Posting from another thread for better organization:

On Tue, Jun 7, 2022 at 10:06 AM Gray, Jonathan
<[email protected]> wrote:

> The data kindof exists.  Servers have Physlocations which have addresses
> of some kind.  Their purpose is presently ambiguous as to if that should be
> used for physical address or mailing address.  Clearing that ambiguity and
> adding lat/long are nice to haves, but that’s really a TO enhancement
> rather than a TP one.  Additionally those physlocation lat/longs wouldn’t
> presently used in a meaningful way since all routing decisions are based on
> the cachegroup.  There are reasons that a cachegroup location might be
> skewed away from the physlocation or serve as an area aggregate.  Today you
> can suboptimally equate physlocation and cachegroup for routing/capacity
> purposes if desired, but that gets pretty far into micromanagement and
> capacity issues.  Analytics I’ve done in the past that were useful were to
> create convex hull polygons of the physlocations associated to a cachegroup
> and check that the areas were reasonable for the tiers as well as to ensure
> the cachegroup lat/long was contained within it or nearby to where it made
> sense.  That’s shown in the past where we’ve had physlocations set wrong on
> hosts due to ambiguous physlocation names.  Although that didn’t affect
> routing just RMA work.
>
> Better visualization around parentage in general would be a definite win.
> Given how we’ve gone from a 1 or 2 tier CDN model with secondary parents
> and MSO to an N-tier model with additional dimensions of permutations
> around topologies as well as capabilities/requirements, as an SRE it’s much
> more difficult to know how traffic ideally should be flowing for diagnostic
> purposes.  That view doesn’t have to be geographic in nature though.  A
> graph/tree would suffice.  From a perspective of self-service though, all
> the topic around parentage is generally too complex for a non-ATC engineer
> and too powerful.  Better visualization around parentage I think benefits
> both user groups.
>
> Jonathan G
>
>
> From: Dave Neuman <[email protected]>
> Date: Tuesday, June 7, 2022 at 9:30 AM
> To: [email protected] <[email protected]>
> Subject: [EXTERNAL] Re: 2022-06-07 TC Working Group Agenda and Meeting
> Notes
> >>   - Zach proposes/iss working on a geographic Topology builder using
> Cache Group latitude/longitude to map them out
>
> I would suggest incorporating the physical locations of the hosts as well.
> Maybe as a drill down?  Cache groups can be as big as a whole country,
> having the actual locations of the caches could be much more helpful.
>
> Just my $.02
>
> --Dave
>

-Zach

On Mon, Jun 6, 2022 at 9:15 AM Wilson, Amy <[email protected]>
wrote:

> To add some additional color, a new delivery service self-service feature
> really means is 2 things:
>
>   *   simplified view for non-super/advanced users
>   *   eliminate the need to queue [server] updates / CDN snapshot to
> propagate changes - https://github.com/apache/trafficcontrol/issues/6183
>
> ~Amy
>
> From: Wilson, Amy <[email protected]>
> Date: Wednesday, June 1, 2022 at 4:04 PM
> To: [email protected] <[email protected]>
> Subject: Call for Traffic Portal Enhancements / New Features
> Good afternoon,
>
> I hope this email finds everyone doing well.
>
> Looking to gather enhancements or new features for Traffic Portal.
> Specifically, around User Management, Tenancy, and Delivery Services
> management.
>
> Please provide your detailed ideas/requirements.
>
> Thanks,
> ~Amy
>

Reply via email to