Re: [Tsv-art] [v6ops] [Last-Call] Tsvart last call review of draft-ietf-v6ops-ipv6-ehs-packet-drops-05

On Thu, Feb 18, 2021 at 1:08 PM Fernando Gont <fgont@si6networks.com> wrote:
>
> Hi, Brian,
>
> On 18/2/21 16:35, Brian E Carpenter wrote:
> > On 19-Feb-21 06:56, Fernando Gont wrote:
> [....]
> >>> * The ID also discusses use of the IPv6  Flow Label: This seems a little off
> >>> topic, but seems linked to EH implications on ECMP.  However, the final
> >>> sentence of this section is a reference to [Jaeggli-2018] which in turn
> >>> concludes that the IPv6 Flow Label should not be used it as a part of hashes
> >>> for load balancing. Yet, as far as I know, this is not the recommendation of
> >>> the IETF in 2020.
> >>
> >> FWIW, we discuss the Flow Label a bit because the usual reaction would
> >> be "why do you process the header chain for load-balancing, instead of
> >> employing the Flow Label?"
> >
> > However, Gorry is right that citing Joel's operational comments is only part
> > of the story. I immodestly suggest citing RFC 7098 too, for a discussion
> > of how the flow label can in principle be used for server load balancing.
>
> Good grief! I thought we were referencing this one already (but looks
> like we're not) Any suggestion on how to reference it?
>
> E.g., we could do:
>
>     Thus, ECMP and Hash-based Load-Sharing [rfc6434] [RFC7098] should be
> possible
>     without the need to process the entire IPv6 header chain to obtain
>     upper-layer information to identify lows.
>
>
> plus adding it ALONG WITH RFC6438 here:
>
>     making Flow Label-based ECMP
>     and Hash-based Load-Sharing [RFC6438] feasible.
>
>
>
> xor add this:
>
>    [RFC7098] discusses how the IPv6 FLow Label can used to enhance layer
>     3/4 (L3/4) load distribution and balancing for large server farms.
>
> right after:
>     Thus, ECMP and Hash-based Load-
>     Sharing should be possible without the need to process the entire
>     IPv6 header chain to obtain upper-layer information to identify
>     flows.
>
I don't why this is only a "should". Hashing of three tuple for load
balancing works as described as evident by the fact it is in wide
deployment (consider it's been in Linux stack for it's internal load
balancing for several years and there's now over a billion devices in
the world that do this).

Tom

>
>
> Alternatively, if you can think of a better way to do it, please do let
> us know.
>
> Thanks!
>
> Regards,
> --
> Fernando Gont
> SI6 Networks
> e-mail: fgont@si6networks.com
> PGP Fingerprint: 6666 31C6 D484 63B2 8FB1 E3C4 AE25 0D55 1D4E 7492
>
>
>
>
> _______________________________________________
> v6ops mailing list
> v6ops@ietf.org
> https://www.ietf.org/mailman/listinfo/v6ops