Implement Ethernet Interface Properties

[Spencer Ku (古世瑜)]

> On Wed, Dec 1, 2021 at 1:59 AM Spencer Ku (古世瑜)
> <Spencer.Ku at quantatw.com> wrote:
> >
> > Hi All,
> >
> >
> >
> > We want to implement two network Ethernet interface properties
> “Bandwidth” and “Drop package rate”, and expose the information via
> Redfish. The goal is that clients can talk to bmc via Redfish and get those two
> information. It will improve the debuggability of BMCs.
> >
> 
> The only thing I would note in your plan is that Redfish generally doesn't
> expose anything as a "rate", it exposes it as a count (packet count, drop count),
> and lets the client determine whatever sampling rate they want.  I suspect
> you should do the same.

Thanks for your explanation! I will follow the feedback to I implement this feature.

> 
> Also, I'm assuming "drop package rate" was a typo for "dropped packets rate"?

Yes , it should be "dropped packets rate", sorry for the typo.

> 
> >
> >
> > About the implement method, we divide the work into two parts:
> phosphor-network(or other suitable service) part, and bmcweb part.
> >
> > For phosphor-network part, we propose to use existing tool like ipert to get
> the Ethernet BW/Drop package rate. Or the other method is that read TX/RX
> from ethernet driver (/sys/class/net/{Ethernet_Name}), then calculate the
> BW/Drop package rate. After that, we will need to register the D-Bus interface
> and sync the property info to that D-Bus.
> >
> >
> >
> > For bmcweb part, as I know so far, there has no suitable field under the
> EthernetInterface to set BW/Drop package rate. The idea is that read the value
> from D-bus then set the properties under the EthernetInterface OEM field.
> Please let me know if I have missed anything.
> 
> Considering that not all NICs will have this telemetry, it probably makes sense
> to have it on a separate dbus interface.

Understand, I will notice this part when I register the D-Bus interface.

> 
> >
> >
> >
> > We are willing to see any suggestions and alternatives as well. Thanks!
> >
> >
> >
> > Sincerely,
> >
> > Spencer Ku