All I think I figured out how “source port entropy” works and provides “better” load balancing then traditional IGP & EGP based ECMP underlay algorithms that are subject to polarization.
So normally w/o source port entropy vxlan feature the overlay NVE Anycast vtep tunnel as a tunnel source and destination so as it’s a single source / destination IP for the vtep tunnel termination, so that would get pinned to a single path similar to L2 VPN service label, single source / destination PE to PE single path. So just as with L2 VPN you have the FAT PW which now you read into the payload and extract the src/dest flows and can now load balance the flows over Ethernet bundles. Similarly with VXLAN “source port entropy” per RFC 7348 the L2/L3/L4 headers 5-tuple hash is used to generate the outer header udp source port which is used as the input key to the hashing function. So one of the major advantages of VXLAN is now traffic can be much more uniformly evenly load balanced now with 5-tuple info over L3 ECMP path as compare to traditional IPv4 flow based per session source / destination hash load balancing. The vxlan 5-tuple hash input key to hash function is also very similar analogous to IPv6 flow label RFC 6437 5-tuple header hash input key to hash function stateless mode flow label uniform load balancing. Kind Regards Gyan On Wed, Jul 14, 2021 at 6:11 PM Gyan Mishra <hayabusa...@gmail.com> wrote: > > Dear BESS Experts > > ?? On NVO3 VXLAN overlay encapsulation > > My understanding of VXLAN source port entropy is to provide uniform load > balancing similar to RFC 6437 IPv6 flow label uniform stateful load > balancing, in NVO3 VXLAN context, using header 5-tuple L2/L3/L4 hash and > generating source port input key to hash function for per packet per flow > uniform load balancing as achieved with EVPN ECMP pr weighted UCMP MHD > MLAG PE-CE AC. > > The problem with L3 ECMP and weighed UCMP is the Day 1 well known TCP > polarization of flows where high bandwidth flows are not evenly distributed > between L3 paths. > > So the question is does source port entropy provide per RFC 7348 excerpt > below provide per packet per flow load balancing or flow based where all > packets that are part of the same flow get hashed to the same path. > > Outer UDP Header: This is the outer UDP header with a source port > provided by the VTEP and the destination port being a well-known > UDP port. > > - Destination Port: IANA has assigned the value 4789 for the > VXLAN UDP port, and this value SHOULD be used by default as the > destination UDP port. Some early implementations of VXLAN have > used other values for the destination port. To enable > interoperability with these implementations, the destination > port SHOULD be configurable. > > - Source Port: It is recommended that the UDP source port number > be calculated using a hash of fields from the inner packet -- > one example being a hash of the inner Ethernet frame's headers. > This is to enable a level of entropy for the ECMP/load- > balancing of the VM-to-VM traffic across the VXLAN overlay. > When calculating the UDP source port number in this manner, it > is RECOMMENDED that the value be in the dynamic/private port > range 49152-65535 [RFC6335 > <https://datatracker.ietf.org/doc/html/rfc6335>]. > > > > Kind Regards > > Gyan > > > -- > > <http://www.verizon.com/> > > *Gyan Mishra* > > *Network Solutions A**rchitect * > > *Email gyan.s.mis...@verizon.com <gyan.s.mis...@verizon.com>* > > > > *M 301 502-1347* > > -- <http://www.verizon.com/> *Gyan Mishra* *Network Solutions A**rchitect * *Email gyan.s.mis...@verizon.com <gyan.s.mis...@verizon.com>* *M 301 502-1347*
_______________________________________________ BESS mailing list BESS@ietf.org https://www.ietf.org/mailman/listinfo/bess
