Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 9 Jan 2004 15:51:43 -0600
From:      "Guy Helmer" <ghelmer@palisadesys.com>
To:        "Richard Bejtlich" <richard_bejtlich@yahoo.com>, <freebsd-net@freebsd.org>
Subject:   RE: Paper on device polling and packet capture performance
Message-ID:  <FPEBKMIFGFHCGLLKBLMMEEIKCDAA.ghelmer@palisadesys.com>
In-Reply-To: <20040109171717.33976.qmail@web60804.mail.yahoo.com>

next in thread | previous in thread | raw e-mail | index | archive | help
Richard Bejtlich wrote on January 09, 2004 11:17 AM

> I was wondering if anyone read the paper by Luca Deri
> (of Ntop fame) on "Improving Passive Packet Capture:
> Beyond Device Polling":
> 
> http://luca.ntop.org/Ring.pdf
> 
> Luca makes some astounding claims regarding packet
> capture performance, with FreeBSD performing very well
> when device polling is enabled.  

Looks interesting.  I would hope the packet loss with FreeBSD
and polling could be eliminated by tweaking the HZ and
kern_frac/user_frac parameters, unless the machine is
just too slow to handle the load.

However, I think tying directly into the network
drivers isn't as general an approach as is working at
a little higher level in the system.  The network
device drivers have been modified in the past to
perform special functions (e.g., polling) but
it is more useful and less bug-prone to push general
functions into the system at other levels (e.g.,
bridging, which used to have hooks into each network
device driver).

I want to look at memory-mapped access to the BPF device.
This would preserve the existing network device drivers
while reducing mbuf copies, context switches/user-kernel
transitions, and latency.  Performance ought to be
comparable to Luca's approach, and this would also
preserve bpf filtering capability.

(If someone else has already done this, I'd love to
know where to find the code!)

> A wrote a short and probably naive synopsis for my
> Blog:
> 
> http://taosecurity.blogspot.com/2004_01_01_taosecurity_archive.htm
> l#107358025105922521
> 
> Does anyone care to comment on the paper?  (I asked
> Luca and he agreed to this post.)
> 
> Thank you,
> 
> Richard Bejtlich
> http://www.taosecurity.com

Thanks for bringing this to my attention!

Guy Helmer



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?FPEBKMIFGFHCGLLKBLMMEEIKCDAA.ghelmer>