Re: I-D Action: draft-retana-rtgwg-eacp-01.txt

[Shane Amante <[e-mail]>]

Curtis,

One correction, below.

On Mar 25, 2013, at 9:42 PM, Curtis Villamizar <[e-mail]> wrote:
> In message <[e-mail]>
> Mingui Zhang writes:
[--snip--]
>>> We also should be
>>> asking whether providers and high end network equipment vendors see a
>>> need for any protocol work in this area.  So far the answer is a
>>> resounding "no".
>> 
>> Energy Efficient Ethernet (EEE) involves a great deal of protocol
>> work. Network equipments supporting EEE have already been shipped
>> around by vendors. Providers may not resist solutions that can save
>> their OPEX, only if vendors can come up with workable ones.
> 
> And supports copper only.  Providers don't use copper Ethernet.

Sure, we do ... at the [very] **edges** of our network.  

I wholly agree with you that Providers (absolutely) DO NOT use copper Ethernet in the "core" of their network, which I would define as:
a) Inter-City links, (P <-> P)
b) Peering links, (ASBR <-> ASBR)
c) P <-> PE links
d) Intra-Metro [WDM] device interconnection links, e.g.: interconnection of Ethernet switches via their East-West ring-side interfaces, (N-PE <-> U-PE *and* U-PE <-> U-PE).

For the above, providers only ever use fiber and/or WDM over fiber for such interconnections.  It's interesting to note that most of those links, (a) thru (c), are the only links over which providers run routing and signaling protocols, (e.g.: IS-IS, LDP, RSVP, etc.).  While technically possible to run routing and signaling protocols for (d), due to a variety of factors, that is not the case.  Instead, it is more common to see some form of Layer-2 ring protocol protocol, (e.g.: RSTP, G.8032, etc.), being run on those interfaces, since there is only a very simple East-West forwarding decision that needs to be made.  Regardless, I agree with your overall points and question what, if any, value is there in modifying routing and/or signaling protocols for this "core" part of the network to accommodate idle power states, given the substantial complexity in attempting to quickly *and* accurately transition from a quiescent state back to a 'nominal' state of full capacity.


I would say that we *DO* use copper Ethernet at the [furthest] "edge" of the network, which I would define as a U-PE <-> CE hand-offs, where the distances are short enough (> 100m) that it is technically feasible to do so.  (Refer to RFC4026 for definitions of U-PE to N-PE).  Interestingly enough, a substantial portion of the *N*-PE <-> CE circuits are running BGP, (think: Internet and/or IPVPN services).  But, as I stated above, the N-PE <-> U-PE interconnects use: i) fiber/WDM-over-fiber; and, ii) are statistically multiplexed in that multiple U-PE's share the same physical "uplinks" into the N-PE.  So, even if it were possible to somehow "turn-off" BGP from N-PE to CE, then I don't see that that could practically lead to power-savings on the N-PE since the N-PE's ring-side interfaces are both optical and need to stay operational/running in order to provide service to other U-PE's.

Thus, the only practical application I can see of power savings would be on copper interfaces at the deepest "edge" of the network, (U-PE to CE), but there's no active routing protocols on those interfaces.  And, although there's Layer-2 control protocols, e.g.: LLDP and the MEF's "Ethernet LMI", but I've not seen either of those achieve widespread deployment mostly because the CE devices do not support it, (yet).  But, we're the IETF, not the IEEE nor the MEF ... so, I'm not clear what the IETF would be able to work on here.

-shane