Re: [EFM] RE: Pause frame usage in transport networks
Carlos,
From a service providers view, all OAM that is used by the service
provider must not be visible to the customer for the reasons that you
stated. However, I think that you are confusing the term "link" as being
used in this thread.
The "link" that you are referring to is, from the data communications view,
is at the transmission convergence sublayer of the Physical Media layer, a
part of what 802.3 would refer to as the PHY. What was being discussed
here is the functionality that exists at the Data Link layer which is part
of the customer's data link frame traffic.
Often the transmission convergence sublayer is referred to as the "mapping"
into the transmission protocols, such as T1/E1, SONET/SDH, etc. In the
case of the 802.3ah copper PHY, the "mapping" is a true transmission
convergence sublayer as defined by ITU. In the case of the other PHYs,
there is none.
In the case of 802.3ah OAM it functions at the Data Link layer, not the
transmission convergence sublayer. As such it is available to the service
provider only for "packet" type services. And for "packet" services the
OAM frames should be for the use of and controlled by the service
provider. For all other services, the OAM frames belong only to the
customer and are not accessible to the service provider.
Thank you.
Roy Bynum
At 04:49 PM 2/21/2003 -0300, Carlos Ribeiro wrote:
> > Shahram Davari wrote:
> > > Hi Siamak,
> > >
> > > What should a customer do if he wants to check the connectivity between
> > > his sites? You need to be able to pass OAM frames through the core.
>
>Sharam,
>
>I think you are confusing link-level OAM with user level network management
>stuff.
>
>Link layer OAM frames must not be acessible by the customer for several
>reasons; in short, it would open the network for various types of exploits,
>and it would do no good for the customer. The best course of action is to
>make sure that the customer will never need to know that OAM was there - in
>other words, it has to be completely transparent from the user side. Note
>that this is one of the few things that are true for both packet based and
>leased services.
>
>As for the real applications today, in packet networks it is common to
>rely on
>IP-level connectivity checks, using a tool such as 'ping' or SNMP queries; if
>you're not using IP, there are equivalents for other L3 protocols such as
>IPX, etc, that work in similar ways.
>
>
>Carlos Ribeiro
>cribeiro@xxxxxxxxxxxxxxxx