Disk deltas in a directory different than .vmem

[UlyssesOfEpirus]

It turns out resume is slower from a compressed .vmem, probably because of excessive fragmentation.

But *REDO* files are better off compressed.

One could compress the *REDO*'s and leave uncompressed the .vmem for faster resumes.

But then when a VM shuts down they are both deleted and recreated at the next startup as compressed if the directory they are in is compressed, or both uncompressed otherwise.

Can the directory of the *REDO*'s be different than the directory of the .vmem, so they are compressed but .vmem is not?

[continuum]

Are you talking about NTFS-compression ?
In the past NTFS-compression used to be a bad idea.
Please explain ...

[UlyssesOfEpirus]

Yes. Was NTFS-compression too slow very long ago?

In an era of slow hard drives and much too fast cpu's like nowadays, shouldn't NTFS-compressed files be faster unless they grow in little increments causing excessive fragmentation?

[continuum]

In my experience using NTFS-compression simply results in a higher probabilty of corruption - only applies to sparse vmdks.

[UlyssesOfEpirus]

Any explanation why the same experience does not occur with other software that use NTFS-compressed files?

[continuum]

Well - even Microsoft warns against using NTFS-compression for storeing database files.

[UlyssesOfEpirus]

What about the vmdk file that stays constant because it is nonpersistent, is it better if it is compressed or not?

[continuum]

???
non-persistent does not mean that it stays constant.
Non-persistent means that you write into a "snapshot" or "delta" or "redolog" - however you may call it. Unlike a regular snapshot the redolog of a non-persistent vmdk will be discarded without question.

So I dont see why NTFS-compression would have dsifferent effects in this context.

[UlyssesOfEpirus]

A nonpersistent disk is made up of a file that looks like *.vmdk and the redolog files that look like *.vmdk.REDO_*.  If you write to the disk, you write to the redolog. Doesn't the *.vmdk part therefore stay constant?

[continuum · Accepted solution]

Ah I see what you are asking ...
I see the risk of higher corruption only for those types of vmdks that are expanded on the fly - while the guestOS is using it.
Thats why I would not see problems with storeing readonly files in compressed NTFS.