Hi All,
I will be spinning up a new node on my existing device running a few other nodes (Win10).
I was curious if there are any formatting settings (allocation size?) that I should do. I know NTFS is recommended for Windows but is there anything else?
Hi!
Perhaps the most popular suggestion would be
do not use Windows and NTFS, if that is not a server version or perhaps do not use it at all and use Linux/FreeBSD instead without using of NTFS
However, you may optimize your NTFS:
See also
Formating just use “default” i recommend.
if You can, use windows 11 instead of 10. It can be held in check much easier than win10.
Win10 can ignore rules because he wants to, and F the user.
Thisiswin11 at github can keep Your win11 intact with few clicks.
(its marvelous, go donate the author in mass pls)
It prevents unnecessary CPU% activity.
There are tools for win10 too, like You type in powershell:
iwr -useb https://christitus.com/win | iex
And Your win 10 will launch a great tool for win10 optimization, but even this didn’t help me to tame some of my win10 instances, and crazy CPU outbreak to 100% im facing right now. im planing to migrate to win11, i already tested a VM with it, and it behaves 2 times better.
Also You make sure You format the hdd under the windows Your are going to use with it. Mixing especially between disks that saw win7 can F’up a LFS ($LogFileformat from 2.0 to 1.1) and cause potential performance problems, and You need performance for storj’s milion small files)
(garbage collector, filewalkers, need to move fast over files, and thats up to OS to allow disk to do that, i did too many test on that)
Also on device manager You can set disk for “Better Performance” because that tells windows to keep LFS 2.0 and not downgrade it to 1.1 (without checking that write cache checkbox, You don’t need that)
Also there is a “NtfsDisableLfsDowngrade” (downgrade, not Upgrade!)
You can find it in registry under local machine and set for 1 just in case.
because when You turn off Your windows 10, it downgrades disk for LFS 1.1 so it can be recognized by windows 7 and older, You don’t need that if You don’t plan to switch disks between machines.
And speaking of which, if its 10, You need to disable “Fast Forward” in Control Panel
And if everything is set up,
Later, you can comapre disk performance on storj’s files read’s speed,
if somehow You feel its slow, Windows Drive optimizer build in tool, can help, if You click on drive and click just “analyze”,
it will scan the drive rather fast (minutes), and can unlock “hidden powers” of performance for the disk, (no need for “optimize” button, its much longer) for real, go check, if You will have some TB of storj files already. You may need to redo that after every disk dismount, or a machine restart to remind windows the performance again (sad but true for win10).
Next You want MFT optimization, by for example UltraDefrag 7.14 or newver (but newer is paid)
It also makes wonders. But takes hours. Best if You could reserve some GB for “MFT zone” (some defrag programs may do that) for like 70GB at least if its 16TB drive.
You can check Your MFT zone size under CMD with admin rights:
fsutil fsinfo ntfsinfo x:
x: is Your drive letter
Next You want periodically defragment data base files, If You keep it together with "D:\storj" folder, (the one with blobs), but thats only effective if You have already a few TB and months old.
Yea I’m aware about the windows not to use, but sadly I am hard locked to it (specifically windows10) for this device. In the future I will make a linux box, but I can still add another drive or so into the current machine so I’ll be sticking to it.
Defrag will be turned on - thanks!
Indexer, 8dot3, accesstime are all good points.
And of course, 1 node per disk, with DB on a separate SSD!
I don’t see this option sadly. The rest I will look into especially the MFT optimization.
But like I said before, I am hard locked to win10 at this point sadly
Lots of RAM or spare ssd space?, UPS? If yes-> Primocache
That’s a nifty diagram! Sadly I maxed out the ram at 32gb ddr3, and no SSD :c extra
If I ever get my current 2 nodes cleaned out + max out the next 2 16TB drives I’ll look into adding a pci card for SSD cache
You also need to perform it…
You may also just enable the write cache (both checkboxes), but only if you have a managed UPS, otherwise you must disable the write cache and do not use a Primocache, otherwise you’re risk to lose your node after the nearest power cut.
