Re: [Lsr] IPv6 Flow Label QOS marking support for 5-tuple ECMP / LAG / MLAG hash

["Acee Lindem (acee)" <[e-mail]>]

Speaking as LSR Co-Chair:

Hi Gyan,
This is more a discussion for the 6MAN WG. Here is the charter for the LSR WG: https://datatracker.ietf.org/wg/lsr/about/
No need to cross-post to the LSR list…
Thanks,
Acee

From: Lsr <[e-mail]> on behalf of Gyan Mishra <[e-mail]>
Date: Monday, November 30, 2020 at 3:22 PM
To: lsr <[e-mail]>
Subject: [Lsr] IPv6 Flow Label QOS marking support for 5-tuple ECMP / LAG / MLAG hash


Dear LSR WG experts,


Does anyone know if vendors have started or plan to start supporting IPv6 flow label 5-tuple dscp marking for ECMP hashing.

IPv6 flow label support for ECMP

https://tools.ietf.org/html/rfc6437

IPv4 has traditionally always utilized recommended BCP of flow based load balancing due to issues related to out of order and reordering of packets.  Although per packet load balancing is supported by most vendors it is not recommended due to forwarding plane impact.

This IPv6 flow label feature of 5-tuple hash provides significant advantages for operators much needed ECMP load balancing entropy as compare to traditional “flow or session” based load balancing which is the case as well with MPLS entropy label RFC 6790 load balancing contrasted below.

IPv6 flow label has significant  benefits for operators deploying  SRv6 which utilizes the IPv6 data plane to now have “native” built in ECMP entropy as part of the protocol as compare to its predecessor IPv4.

This gives SRv6 another significant edge over MPLS predecessor.

Excerpt from RFC 6437:


      Forwarding nodes such as routers and load distributors MUST NOT

      depend only on Flow Label values being uniformly distributed.  In

      any usage such as a hash key for load distribution, the Flow Label

      bits MUST be combined at least with bits from other sources within

      the packet, so as to produce a constant hash value for each flow

      and a suitable distribution of hash values across flows.

      Typically, the other fields used will be some or all components of

      the usual 5-tuple.  In this way, load distribution will still

      occur even if the Flow Label values are poorly distributed.



   Although uniformly distributed flow label values are recommended

   below, and will always be helpful for load distribution, it is unsafe

   to assume their presence in the general case, and the use case needs

   to work even if the flow label value is zero.



   As a general practice, packet flows should not be reordered, and the

   use of the Flow Label field does not affect this.  In particular, a

   Flow label value of zero does not imply that reordering is

   acceptable.





Below comparison of IPv6 flow label benefits over MPLS entropy label:





! MPLS Entropy label

https://tools.ietf.org/html/rfc6790







As a comparison to MPLS entropy label, the mpls entropy label reduces the control plane label binding and LFIB forwarding plane data structure by not having a per ECMP path label allocation per FEC by adding an additional entropy label to the label stack.





However MPLS entropy label is still uses the traditional flow or session based load balancing algorithm which results in

uneven load balancing.





Kind Regards



Gyan



--

[Image removed by sender.]<http://www.verizon.com/>

Gyan Mishra

Network Solutions Architect

M 301 502-1347
13101 Columbia Pike
Silver Spring, MD