Hi Kireeti, I quite disagree about the more state argument.
If you treat mpls label as prefix with mask which is all this boils down to there is no more state either in data or control plane. As to the point of not perfect load balancing it all depends on your hash function. If you always advertise at least as big block as you have max parallel links on any interface you should be fine. Best , R. On Jul 24, 2014 2:12 PM, "Kireeti Kompella" <kire...@juniper.net> wrote: > Hi Robert, > > On Jul 24, 2014, at 03:21 , Robert Raszuk <rob...@raszuk.net> wrote: > > Therefor an alternative of using any form of entropy labels in data > plane all together would be to allocate a SID blocks (say 64 or 128 wide) > and allow SPRING header imposition to use such pools of SIDs per group of > flows to effectively allow for efficient load balancing in the network. > > > We’re rehashing (pun unintended) the entropy label discussion. > > Before we settled on entropy labels the way we did, we discussed > advertising label blocks in LDP. There are two problems: > a) as Stephane mentions, ELs are stateless, but pools of SIDs impose more > state in the forwarding plane (and to some extent, in the control plane). > b) if you use small blocks, you get uneven load balancing; if you use > large blocks, you blow up the state more. > > Kireeti > >
_______________________________________________ spring mailing list spring@ietf.org https://www.ietf.org/mailman/listinfo/spring
