Re: [Pce] [Detnet] new draft on segment routing approach to TSN

Toerless Eckert <> Sat, 27 February 2021 18:19 UTC

Return-Path: <>
Received: from localhost (localhost []) by (Postfix) with ESMTP id D18523A11E0; Sat, 27 Feb 2021 10:19:16 -0800 (PST)
X-Virus-Scanned: amavisd-new at
X-Spam-Flag: NO
X-Spam-Score: 0.613
X-Spam-Status: No, score=0.613 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, HEADER_FROM_DIFFERENT_DOMAINS=0.25, LOTS_OF_MONEY=0.001, MONEY_NOHTML=1.481, SPF_HELO_NONE=0.001, SPF_NEUTRAL=0.779, URIBL_BLOCKED=0.001] autolearn=no autolearn_force=no
Received: from ([]) by localhost ( []) (amavisd-new, port 10024) with ESMTP id wXi4Tm8b1IAz; Sat, 27 Feb 2021 10:19:14 -0800 (PST)
Received: from ( [IPv6:2001:638:a000:4134::ffff:40]) (using TLSv1.2 with cipher ADH-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by (Postfix) with ESMTPS id E4D5F3A11DD; Sat, 27 Feb 2021 10:19:13 -0800 (PST)
Received: from ( []) by (Postfix) with ESMTP id 9A6A3548015; Sat, 27 Feb 2021 19:19:08 +0100 (CET)
Received: by (Postfix, from userid 10463) id 917E7440166; Sat, 27 Feb 2021 19:19:08 +0100 (CET)
Date: Sat, 27 Feb 2021 19:19:08 +0100
From: Toerless Eckert <>
To: "Pascal Thubert (pthubert)" <>
Cc: Tianran Zhou <>, Yaakov Stein <>, "" <>, "" <>, "" <>
Message-ID: <>
References: <> <> <>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <>
User-Agent: Mutt/1.10.1 (2018-07-13)
Archived-At: <>
Subject: Re: [Pce] [Detnet] new draft on segment routing approach to TSN
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Path Computation Element <>
List-Unsubscribe: <>, <>
List-Archive: <>
List-Post: <>
List-Help: <>
List-Subscribe: <>, <>
X-List-Received-Date: Sat, 27 Feb 2021 18:19:17 -0000

To add to Pascals reading list, check out:

This isn't meaning to endorse all the opinions and conclusions offered, but while probably
not being complete, i found it to be is AFAIK the most comprehensive survey for large scale
network bounded latency.

On Thu, Feb 25, 2021 at 05:40:29PM +0000, Pascal Thubert (pthubert) wrote:
> Hi Yaakov and all:
> Whereever Yaakov decides to place it I'll be there supporting the work. The draft itself is incredibly well-written and information-rich.
> Note that there's also work in RAW that mentions SR operation DetNet related operations (draft-pthubert-raw-architecture<>)5>). RAW has vested interest in intelligent forwarding decision, that would be the trademark vs. DetNet. With this draft, the forwarding is not based on Qbv schedule but the forwarder has some latitude as long as it matches the hop deadline. So RAW may be a good place.
> And then there's draft-chen-detnet-sr-based-bounded-latency<>01>. Ideally all these related items would progress in the same room.
> Also a few notes on the draft itself:
> - maybe use latency instead of delay; it would be nice to maybe define delay as something else, e.g., the delay representing the time the packet spends queued in one hop vs. the latency that is end to end?
> - not sure the term green wave is well understood by the public here; the draft gives the impression that the TSN path is faster than the best effort and involves no queueing. For the most part that is untrue; the latency is bounded but for most flows it is longer than best effort. Best effort can be really fast with passthrough in an empty network. The problem is the long tail and possibly congestion loss. For TSN, there can be very special flows that will traverse the city with all the lights green, but usually there'll be queuing. The difference is that the queueing latency is constant and the overall latency is withing bounds.
> - Time triggered is not the only TSN operation. I wonder what the draft would become with asynchronous shaper in mind. We designed (and as I must announce, patented as US9602420<>) a system very similar to the one proposed in the draft, but that is designed to adapt QoS depending on whether the packet is early or late vs. its schedule, and not tagging the schedule in the since the latency is considered end to end not hop by hop. The use case is slightly different since we apply this without a global controller and a provable guarantees all flows will meet the deadline - so not really detnet-, but more like a best effort that all flows meet their deadline in a stochastic environment. If Yaakov is interested, we can contribute on that aspect.
> Good luck with the draft,
> Pascal
> From: detnet <> On Behalf Of Tianran Zhou
> Sent: jeudi 25 février 2021 9:14
> To: Yaakov Stein <>om>;;;
> Subject: Re: [Detnet] new draft on segment routing approach to TSN
> Hi Yaakov,
> This is an interesting topic.
> After a quick review, there are several questions as follows:
> 1. It's clear to me to have a deadline for each packet. So that router can schedule the packet based on the urgency. But what's the motivation to split the end to end deadline to several local ones?
> 2. How to divide an end to end deadline into several local deadlines? Is there any example algorithm that could be used by the controller?
> 3. As far as I know, most devices do not support edf. I am not sure whether your proposal based on edf could really be useful.
> Cheers,
> Tianran
> From: Pce [] On Behalf Of Yaakov Stein
> Sent: Tuesday, February 23, 2021 9:14 PM
> To:<>;<>;<>
> Subject: [Pce] new draft on segment routing approach to TSN
> All,
> I would like to call your attention to a new ID
> which describes using a stack-based approach (similar to segment routing) to time sensitive networking.
> It furthermore proposes combining segment routing with this approach to TSN
> resulting in a unified approach to forwarding and scheduling.
> The draft is information at this point, since it discusses the concepts and does not yet pin down the precise formats.
> Apologies for simultaneously sending to 3 lists,
> but I am not sure which WG is the most appropriate for discussions of this topic.
>   *   DetNet is most relevant since the whole point is to control end-to-end latency of a time-sensitive flow.
>   *   Spring is also directly relevant due to the use of a stack in the header and the combined approach just mentioned.
>   *   PCE is relevant to the case of a central server jointly computing an optimal path and local deadline stack.
> I'll let the chairs decide where discussions should be held.
> Y(J)S

> _______________________________________________
> detnet mailing list