Tweak Win 10 NTFS by disabling 8dot3 and stripping 8dot3 - HUGE PER GAIN


Recommended Posts

There are performance and security benefits to disabling short file names (or 8.3 file names) on Windows. From the graph below, you can see the performance benefits of disabling the 8.3 short file names. On Windows 8/2012 serv, the 8.3 naming is disabled when you format a new volume. When you’ve upgraded from previous versions of Windows, it’ll inherit your previous settings.

 

8484.image_thumb_0763777C.png

 

Run cmd with admin privileges:

fsutil behavior set disable8dot3 1 OR  fsutil behavior set disable8dot3 3

(1 = all volumes / 3 = all but system colume)  - then reboot.

 

Run cmd again with admin credentials and type:

fsutil 8dot3name strip /f /s C:

Repeat strip for every drive/partition.

 

 

NTFS keeps the filenames inside the MFT. http://en.wikipedia....ster_File_Table

The links within the MFT are organised so the long filenames form an index based on a b-tree. http://en.wikipedia.org/wiki/B-tree

To find any long filename an efficient algorithm called a binary search (can also be called binary chop) is used.

All very well and good so far. Now the trouble with 8dot3names is that they are not part of any index, they are just another MFT attribute.

To find an 8dot3name requires a sequential search of the MFT records that belong to the relevant directory  - leading to I/O mess.

Even if the numbers above are true, the performance gain will be negligible in real-life operations. One does not create 100K in a 1M files directory on a regular basis. Let alone a fraction of that. Same reasoning goes for directory enumeration.

All in all, this tweak might break more than it fixes.

18 minutes ago, Raphaël G. said:

Even if the numbers above are true, the performance gain will be negligible in real-life operations. One does not create 100K in a 1M files directory on a regular basis. Let alone a fraction of that. Same reasoning goes for directory enumeration.

All in all, this tweak might break more than it fixes.

I agree. Whilst I appreciate the efforts to share knowledge, if you want to demonstrate a benefit, try to provide a real-world example as this one is grossly exaggerated. Maybe do some load time benchmarks, or something, even if synthetic.

Quit pushing this snake oil... you were told many years ago that the performance gain is minimal and does nothing for the end user but cause problems down the line. Your screenshot is also misleading because those results are from Server 2012, not Windows 10.

 

13 hours ago, hardbag said:

I think 8dot3 has been disabled since WIn 8/2012...

It might be disabled OOB on the Server editions but on 8, 8.1, and 10 it is enabled. This is pure snake oil at this point.

 

  • Like 2
  • Thanks 1

What utter BS. The possible (and I mean vaguely possible) gains are so minute compared to the possible (and I mean likely if you run any legacy software) problems.

 

Stop spreading fud!

  • Like 2

Hello,

 

If you are going to do this, check all your applications carefully to see if anything is broken.  It is possible that a modern Windows application may still have a setup or uninstallation program that makes use of short filenames, etc.

 

Regards,

 

Aryeh Goretsky

 

Hups, on topic title I meant "PERF" gain :D not "PER" whatever that is.

 

It's not snake oil. It wasn't that either back in 2011 and certainly isn't _today_

 

If there is such a stuid app that wholesomelly relies on 8dot3 to keep gum in glue and vice versa (like Office was a decade ago :D :D) it's time to face the music. That would mean that those apps don't work on server... Perhaps not.

 

Tech ARP - ED#179 : Why It's Really Time To Ditch (And Strip) The ...

https://archive.techarp.com/showarticle53b4.html?artno=827&pgno=1

... well creating a lot of files isn't that uncommon. A regular front end project usually creates 10-20 000 files on npm install and it was significantly worse in previous npm versions.

I would say stick with the default option for newly formated partition but fs tweaking can give gains. For example noatime/relatime was common change in fstab under linux (also barriers, log sizes etc).

17 hours ago, hardbag said:

Hups, on topic title I meant "PERF" gain :D not "PER" whatever that is.

 

It's not snake oil. It wasn't that either back in 2011 and certainly isn't _today_

 

If there is such a stuid app that wholesomelly relies on 8dot3 to keep gum in glue and vice versa (like Office was a decade ago :D :D) it's time to face the music. That would mean that those apps don't work on server... Perhaps not.

 

Tech ARP - ED#179 : Why It's Really Time To Ditch (And Strip) The ...

https://archive.techarp.com/showarticle53b4.html?artno=827&pgno=1

Then provide us with modern benchmarks that show a performance gain. This article you keep regurgitating is from 2014, far from what we use today. If you can provide meaningful benchmarks I'll quit calling it snake oil, until then this is nothing but FUD spread by you.

Snake oil...as CircaFlex pointed out.  You're more apt to break something vs. having any real world performance increase.

 

I thought this was dismissed years ago?  Just don't.

 

Edit:  ... and OP posted a similar topic in 2011...and got roasted.  

 

  • 1 month later...
  • 2 months later...

This may have had some albeit small effect back when we all ran with HDD's for boot/data disks but now we use SSD's there really is no gain to doing this so lets just resign this tweak to where it now belongs which in a shoe box at the back of the cupboard out of sight and out of mind

This topic is now closed to further replies.
  • Recently Browsing   0 members

    • No registered users viewing this page.