[Btrfs-devel] btrfs and Solid State Disks (SSD).

[Claudio Martins]

On Monday 11 February 2008, Miguel Figueiredo Mascarenhas Sousa Filipe wrote:
>
> >From what I read, current SSD are characterized by:
>
> - poor performance in random writes (because of block erasure)
> - require wear leveling, even those that emulate sata/ide/scsi disk
> with onboard wear leveling logic.

 I thought that SSDs that emulate scsi disks (like USB pen drives, etc) and 
also CompactFlash and SD cards, had no need of additional wear leveling since 
they have, as you say, onboard wear leveling logic because they were 
basically designed to work well with FAT16/32.
Though I'd also really like to hear if anyone on the list has more information 
about this problem.

> - excellent seek latency
> - excellent read (random and sequential) performance.
> - good at sequential writes.
>
> I've read that journaling file systems are usually bad for SSD because
> of (from what I suppose are) two things:
> - increased "random" write load (journal + proper data)
> - write hot spot on the journal, causing lots of write cycles on a
> given set of blocks.

 From what I understand of the design (please correct me if I'm wrong), btrfs 
shouldn't have much of a problem in this particular area, since there really 
is no specific journal area. When something is changed (data or metadata) the 
data is not rewritten in place. Instead, new blocks are allocated and then 
the tree is updated all the way up to the root node.

 Best regards

Cláudio Martins