Re: [Lsr] IP layer metrics collected by Edge routers - draft-dunbar-lsr-5g-edge-compute-ospf-ext (was: LSR Presentation Slot request)

[Linda Dunbar <[e-mail]>]

Les,

Thank you for the comments.
Replies are inserted below:


From: Les Ginsberg (ginsberg) <[e-mail]>
Sent: Monday, July 12, 2021 8:32 PM
To: Acee Lindem (acee) <acee=[e-mail]>; Linda Dunbar <[e-mail]>; Yingzhen Qu <[e-mail]>; [e-mail]
Cc: [e-mail]
Subject: RE: [Lsr] IP layer metrics collected by Edge routers - draft-dunbar-lsr-5g-edge-compute-ospf-ext (was: LSR Presentation Slot request)

Linda -

Picking up on this comment from Acee:

"Note that routes are based on IP prefixes and not applications while the draft uses these two interchangeably. "
[Linda] Yes, the application is identified by their IP addresses.

As regards how to advertise the new metric, to the best of my understanding what you want to advertise is the cost to reach an anycast address - which argues for a prefix Reachability advertisement. And indeed, that is what you chose to use for OSPF. It therefore makes no sense to me why you would use a link attribute advertisement when advertising the same information in IS-IS.
[Linda]  Is it better to use the TLV 22 (Extended IS Reachability) to carry the Site-Cost subTLVs specified in the draft?  Or is TLV 23  (IS Neighbor Attribute ) more appropriate?


I also think you are quite confused about the application part of "ASLA" as defined in RFC 8919/8920. The application identifies which applications are using the link attribute advertisements - it does not define the attribute itself - which potentially could be used by any application.
[Linda] since not every node will utilize the detailed IP Layer Metrics carried in the Site-COST,  the BIT MASK is to indicate if a Node should even process the detailed IP layer metrics. Is "ASLA"  intended to be?

If you need to advertise a new type of metric, the identification of that metric derives from the (sub-)TLV codepoint used - not from any of the applications bits. Your proposal to define a new application "C" therefore seems inappropriate.

+1 to Acee's other comments.

   Les


From: Lsr <[e-mail]<mailto:[e-mail]>> On Behalf Of Acee Lindem (acee)
Sent: Monday, July 12, 2021 5:06 PM
To: Linda Dunbar <[e-mail]<mailto:[e-mail]>>; Yingzhen Qu <[e-mail]<mailto:[e-mail]>>; [e-mail]<mailto:[e-mail]>
Cc: [e-mail]<mailto:[e-mail]>
Subject: Re: [Lsr] IP layer metrics collected by Edge routers - draft-dunbar-lsr-5g-edge-compute-ospf-ext (was: LSR Presentation Slot request)

Hi Linda,

From: Linda Dunbar <[e-mail]<mailto:[e-mail]>>
Date: Monday, July 12, 2021 at 5:41 PM
To: Acee Lindem <[e-mail]<mailto:[e-mail]>>, Yingzhen Qu <[e-mail]<mailto:[e-mail]>>, "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>
Cc: "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>
Subject: IP layer metrics collected by Edge routers - draft-dunbar-lsr-5g-edge-compute-ospf-ext (was: LSR Presentation Slot request)

Acee,

The draft-dunbar-lsr-5g-edge-compute-ospf-ext has two parts:

-          Aggregated Cost Advertisement

-          IP Layer App-Metrics Advertisements by OSPF

"Aggregated Cost" is only applicable to scenario where all the A-ER can have a consistent algorithm to compute the Aggregated cost.

When it is not possible for all the egress edge routers to have a consistent algorithm to compute the aggregated cost or some routers need all the detailed IP Layer metrics for the App Servers for other purposes, the raw IP layer metrics collected A-ER will be distributed. Only the nodes that are capable of utilizing the metrics will process the sub-TLV.

So, why would these "capable" nodes have a consistent algorithm for using this complex set of metrics but the A-ERs would not have a consistent algorithm for aggregating the cost?

Since only a subset of routers within an IGP domain need to know those detailed metrics, the draft used your suggested  OSPFv2 Extended Prefix Opaque LSA for IPv4 and OSPFv3 Extended LSA with Intra-Area-Prefix TLV to carry the detailed sub-TLVs.  For routers that don't care about those metrics, they can ignore them very easily.

This just doesn't work. All routers in an IGP domain must use the same algorithm. You can't just draw a picture with an LDN directly connected to a couple A-ERs say that the LDNs can use your metrics to route application specific traffic. The problem could possibly be solved with flex algorithm but it would require a lot more specification. I guess with your simple topology different LDNs could use the metrics differently as well? This would explain why you are not concerned with "consistency"...

It worth noting that not all hosts (prefix) attached to an A-ER are ANYCAST servers that need network optimization. An A-ER only needs to advertise the App-Metrics for the ANYCAST addresses that match with the configured ACLs.

Note that routes are based on IP prefixes and not applications while the draft uses these two interchangeably.

Thanks,
Acee


Any other concerns?

Thank you
Linda Dunbar

From: Acee Lindem (acee) <[e-mail]<mailto:[e-mail]>>
Sent: Monday, July 12, 2021 4:04 PM
To: Linda Dunbar <[e-mail]<mailto:[e-mail]>>; Yingzhen Qu <[e-mail]<mailto:[e-mail]>>; [e-mail]<mailto:[e-mail]>
Cc: [e-mail]<mailto:[e-mail]>
Subject: Re: [Lsr] LSR Presentation Slot Requests - IETF111

Speaking as WG member:

Hi Linda,
Even if you've added some IS-IS encodings, the draft still suffers from the fundamental problem of the previous draft. If you can't rely on the A-ERs to consistently calculate an aggregated metric, how can you rely on all the routers in the IGP routing domain to use complex set of metrics to reach the least-loaded app server? Do we really want to talk about this again?
Thanks,
Acee

From: Linda Dunbar <[e-mail]<mailto:[e-mail]>>
Date: Monday, July 12, 2021 at 4:27 PM
To: Yingzhen Qu <[e-mail]<mailto:[e-mail]>>, "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>
Cc: "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>
Subject: RE: [Lsr] LSR Presentation Slot Requests - IETF111
Resent-From: <[e-mail]<mailto:[e-mail]>>
Resent-To: Acee Lindem <[e-mail]<mailto:[e-mail]>>, Christian Hopps <[e-mail]<mailto:[e-mail]>>, Yingzhen Qu <[e-mail]<mailto:[e-mail]>>
Resent-Date: Monday, July 12, 2021 at 4:27 PM

Yingzhen and LSR Chairs,

We need a 10 minutes slot at IETF111 to present https://datatracker.ietf.org/doc/draft-dunbar-lsr-5g-edge-compute/<https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fdraft-dunbar-lsr-5g-edge-compute%2F&data=04%7C01%7Clinda.dunbar%40futurewei.com%7C7924484d75ec44a8892f08d9459e0034%7C0fee8ff2a3b240189c753a1d5591fedc%7C1%7C0%7C637617367189464541%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&sdata=Vjp9BD8jGQfTuvixVyfWDzDnv0tlEkFhYp18Zh%2F0KPE%3D&reserved=0>
Speaker: Linda Dunbar.

This draft adds the IS-IS extension to the draft-dunbar-lsr-5g-edge-compute-ospf-ext-04.

Thank you
Linda Dunbar


From: Lsr <[e-mail]<mailto:[e-mail]>> On Behalf Of Yingzhen Qu
Sent: Wednesday, June 30, 2021 4:00 PM
To: [e-mail]<mailto:[e-mail]>
Cc: [e-mail]<mailto:[e-mail]>
Subject: [Lsr] LSR Presentation Slot Requests - IETF111

Hi all,

The draft agenda for IETF111 has been posted: https://datatracker.ietf.org/meeting/111/agenda<https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdatatracker.ietf.org%2Fmeeting%2F111%2Fagenda&data=04%7C01%7Clinda.dunbar%40futurewei.com%7C7924484d75ec44a8892f08d9459e0034%7C0fee8ff2a3b240189c753a1d5591fedc%7C1%7C0%7C637617367189474498%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&sdata=Jya3cOSyJ%2BjcKzsddeFhE6DxFEYA89j4E3LFlL1HROo%3D&reserved=0>.

LSR will have one meeting session: Friday, July 30, 2021 16:00-18:00  Session III PDT

Please send slot requests to [e-mail]<mailto:[e-mail]>. Please include name of the presenter, pointer to the draft and time estimation including Q&A.

Thanks,
Yingzhen