readv: parallel or sequential?

[Peter Schuller]

> Yes, but how can it now that it is on a RAID0 and taking advantage of
> multiple spindles instead of making it worse?
> The FS has to do sensible things for single spindle as well.
> And normaly disks are fastest when reading linear and with disk read
> caches this doesn't even have to be interleaved.
> I don't see any potential for parallell access within the same file
> beside some special constructed cases maybe.

I strongly disagree here. I would expect the operating system to not serialize 
all I/O to the same file descriptor/file, if done in paralell from multiple 
threads or through AIO. I have at least on use case heavily dependent on 
this - does this mean FreeBSD would not be usable for this?

Not to mention things like databases (e.g. high concurrency I/O in PostgreSQL 
or whatever). You most definitely want concurrent I/O if supported by the 
underlying device.

Are we saying here that I/O scheduling happens at a just below the file 
system, rather than at the individual device level? I would have assumed that 
a particular geom class would just forward (modified) I/O requests to the 
underlying device, such that when the request reaches a device actually doing 
I/O to some external entity, it can intelligently schedule in a way that is 
optimal.

Or am I misunderstanding something in this discussion?

-- 
/ Peter Schuller

PGP userID: 0xE9758B7D or 'Peter Schuller <peter.schuller at infidyne.com>'
Key retrieval: Send an E-Mail to getpgpkey at scode.org
E-Mail: peter.schuller at infidyne.com Web: http://www.scode.org

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 187 bytes
Desc: This is a digitally signed message part.
Url : http://lists.freebsd.org/pipermail/freebsd-fs/attachments/20071220/0bdf49de/attachment.pgp