>I'm talking about logical volumes, which usually map to partitions. One significant exception is VHD(x)s, which I also consider to be logical volumes.
>Hmm, on thinking about it further there may be a fundamental difference between HDDs and SSDs:

The above was my real question, even though in re-reading was not really well flushed out ;-)

>- Partitions on HDDs are fixed. They are always the same range of cylinders and sectors. When those fill up, fragmentation is inevitable
>- Partitions on SSDs are always logical. When a partition is created it may refer to a certain range of memory cells but over time that is expected to change:
> - to implement wear leveling, with or without over-provisioning
> - (possibly) to optimize performance by moving blocks between channels in a multi-channel SSD architecture
>
>IOW an SSD is a fully remappable block store, while an HDD is not.
>So I think I'll amend my rule:
>- 5/10% for any given partition on an HDD
>- 5/10% for the entirety of an SSD
>
>One likely side effect is that if an HDD partition gets full, other partitions aren't directly affected, except that the drive is likely to spend a lot of time in the crowded partition, which will slow down average access to any other partitions.
>
>If an SSD reaches these usage levels then all partitions on the drive will slow down.

Even better, my tentative change from having full RO partitions, spacey OS partitions and perhaps dedicated mostly empty scratch partitions on HDD to combined OS / Scratch partition with 30% of total SSD free is validated. Will probably switch over to special scratch device by replacing DVD with 2nd SSD - those few times I need them will be served well enough by external DVD. Less worry of havig to replace with somewhat stale backup and with mSdxc in laptop as well reminiscent of good old tower days with 2 dedicated scratch disks...

thx

thomas