Re: nfe driver 6.2 stable
- From: Pyun YongHyeon <pyunyh@xxxxxxxxx>
- Date: Sat, 29 Sep 2007 10:28:01 +0900
On Fri, Sep 28, 2007 at 07:05:35PM +0100, Chris wrote:
On 28/09/2007, Pyun YongHyeon <pyunyh@xxxxxxxxx> wrote:
On Thu, Sep 27, 2007 at 06:30:20PM +0100, Chris wrote:
> On 27/09/2007, Pyun YongHyeon <pyunyh@xxxxxxxxx> wrote:
> > On Mon, Sep 24, 2007 at 04:44:09PM +0100, Chris wrote:
[...]
> >
> > > The card is more stable when using nve driver no crashes still but the
> > > performance is around the same maxing out at around 200mbit. I was
> > > expecting nearer 400-500mbit.
> > >
> > > e1000phy0: <Marvell 88E1116 Gigabit PHY> on miibus0
> > >
> >
> > There had been issues nfe(4) with 88E1116/88E1149 PHYs but your issue
> > seems to be different one. How about manually set media configuration?
> > For example, "ifconfig nfe0 media 1000baseTX mediaopt full-duplex"
> >
[...]
>
> I typod I meant to say nve was unstable and nfe has been rock solid so
> great work with the driver. I enabled mpsafe now and still no
> problems and its gave me a little more performance as well, a shame my
> nic has no hardware features (unusual for a gigabit nic) but overall
> glad its at least stable.
>
How did you check network performance?
Maxing out at around 200Mbps seems weird. Personally, I had never seen
GigE hardwares that saturate at 200Mbps. One of causes I can think of
is speed/duplex mismatches with link partner. Manually setting
speed/duplex might fix your performance issue, I guess.
--
Regards,
Pyun YongHyeon
These are ftp transfers it is running on a amd62 x2 dual core
processor and gig of ram, when at max speed cpu usage is very high in
excess of 80% but not completely maxed out. It now seems to be able
to sustain around 30meg/sec the highest I have seen I havent done any
other testing so if you have a better way it would be good to know
ftp transfers involve disk activities so you're not measuring NIC
performance. Try one of benchmark programs in ports/benchmarks
(e.g. netperf, iperf, ttcp etc).
ATM nfe(4)'s interrupt moderation mechanism doesn't seem to work
at all so nfe(4) generates too many interrupts. However I don't
think it wouldn't be major bottleneck of the performance.
thanks. Please bare in mind I have no local access to the server so
crashing it eg. is expensive for as will have to pay for a kvm switch
netstat -i indicates no collisions for a duplex mismatch so not keen
on trying a manual negotiotian again for reasons above. So even for a
card with no hardware features you would expect it to exceed 200mbit
easily?
Yes. You have a gigabit ethernet controller and fast CPU.
Run one of benchmark programs and get a number.
I havent enabled net isr setting I wonder if that will help.
That wouldn't help a lot, I guess.
Chris
--
Regards,
Pyun YongHyeon
_______________________________________________
freebsd-stable@xxxxxxxxxxx mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "freebsd-stable-unsubscribe@xxxxxxxxxxx"
- Follow-Ups:
- Re: nfe driver 6.2 stable
- From: Chris
- Re: nfe driver 6.2 stable
- From: Jos Backus
- Re: nfe driver 6.2 stable
- References:
- nfe driver 6.2 stable
- From: Chris
- Re: nfe driver 6.2 stable
- From: Jeremy Chadwick
- Re: nfe driver 6.2 stable
- From: Chris
- Re: nfe driver 6.2 stable
- From: Pyun YongHyeon
- Re: nfe driver 6.2 stable
- From: Chris
- Re: nfe driver 6.2 stable
- From: Pyun YongHyeon
- Re: nfe driver 6.2 stable
- From: Chris
- nfe driver 6.2 stable
- Prev by Date: Re: Progress with usability of AMD64
- Next by Date: Re: nfe driver 6.2 stable
- Previous by thread: Re: nfe driver 6.2 stable
- Next by thread: Re: nfe driver 6.2 stable
- Index(es):
Relevant Pages
|
|