The use of 802.1 LLDP or 802.3 OAM for power negotiation
All,
I apologize for the crossposting of this to the wider 802.1 and 802.3
groups. If it leads to long detailed discussions that are not of
interest to the broad membership then I will take it to the 802.3at
reflector. However, the issue is beyond the ken of the
power-over-Ethernet experts and needs some input from standards and
protocol geeks, to whit:
The 802.3at Task Force would like to define a mechanism for negotiating
changes in power state. I have proposed a simple state machine to handle
the request/acknowledge mechanism but the method of communicating the
states and the power objects is causing some controversy.
Proposal 1.
Extend the current definition of LLD-MED for power objects to include
the state information and extended objects as defined by 802.3at.
Advantages: Some (but not all) inline power systems already require the
use of LLDP for power information in a stateless manner.
Disadvantages: LLDP has been defined as a stateless protocol, its use
for this stateful mechanism violates a philosophical rule (potentially
opening the door for abuse...). Also, changing the definition for LLDP
in 802.1AB (10.2.4) would require opening a project in 802.1 to run
parallel to 802.3at.
Proposal 2.
Use a variation of the 802.3ah OAM protocol to transport the state
information and power objects. This would create a new slow protocol,
OAM subtype, code - as defined in 802.3 Clause 57.
Advantages: 802.3ah OAM is specifically designed for handling the
MAC/PHY management of a single link segment. The definition is contained
within 802.3 and would create no confusion with currently defined
stateless power management (e.g. TR41 & LLDP-MED).
Disadvantages: The effort may be seen as "re-inventing the wheel" or may
be taken as a snub to 802.1. Some devices might be required to support
both the 802.1 and the 802.3 protocols.
Discussion.
In my view, the definition in 802.1AB does not preclude its use for
transporting state information - as long as the state machine is clearly
defined outside the standard. I believe that we could define the power
mode control state machine in 802.3 (Clause 33) and include an annex
describing the use of 802.1 LLDP to carry the state information and the
extended power objects. No project in 802.1 should be required.
On the other hand, if there is a philosophical barrier to using LLDP for
this application then the additional burden of supporting 802.3ah will
be relatively small for a small number of devices for a short time. A
device that is sufficiently complex to support a layer 2 management
protocol can generally support 2 similar protocols with little extra
complexity. Furthermore, most of the devices that might require both
LLDP and OAM would also be required to support SNMP - at a greater level
of complexity.
Hugh.
---- IEEE 802.1 Email List ----
DELETE THIS FOOTER from copies, forwards, & replies.
Working Group 802.1 Web pages:
http://www.ieee802.org/1/
List subscriber pages:
http://www.ieee802.org/1/email-pages/vnpnn806.html