svn commit: r271946 - in head/sys: dev/oce dev/vmware/vmxnet3 dev/xen/netfront kern net netinet ofed/drivers/net/mlx4 sys

[Adrian Chadd]

On 15 September 2015 at 11:34, Hans Petter Selasky <hps at selasky.org> wrote:
> Hi Adrian,
>
> On 09/15/15 19:13, Adrian Chadd wrote:
>>
>> I know all of this. What I'm asking is different - what about the
>> change(s) that broke/fixed the Xen network performance actually caused
>> the change in behaviour? I know things are sensitive to how the mbufs
>> are broken up - I'd like to see exactly this.
>
>
> You need to add statistic counters to the network drivers in question, how
> frequently packets are defragged and/or dropped in the TX path to figure
> that out. Can we continue this discussion elsewhere?

Sure thing.

Yes, maybe having a per-interface histogram of mbuf / segment sizes
would be good - something we can run live on a box rather than trying
to use dtrace. I wonder how difficult it'd be to slip something in as
part of if_transmit and ifnet..



-adrian