On Thu, Dec 17, 2009 at 03:54:54PM +0100, Robert wrote:
quoted text
> David Gwynne wrote:
> >let's not wave our hands around in the air for no good reason. surely there
> >are more important things to get upset about.
> 
> I have to object here.
> 
> If I understood the original article correct this would mean that
> the writing of 1 sector could in fact be 2 reads + 1 write
> (internally in the disk), which would be a *serious* performance hit
> (we are not talking 5-10% here...).
> 

How can a device with a 4k internal blocksize need 2 reads + 1 write to
write a 512byte block. It will be 1 read + 1 write max. Now if you can
figure out how often 512byte blocks are written in OpenBSD then you could
actually figure out how bad the situation is.
Hint our ffs parameters default to 16KB blocks and that's what the
buffercache is queuing towards the disks. So I guess in > 99.5% there will
be no partial write to a 4k block.

quoted text
> I don't know about the effect on RAID controllers that you
> mentioned, but I guess their cache helps to hide those effects and
> also they are not in use on every PC.
> 
> The affected SATA disks on the other hand will be in use on
> basically every new i386/amd64 system in the near future and I don't
> know how much their internal cache will help. I see a big problem
> coming up... ?
> 

Why should we care about the last few percent of performance on a desktop
PC. Seriously it will be fast enough even for HD porn.
In the end the pathetic seek times of desktop SATA disks will be the most
limiting factor when you hit them with nearly random access patterns.

-- 
:wq Claudio