Help choosing a filesystem for plots
-
This post is deleted!
-
I'm using a regular Linux distro so I can use any filesystem. But if no compression is possible, would using an non-formatted partition help? It would be the fastest way of storing sequential and read-only data AFAIK.
-
The following info is from my own personal experiences with personal computers and results can vary depending on the grade of hardware and should be treated as a guide only.
i haven't ventured into unix about burst so i cannot comment on unix / linux file systems. However, on windows systems I have found that external drives work wonderfully as extensible file system who's acronym's are referred to exFAT FAT64. Internal windows pc drives I have found run best with NTFS (since vista era). Both formatted with largest cluster sizes available in quick mode. The fact xplotter uses NTFS; I therefore transfer plots from NTFS to exFAT disks manually using a tool called fastcopy when the tool completes the plot which can be set with a nice large memory buffer to prevent or limit fragmentation of the transferred plot. I always do my best to transfer from faster to slower disks to ensure optimal contiguous file transfer. NTFS has security features but plots are already secure. USB drives generate less overhead without NTFS where possible. NTFS inside the PC because the internal hard disk controllers are more equipped to handle it with dedicated components whilst USB isn't dedicated (we all use many kinds of other peripherals on usb now) extensible file system fuse bundle can be obtained to read them on linux
-
@josem11200 Don't know if you can use a non-formatted partition, I'd assume there has to be some filesystem there. I'd probably go with Ext3.
-
I'll try non-formatted in Linux (after all, for Linux an empty partition is just a very big single file) and see how it works. If not, I can always switch to a regular filesystem :)
And ZapbuzZ, I'll keep in mind your tips when setting up mining on Windows (it's powered on around 10 hours/day so it'll do something...)
-
I wonder if Apple computer disks are formatted?
-
@josem11200 ext3/ext4 are your options. non-formatted won't be readable/writable by the system.
-
@rnahlawi In unix systems, non-formatted partitions are like fixed size files. You can read and write, but not resize. The same applies for lots of things: terminals are files, disks are files, input devices are files...
I got it working with a 1 GB plot, so I'll stick with it.
-
@josem11200 i think windows is great for plotting in general that the tools are more accelerated and / or optimized on that platform and its file systems can be used in Linux anyhow. Hey I wonder if anyone's tried Linux wine with the windows plotters gotta be worth a tinker sometime! i might just try it one day .... get xplotter working on wine 64 plot an ntfs disk then transfer the plot to a ext3/ext4 disk just for kicks.
-
@josem11200 said in Help choosing a filesystem for plots:
@rnahlawi In unix systems, non-formatted partitions are like fixed size files. You can read and write, but not resize. The same applies for lots of things: terminals are files, disks are files, input devices are files...
I got it working with a 1 GB plot, so I'll stick with it.
All the miners expect a directory path, and they need to parse the filename(s) to correctly "identify" the nonces in the file and the internal structure (stagger). If you store the nonces directly on a raw block device (accessible via /dev/sdNN, for example), the miner wouldn't know which nonces are "in the file". To get around this you create a (symbolic) link
ln -s /dev/sdNN /path/for/miner/id_start_count_stagger
and feed "/path/for/miner/" to the miner.
What you gain here is a little less i/o for metadata. I'm interested in the results, especially comparing equal-sized ext2/raw device.
