> Laxmi Prasad Botla writes:
...
> So, yes, until we've got some better monitoring
> answer (and I don't
> know of anyone working on one right now), bufmod(7M)
> and pfmod(7M)
> along with dlpi(7P) are the way to go.

Casper's membuf (replacing bufmod) is one area where
performance can be made much better but he's so far
resisted suggestions to put it on a path into Solaris.

I've also spent a weekend working on the HP-UX bpfmod
and had that functioning except for one major area:
delayed delivery of packets up a STREAM.  I've not
posted the results anywhere public because I'm not
clear on the licensing.

I've also raised some suggestions of being able to
do things like DMA directly into memory mapped in
to or from user space at the NIC level, rather than
have to copy it.

Crossbow also has the potential to aid in watering
down the packet flow being captured if the expression
to match on fits within the capabilities of the NIC's
inbuilt classification capabilities.

Darren
(..time to write a blog entry on this stuff...)
 
 
This message posted from opensolaris.org
_______________________________________________
networking-discuss mailing list
[email protected]

Reply via email to