Reading speed

Discuss new features and functions
Posts: 8
Joined: 11 Apr 2018

mumen2

Zenju, you wrote to warn you if we encountered a faster synchronization than yours. That's not exactly the case, since it's not synchronization, but something gave me to think about.

In fact I find your comparison speeds amazing, but when I saw the reading speed of Everything software I didn't understand: I thought "when does it make its indexes?" when they were already made.

I did some testing on removable disks to be able to empty the caches easily:

- an HDD with 150,000 elements / 300GB. FFS : 1 minute, Everyting : 9 seconds reading + 5 indexing.
- a USB key with 90,000 elements / 60GB. FFS : 30 seconds, Everyting : 4 seconds reading + 3 seconds indexing.

FFS performance is equivalent to Windows performance, I did the tests too (asking for all folders size).

Both media are in NTFS, I suppose this have to do with NTFS indexes? What do you think about this?
User avatar
Site Admin
Posts: 7210
Joined: 9 Dec 2007

Zenju

I believe this is a case of speed vs correctness. Their directory traversal speed is impressingly fast, but it is not considering the operating system buffers before they are flushed out to the disk. Also IIRC the indexing requires admin rights for the low-level HDD access and it almost certainly does not respect ACLs. In other words, it's missing everything Windows does to the data after its read from disk (and also what Windows will do in the future).
The reading technique could easily be implemented in FFS, too, but I'm not convinced it is correct for the purpose of file sync, where correctness is more important than speed.
Posts: 8
Joined: 11 Apr 2018

mumen2

Thanks, that's clearer for me.

So Everything is based only on reading, not writing, and fine precision is not critical and neither are rights. This clearly explains why Microsoft does not use such an expedient.

For information, Everyting does not necessarily claim the right of admin at launch, we can uncheck this possibility and the speed remains the same.