|
Posted by anoop on August 19, 2006, 12:32 pm
If you were Registered and logged in, you could reply and use other advanced thread options
Thomas Bahls wrote:
> Having fetched the draft 6.1 version of IEEE 802.1ag and trying to
> understand it, I ran into some conceptual questions I cannot answer
> myself. Maybe there's someone out here able to help?
Let me give it shot. :-)
> (1) 802.1ag is set out for end-to-end management and I would assume IEEE
> to make things as backwards compatible as possible, that is, allow
> 802.1ag-enabled clouds to be connected via 802.1ag-agnostic older
> equipment. I would have believed 802.1ag frames just traverse such
> agnostic equipment transparently. (Maybe this is wrong. Then, my problem
> had vanished, of course.)
> Now seeing they use dstMAC==01-80-C2-00-00-0x I have doubts about my
> guess as I happen to remember such MACs MUST NOT be forwarded at all, per
> 802.1D. So what about backwards compatibility and 802.1ag-agnostic
> equipment then?
If you pay close attention, you will see that the addresses used for
CFM are not 01-80-C2-00-00-0x, but 01-80-C2-xx-xx-xy. This means
that they use the IEEE OUI but are not from the set of reserved
addresses that bridges shall not forward. Therefore these frames
would pass transparently through legacy equipment.
> (2) Again addressing. It appears to me that the ME level shall be
> signified in the last three bits of the MAC address, with ME level from
> 0..7. That'd make 01-80-C2-00-00-00 to -00-07, which I believe to conflict
> with STP, PAUSE/.3x and LACP. What's wrong or missing here (don't believe
> IEEE to make conflicting specs ;-)?
See above.
> (3) Let's put 802.1ag and STP together, will that go?
Absolutely! It works in the presence or absence of STP. If using
STP, one should make sure that the CCM interval is high enough
that STP finds the problem and fixes it before CCM detects it.
Alternatively, CCM can detect the problem, but it doesn't really
fix it, and so the problem will eventually go away when STP reconverges
even though a defect may be indicated in the interim.
> If we have a
> topology with loops (presuming this is supported), will 802.1ag care about
> it and resolve or will it let STP do the job? My guess is that having both
> active simultaneously, STP may rearrange topology underneath 802.1ag,
> which in turn tries to find an alternative path via AIS and LTMs at the
> same time that STP processes a TC, and that looks strange to me now. Any
> ideas?
AIS has been removed from the spec several versions ago.
One cannot use "LTMs" or anything else to find paths. If one
doesn't use STP, the only way to create loop-free paths in
network topology with loops is to pre-provision them. That
is certainly an option, but the details of that are out of scope
for this spec.
> (4) 802.1ag says that it makes use of an extended discovery mechanism that
> can be found in chapter 12 of 802.1Q -- fine. But given .1Q invented some
> discovery mechanism prior to .1ag, what did they intend it for, wo used it
> and what for? I must admit never having seen such .1Q discovery so far.
Can you provide a more specific reference on this?
Anoop
| 
Yahoo! 
Google 
Windows Live 
del.icio.us 
digg 
Netscape 
XML site map