Colby, I was mainly talking about shaping into LSP (rate limiting). And it is a feature not a bug to reoptimize the TE-LSPs in the network in the order of tens of minutes or longer. Transient short lived peaks should not trigger such reoptimization.
In any case I stop now. My concern has been raised. You are likely going to ship it anyway and consider TTE to be a cool thing. I can only hope that if that draft is going to be published as an informational RFC it will contain an honest section of deployment risks associated with enabling such automation of blind load shift. Cheers, Robert. On Tue, Apr 11, 2023 at 2:32 PM Colby Barth <[email protected]> wrote: > Hi Robert- > > > > *From: *Robert Raszuk <[email protected]> > *Date: *Tuesday, April 11, 2023 at 5:59 AM > *To: *Tony Li <[email protected]> > *Cc: *Colby Barth <[email protected]>, RTGWG <[email protected]> > *Subject: *Re: TTE > > *[External Email. Be cautious of content]* > > > > Tony, > > > > No, we’re trying to protect against unpredicted load. That doesn’t mean > that TE was very badly provisioned, just that it wasn’t perfect. Whatever > the root cause, congestion does happen and we want to try to address it. > > > > Since very early days of RSVP-TE one fundamental recommendation we were > giving operators was to always do policing/rate-limiting what goes into a > TE-LSP on ingress. There is nothing more or less perfect about it - it is > either in place or not. Such policing/rate-limiting would also auto adjust > when auto-bandwidth has been enabled. So if the unexpected traffic entered > the network it was either dropped or shaped on ingress. > > > > <cb> Ingress rate limiting doesn’t adapt to load. It just drops or passes > (may remark in some cases contributing to congestion further). Auto-bw > does adapt but as has been previously mentioned, auto-b/w adjustments can > take time on the order of minutes. TTE is meant to bridge the time gap > while auto-bw (or something like auto-bw) does it’s optimization. > > > > --Colby > > > > You are talking about unpredicted load suddenly appearing in the middle of > the network and NOT as a result of already single network failure. I guess > you are still confirming that you will not protect with TTE traffic which > is already protected with vanilla FRR due to node/link failure. > > > > Regards, > > R. > > > > Juniper Business Use Only >
_______________________________________________ rtgwg mailing list [email protected] https://www.ietf.org/mailman/listinfo/rtgwg
