“SRV6+” complexity in forwarding

["Darren Dukes (ddukes)" <[e-mail]>]

Hi Ron, I agree ASICs are always improving, indeed this is evident in the number of successful SRv6 deployments and multiple vendor implementations at line rate on merchant silicon, and multiple vendor ASICs.

Is “SRv6+” (PSSI+CRH+PPSI) implemented and deployed at line rate?
You’ve been asked this several times.  Since you’re the only implementor(?) and one operator is claiming deployment or testing, I am curious.

Regardless of ASIC capabilities there are two performance penalties you will not escape with PSSI+CRH+PPSI: TLV parsing and multiple lookups.

Requiring all segments in a CRH segment list to process an arbitrary length DOH+set of PSSI’s and other options is always very expensive.
- It is expensive in SRAM as previously discussed in these threads.
- It is expensive in parsing logic to know and process a set of TLVs in any ASIC or NP.

Spreading PSSI, CRH, PPSI operations in multiple headers and multiple identifiers you now have multiple lookups at a node.
1 - lookup destination address
2 - lookup one or more PSSI and future destination options.
3 - lookup the CRH label or PPSI label.
4 - lookup new destination address

Compare this with SRv6.
1 - lookup destination address
2 - lookup new destination address

While ASICs are more capable and will continue to be more capable, these technical performance problems you introduce with PSSI+CRH+PPSI will not go away.

Darren

On Sep 12, 2019, at 12:34 PM, Ron Bonica <rbonica=[e-mail]<mailto:rbonica=[e-mail]>> wrote:

Mark,

I think that you may have exposed yet another elephant in the room……

IPv6 defines a robust extensibility architecture, that includes multiple extension headers. Initially, IPv6 adoption was slow and router vendors were not highly motivated commit extension headers to ASICs. Also, in those days, ASICs were not so capable as they are today.

From the above-mentioned data points, we should not infer that it is beyond the capability of a modern vendor to develop an ASIC that supports a more complete set of extension headers. Two things have changed. As IPv6 adoption progresses, vendors are becoming more committed to IPv6. Beyond that, ASICs have become more capable over the decades.

We shouldn’t abandon the IPv6 extensibility architecture based on a claim that ASICs cannot and will never be able to  process multiple extension headers.

                                                                                                  Ron




From: ipv6 <[e-mail]<mailto:[e-mail]>> On Behalf Of Mark Smith
Sent: Thursday, September 12, 2019 1:30 AM
To: EXT - [e-mail]<mailto:[e-mail]> <[e-mail]<mailto:[e-mail]>>
Cc: Rob Shakir <[e-mail]<mailto:[e-mail]>>; SPRING WG <[e-mail]<mailto:[e-mail]>>; 6man <[e-mail]<mailto:[e-mail]>>; Robert Raszuk <[e-mail]<mailto:[e-mail]>>; [e-mail]<mailto:[e-mail]>; Tarek Saad <[e-mail]<mailto:[e-mail]>>
Subject: Re: [spring] Beyond SRv6.

It's simple because IPv6 doesn't look past the fixed IPv6 header to perform its forwarding, and matches on the Destination Address to determine if to perform deeper packet host processing.


You're building much more complicated forwarding services if you're going to be marching on TLVs etc. past the IPv6 fixed header.

However vendors and carrier engineering groups like selling and deploying new gear, so I suppose that's ok. They don't have to operate it or fix it when it breaks.
On Thu, 12 Sep 2019, 13:41 Bernier, Daniel, <[e-mail]<mailto:[e-mail]>> wrote:

+1



The ability of using a single SRH to convey behaviour information wether they are per-segment or per-path has proven to be very simple and quick to define in various data plane targets.



At first analysis, trying to replicate with CRH + DOH variants, the logic required for service programs is more com​plex.



What happens if I need TLVs mid-point in a path but not at its end (e.g. referring to the Ole's ACME analogy) ? Would they now be defined in a seg-end-opt or a vpn-dest-opt ? If seg-end-opt then it means non-interested midpoint devices will have to lookup beyond the TLVs to get to CRH for next segment ?



Similar question would be on how would we implement proxy behaviours either dynamic or static ? If I read https://tools.ietf.org/html/draft-bonica-6man-seg-end-opt-04<https://urldefense.com/v3/__https:/tools.ietf.org/html/draft-bonica-6man-seg-end-opt-04__;!8WoA6RjC81c!WhasJYFTRmXzKd1g-oMU5hza4EoH-63AFe6qzFFZtlfTRAiabJjCZB0f5dp14y8L$> correctly, we then need NSH for richer service chains constructs (think Gi-LAN). This means I need CRH, some variants of DOH + NSH​.



I fail to see the simplicity there.



Dan B

________________________________
From: spring <[e-mail]<mailto:[e-mail]>> on behalf of [e-mail]<mailto:[e-mail]> <[e-mail]<mailto:[e-mail]>>
Sent: Wednesday, September 11, 2019 8:57 PM
To: List
Cc: Rob Shakir; 6man; Tarek Saad; Robert Raszuk
Subject: [EXT]Re: [spring] Beyond SRv6.


Hi, folks,
Last year China Telecom begun to implement SRv6 trial in Sichun province for the bearing and interconnection of video platforms which are  located in different cities, service agilities,secure sepertion, traffic steering and other features of SRv6 have been demonstrated in this trial. Based on this, SRv6 will be implementated in larger-scale in CT.
No technologies is 100% perfect,I agree that compression mechanism is needed to reduce the the overhead of long SID in the long term, but it is better to be compatible withe SRH, instead of designing a complete new one. CRH is a complete new design, it move the complexities from SRH to DOH.
Moreover, I hope more efforts of SRv6 should be given on how to support new services,for instance, Application-aware network (APN). Meanwhile, we should consider more on how to implmement smooth transition and protect the network assetof carriers.

Best regards
Chongfeng


From: Dirk Steinberg<mailto:[e-mail]>
Date: 2019-09-09 21:31
To: Gyan Mishra<mailto:[e-mail]>
CC: SPRING WG List<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>; Robert Raszuk<mailto:[e-mail]>; Rob Shakir<mailto:[e-mail]>; Tarek Saad<mailto:[e-mail]>
Subject: Re: [spring] Beyond SRv6.
There seems to be some confusion regarding TI-LFA.
A couple of comments:

- Remote LFA tunnel is not used with SR, only TI-LFA which
  only operates on the node that is the PLR (point of local repair).

- Any encapsulation on the ingress PE with or without EH has nothing
  to do with TI-LFA except for the special case where the ingress PE
  itself is the PLR.

- TI-LFA is not an IGP extension and does not require one.
  It is a purely local computation based on IGP topology.

- The PLR for TI-LFA may need to insert some SIDs into the SID
  list to steer the packet around the failure. For the LFA base case
  no SIDs are needed at all. If SID insertion is needed, the PLR
  will push the required number of labels in the MPLS case.

  For SRv6, the equivalent operation to the label push is to
  insert an EH with the required SID list. The packet will already
  have been encapsulated on the ingress PE and in the most
  common Internet or VPN base use case it will not even have
  an EH so that this EH insertion will result only in a single EH..

  Alternatively, the PLR could also be configured to perform
  encapsulation with a new IPv6 header using the repair SID
  as IPv6 destination address, without needing any EH.
  This will work for the vast majority of cases.
  Remember that one 128-bit SID in SRv6 is in most cases
  equivalent to 2 MPLS labels, i.e. a node label plus an
  adjacency SID can be encoded in a single SRv6 SID.

  Only in extreme cases would the PLR need to add an
  EH to the new IPv6 header with more SIDs.

- EH insertion for TI-LFA has nothing to do with stitching SRv6 domains.

Hope it helps.

Cheers
Dirk


Am 08.09.2019 um 09:19 schrieb Gyan Mishra <[e-mail]<mailto:[e-mail]>>:

>From reading through all the discussion threads the SR insertion is two fold one being for FRR capabilities using Ti-LFA or remote LFA tunnel so end up requiring double EH insertions on the Ingress PE tunnel head end SRv6 source node and then second scenario being a possible EH insertions occurrence on intermediate nodes.  I have not read through the drafts or RFC regarding Ti-LFA with SR but since that is an IGP extension I am guessing an opaque LSA and is not the traditional MPLS FRR link/node/path protection that adds an additional mpls shim so not sure why an EH insertion needs to occur for Ti-LFA.  Can someone clarify that use case for me.  Also the EH insertion on intermediate node what is the use case or reason for that.  My guess is it’s for special use case of stitching SRv6 domains together.  Please clarify..


_______________________________________________
spring mailing list
[e-mail]<mailto:[e-mail]>
https://www.ietf.org/mailman/listinfo/spring<https://urldefense.com/v3/__https:/www.ietf.org/mailman/listinfo/spring__;!8WoA6RjC81c!WhasJYFTRmXzKd1g-oMU5hza4EoH-63AFe6qzFFZtlfTRAiabJjCZB0f5aWDOXgb$>

Juniper Business Use Only
--------------------------------------------------------------------
IETF IPv6 working group mailing list
[e-mail]<mailto:[e-mail]>
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------