Brad,
(I
changed the subject to better reflect the current topic)
That
the point of what I was suggesting. The size of the Clause 22 register
area is pretty small. 10GBASE-T management would benefit from the larger space
available with the Clause 45 register addressing. Also, as you point out,
Clause 45 is probably the way of the future.
My
point is that it isn't necessarily an either/or decision. For a
10/100/1000/10G Phy (or some subset of those speeds), one could use
Clause 22 registers to control the autonegotiation and, in the cases
where the negotiation chooses one of PHYs currently covered by Clause 22, one
would use the Clause 22 registers to manage the PHY. If the negotiation
chooses 10GBASE-T, then one can use Clause 45 to manage its
operation.
The
alternative is to move all the Clause 22 functionality into registers in a
Clause 45 device (xxBASE-T PCS?) so that
everything could be managed through that register set.
Is
it better to have two different register maps to access management information
for a 1000BASE-T device depending on whether it is packaged with a 10GBASE-T
device or not, or is it better to keep one register map per PHY type and have
ICs supporting old and new PHY types support two forms of register
addressing?
Pat
-----Original
Message-----
From: Booth, Bradley
[mailto:bradley.booth@intel.com]
Sent: Tuesday, March 25, 2003 1:02
PM
To: stds-802-3-10gbt@ieee.org
Subject: RE: [10GBASE-T]
EFM OAM...
Pat,
Another area to consider is that Clause 28
auto-negotiation is written to only support interaction with Clause 22
registers, yet Clause 45 is more likely to be used as Ethernet moves
forward. This is where 10GBASE-T may need to reflect the registers
used by our previous BASE-T PHYs in the 10GBASE-T register set. Or am
I just making a mountain out of a fire ant hill?
Cheers,
Brad
Brad,
The clause 45 register addressing was constructed so that it
didn't interfere with clause 22 register access. If 1000BASE-T devices can
auto-negotiate, they could use clause 22 defined registers for addressing
that auto-negotiation mechanism and use clause 45 registers for 10 Gig
specific management.
The only conflict one would have to resolve is in the electrical
interface to the management registers. Clause 22 specifies higher voltage
than Clause 45. Given current technology, one probably would want to use
the Clause 45 electrical specification.
Regards,
Pat
-----Original
Message-----
From: Booth, Bradley
[mailto:bradley.booth@intel.com]
Sent: Monday, February 24, 2003
7:57 PM
To: stds-802-3-10gbt@ieee.org
Subject: RE:
[10GBASE-T] EFM OAM...
Ahmet,
Valid points. Some may argue that EFM OAM
didn't exist when 1000BASE-T or 1000BASE-X were written, but that has
not prevented EFM OAM from having some affect on 1000BASE-X. When
802.3ae was starting, 802.3ad (Link Aggregation) had not yet completed,
but there was a decision to support Link Agg in 802.3ae. The
10GBASE-T Study Group will have to make decisions about what portions of
802.3, 802.3ae, 802.3af and 802.3ah that we feel we should support or at
least consider supporting moving forward. For example, do we use
Clause 22 management or Clause 45 management? If we use Clause 45
but wish to support auto-negotiation to lower speeds, then we may need
to consider supporting Clause 22 access to Clause 45 registers.
802.3 is a living document, so we need to be careful about what parts we
do and do not want to consider in our effort.
As the Chair, I want to make sure we consider
all aspects of creating a standard, from the technical feasibility up to
the management requirements. All opinions are valid, and open
discussion is a great way that we make sure we haven't left stones
unturned.
Thanks,
Brad
Brad, if this isn't an invitation to get one into trouble, I
don't know what is. In any case, my input would be that we need
OAM no more than 1000BASE-T does, so I'd favor leaving out of the
scope of the 10GBASE-T PHY. If someone wants to implement some
level of OAM in a derivative PHY device or use MAC level OAM
functions they're free to do so.
Also, from my limited understanding of 802.3ah, OAM
for the PHY layer is TBD, so I don't think we can consider
the work being done in EFM as useful precedence at this
point.
-ahmet
Geoff,
Would you like to make a presentation to
that effect? :-)
I see the Study Group as having three
options related to OAM in our objectives:
1) state compliance with EFM OAM (and
therefore possibly use it in our effort)
2) state that EFM OAM is beyond the
scope of 10GBASE-T, excluding it from use within our
effort
3) say nothing, and leave the use of
EFM OAM capabilities up to those implementing the
systems
Which of the three options would you
prefer?
Would anyone else like to state a preferred
option?
Thanks,
Brad
Brad-
I
would say that
since...
the
same entity is likely to own both ends of the
link
AND
both
ends of the link are expected to be in the same
building.
AND
both
ends of the link are likely to be in the same room
that there
is no need for management beyond that required for existing
enterprise links.
Geoff
At 08:32 PM 2/18/2003
-0800, Booth, Bradley wrote:
Study group members,
As some of you may
know, EFM (Ethernet in the First Mile or 802.3ah) has added
Operation, Administration and Management (OAM) capabilities to
their specification. Like 802.3af DTE power, the study
group needs to decide whether or not compliance with 802.3ah is
within the scope of our effort, and most specifically the OAM
capabilities. This relates to compatability with our
existing standards. If there is anyone that would like to
make presentations for or against compliance with 802.3ah or
802.3ah OAM, please let me know.
Thank you,
Brad
Chair, 10GBASE-T
Study Group
bbooth@ieee.org