| 1 |
On Mon, Apr 27, 2020 at 3:07 PM antlists <antlists@××××××××××××.uk> wrote: |
| 2 |
> |
| 3 |
> On 27/04/2020 17:59, Rich Freeman wrote: |
| 4 |
> > Really though a better solution than any of this is for the filesystem |
| 5 |
> > to be more SSD-aware and just only perform writes on entire erase |
| 6 |
> > regions at one time. If the drive is told to write blocks 1-32 then |
| 7 |
> > it can just blindly erase their contents first because it knows |
| 8 |
> > everything there is getting overwritten anyway. Likewise a filesystem |
| 9 |
> > could do its own wear-leveling also, especially on something like |
| 10 |
> > flash where the cost of fragmentation is not high. I'm not sure how |
| 11 |
> > well either zfs or ext4 perform in these roles. Obviously a solution |
| 12 |
> > like f2fs designed for flash storage is going to excel here. |
| 13 |
> |
| 14 |
> The problem here is "how big is an erase region". I've heard comments |
| 15 |
> that it is several megs. |
| 16 |
|
| 17 |
I imagine most SSDs aren't that big, though SMR drives probably are |
| 18 |
that and more. |
| 19 |
|
| 20 |
But I agree - for anything like this to work it really needs to be a |
| 21 |
host-managed solution ideally, or at least one where the vendor has |
| 22 |
published specs on how to align writes/etc. |
| 23 |
|
| 24 |
-- |
| 25 |
Rich |
Archives