 |
 |
 |
 |
|
 |
|
 |
|
 |
|
 |
|
|
|
|
|
|
|
|
|
|
|
|
George, On Jul 9, 2008, at 3:50 PM, George Swallow wrote:
I disagree with the text below. It is my understanding that a lot of equipment does ECMP by looking for EOS, checking the first nibble for 0100b and if true running the usual IPv4 load-balancing. Only if it is not 0100b (and maybe 0110b) does it actual use some portion of the label stack. This is why in PW, the first nibble of the CW is 0000b.
It seems to me the only way to be sure that this will work is to define a new FEC in LDP. This would be equivalent to an IP FEC, but say that you must load balance on the label stack only.
Although an operator could, at the appropriate point in time, (manually) configure the transit LSR's to only pay attention to the label stack for load-balancing, this suggestion is likely better for backwards compatibility, so we'll take it under advisement.
Thanks, -shane
...George 4.2. Transit LSR Transit LSRs have no change in forwarding behavior. For load balancing, transit LSRs SHOULD use the whole label stack (e.g., for computing the load balance hash). Transit LSRs MAY choose to look beyond the label stack for further load balancing information; however, if entropy labels are being used, this may not be veryuseful. In a mixed environment (or for backward compatibility), thisis the simplest approach. Thus, transit LSRs are almost unaffected by the use of entropylabels. If transit LSRs were programmed to use a subset of the labelstack, they may have to be reconfigured to use the full stack. But otherwise, no changes are needed.
_______________________________________________ mpls mailing list mpls at ietf.org https://www.ietf.org/mailman/listinfo/mpls