 1: 3 plots are created and moved by PLOW at the same time.
Result: 160-170 MB/s
Picture 2: Like picture oneā¦ now only additionally chia_Cuda started.
Result: approx. 90 MB/s
Iāve no idea how youāve tried a version of my plot mover, as the only copies are on my computer.
Iāve been running Chia ever since launch (over two years ago), my server runs other stuff as well, even with compressed plots lookup times are below 5 seconds. Yes look up times do increase when copying plots to the diskās, and thatās perfectly expected, and not too bad either. I actually used to run multiple forks in a VM on the same server and look times were still perfectly acceptable.
I bought my workstation well over a year ago, adding to it as Iāve gone along, yes Iāve spent far too much on it, but I expect most have.
Hello Ronski,
Iām also on windows and I canāt find any good software to move my plots to my hard drive, maybe you want to share it.?
THANKS