[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20130419134925.4981bdfc@nehalam.linuxnetplumber.net>
Date: Fri, 19 Apr 2013 13:49:25 -0700
From: Stephen Hemminger <stephen@...workplumber.org>
To: David Miller <davem@...emloft.net>
Cc: eric.dumazet@...il.com, gregkh@...uxfoundation.org,
netdev@...r.kernel.org
Subject: Re: [RFC 1/2] netmap: infrastructure (in staging)
On Fri, 19 Apr 2013 16:38:31 -0400 (EDT)
David Miller <davem@...emloft.net> wrote:
> From: Eric Dumazet <eric.dumazet@...il.com>
> Date: Fri, 19 Apr 2013 13:31:23 -0700
>
> > On Fri, 2013-04-19 at 13:16 -0700, Stephen Hemminger wrote:
> >
> >> I can not get line rate output with pktgen on existing kernels today.
> >
> > I have no trouble saturating at line rate with pktgen, and using
> > multiqueue NIC.
>
> +1
>
> > I really hope we do not use pktgen as an argument for having netmap in
> > the kernel.
>
> Me too.
I get 7Mpps (single queue) with ixgbe and pktgen.
Easily hit 14.8 Mpps (single queue) with netmap.
The real problem is that DPDK and netmap can do multiple packets per request
to driver. Right now their is one PCI bus transaction per packet with current
driver model.
But, I am not convinced that netmap is the right solution either, this is purely
an RFC to get some attention on doing better at small packet performance.
If you look at netmap right now, it has really ugly BSD wrapper code
and does lots of assumptions and bypassing of network stack; ie. fugly.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists