FNN vs GigabitEther & Myrinet

Mark Hahn hahn at physics.mcmaster.ca
Thu Oct 25 09:50:03 PDT 2001


> However, IMHO doing this in Linux at the moment is not future-proof; apart
> from the ever changing VM subsystem, some people favour iovecs to do
> kernel-userland data passing; OTOH when the zero-copy network changes were
> introduced (around 2.4.3 or so), there were discussions about using iovecs
> - some networking guys said that they are too complex and time-consuming
> to set up and use and so they introduced yet another mechanism which is
> now specific to networking...

I think this is a bit distorted (and seemingly reactionary).  my impression
from monitoring linux-kernel is that zero-copy stuff was originally done by
people who wanted to spew large chunks of databases onto disk.  since disks
are pathetically slow, and their data is huge, they didn't mind doing things
like having huge, fixed-size vectors of the pages to be operated upon.  disk
IO is also highly constrained in its alignment, etc.  these factors were too
specialized to reasonably apply to zero-copy networking.  

for 2.5, expect a modestly modified version of the networking zero-copy stuff
to replace the old disk-oriented stuff.

regards, mark hahn.





More information about the Beowulf mailing list