Re: [Lsr] New draft on Flex-Algorithm Bandwidth Constraints

Peter Psenak <ppsenak@cisco.com> Thu, 04 March 2021 11:06 UTC

Return-Path: <ppsenak@cisco.com>
X-Original-To: lsr@ietfa.amsl.com
Delivered-To: lsr@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id A98883A1923 for <lsr@ietfa.amsl.com>; Thu, 4 Mar 2021 03:06:06 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -9.602
X-Spam-Level:
X-Spam-Status: No, score=-9.602 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, NICE_REPLY_A=-0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001, USER_IN_DEF_DKIM_WL=-7.5] autolearn=unavailable autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=cisco.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id MPmNSYK9Y9Yx for <lsr@ietfa.amsl.com>; Thu, 4 Mar 2021 03:06:04 -0800 (PST)
Received: from aer-iport-3.cisco.com (aer-iport-3.cisco.com [173.38.203.53]) (using TLSv1.2 with cipher DHE-RSA-SEED-SHA (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 1465F3A18EE for <lsr@ietf.org>; Thu, 4 Mar 2021 03:06:02 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=cisco.com; i=@cisco.com; l=14207; q=dns/txt; s=iport; t=1614855963; x=1616065563; h=subject:to:cc:references:from:message-id:date: mime-version:in-reply-to:content-transfer-encoding; bh=D4rWQW90mZGQcBxvHenCYJfNRG9cg3m5HeiIcN6WmUY=; b=gbJXlEStSvqwYpwlRymdZjbpfDNJq/tT9d7BbTU5IsANvvXHJdIW2cZh 5XOH93VbotoPY9oW8mu1di++JRV0v6V3G09w6YqHkRVbnlc5ucEhlw0l2 ztr5CJ4lGGve47aCLJRCtMG2zeocBoFUeKx0QsFQ7bsRIzTC3I/mBg96v s=;
X-IPAS-Result: =?us-ascii?q?A0CWAQBkvkBg/xbLJq1iGgEBAQEBAQEBAQEDAQEBARIBA?= =?us-ascii?q?QEBAgIBAQEBQIFPg3cBJxIxhEGJBIgoCCUDnEsLAQEBDzQEAQGETQKBeyY4E?= =?us-ascii?q?wIDAQEBAwIDAQEBAQUBAQECAQYEcYVuhkQBAQEDASMPAQVBBQsLGAICIwMCA?= =?us-ascii?q?kYRBg0GAgEBgmyCZyGteHaBMoVYg0eBRIEOKowTNnpCgUlCgREnDII5Lj6HV?= =?us-ascii?q?IJfBIFVawYHOSgiDQwIMAIuB20VBCYvEy+QAwclK4I7AaYogwaDL5V8gmkFB?= =?us-ascii?q?wMfgzeKT4VPkAG3G4FrI4FXMxoIGxWDJFAZDY4rFo4nQAMvOAIGAQkBAQMJj?= =?us-ascii?q?BMBAQ?=
X-IronPort-Anti-Spam-Filtered: true
X-IronPort-AV: E=Sophos;i="5.81,222,1610409600"; d="scan'208";a="31475617"
Received: from aer-iport-nat.cisco.com (HELO aer-core-2.cisco.com) ([173.38.203.22]) by aer-iport-3.cisco.com with ESMTP/TLS/DHE-RSA-SEED-SHA; 04 Mar 2021 11:05:58 +0000
Received: from [10.60.140.52] (ams-ppsenak-nitro3.cisco.com [10.60.140.52]) by aer-core-2.cisco.com (8.15.2/8.15.2) with ESMTP id 124B5w0o025157; Thu, 4 Mar 2021 11:05:58 GMT
To: Robert Raszuk <robert@raszuk.net>
Cc: Tony Li <tony.li@tony.li>, Gyan Mishra <hayabusagsm@gmail.com>, DECRAENE Bruno IMT/OLN <bruno.decraene@orange.com>, Shraddha Hegde <shraddha@juniper.net>, Rajesh M <mrajesh@juniper.net>, "lsr@ietf.org" <lsr@ietf.org>, William Britto A J <bwilliam=40juniper.net@dmarc.ietf.org>
References: <161401476623.19237.3808413288895066510@ietfa.amsl.com> <DM5PR0501MB38006C4B638AD2AB6A7731B5CD9A9@DM5PR0501MB3800.namprd05.prod.outlook.com> <7C67D01F-24DB-4450-8587-E004CAFBBEBC@tony.li> <CAOj+MMGZppwYtNr4t0rJoy3BKWaBYqHiJ_esM1XNFTNxbm8c5w@mail.gmail.com> <08882555-009B-4068-ABB0-20B0D165D722@tony.li> <2c2605a8-95c6-a477-b1b5-5ae4d4de222a@cisco.com> <52B3A5ED-6ACC-4772-BEF7-085A33A53F31@tony.li> <e5190522-3a8b-2d6e-c2fe-646049689cc4@cisco.com> <1EABA651-2F05-415B-97EF-054507FADEAC@tony.li> <f935dbc4-6220-5f47-65a4-f642823f594f@cisco.com> <CAOj+MMHXd5j8B9a13E90HQVB=SUOkQ=fqhyJEgTf-Y7Tp5eiBQ@mail.gmail.com> <9d66e5af-414a-42b7-6af5-388974785b8f@cisco.com> <CAOj+MMGggyuzqD=ZbDxNG_10ThHykOO7y5rYWphvSk0rs+4CbQ@mail.gmail.com> <bb9555c4-8007-4d0e-2715-39c4069fc61f@cisco.com> <CAOj+MMFP+Tx13v99GaOc8vQaN6UPkSiS7UX1hRFwevKd8Oj13Q@mail.gmail.com> <532100cf-0252-73f3-046c-218de2fc26db@cisco.com> <CAOj+MMEYAc=LN6JfTto7fwcSzeQuyUdeTTC262c5LwBfPFgRRg@mail.gmail.com>
From: Peter Psenak <ppsenak@cisco.com>
Message-ID: <f21c7460-86ff-6169-1bf8-3e95e843e9df@cisco.com>
Date: Thu, 4 Mar 2021 12:05:58 +0100
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.14; rv:60.0) Gecko/20100101 Thunderbird/60.9.1
MIME-Version: 1.0
In-Reply-To: <CAOj+MMEYAc=LN6JfTto7fwcSzeQuyUdeTTC262c5LwBfPFgRRg@mail.gmail.com>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 8bit
X-Outbound-SMTP-Client: 10.60.140.52, ams-ppsenak-nitro3.cisco.com
X-Outbound-Node: aer-core-2.cisco.com
Archived-At: <https://mailarchive.ietf.org/arch/msg/lsr/MMqdUZBE3J_SgBYOsxMuiszcYZg>
Subject: Re: [Lsr] New draft on Flex-Algorithm Bandwidth Constraints
X-BeenThere: lsr@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Link State Routing Working Group <lsr.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/lsr>, <mailto:lsr-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/lsr/>
List-Post: <mailto:lsr@ietf.org>
List-Help: <mailto:lsr-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/lsr>, <mailto:lsr-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 04 Mar 2021 11:06:07 -0000

On 04/03/2021 11:52, Robert Raszuk wrote:
> I guess now you are not listening ;)
> 
> I am saying it is about finding the right normalization timers and 
> values which can satisfy the need. And those should reside on the senders.

that's about what tend to agreed on so far.

But that will not address the entire problem. No timers would make it 
work in environment where the delay fluctuates constantly and 
significantly. There timers would only make sure the stability is not 
compromised. But the delay optimized forwarding will not be achieved.

> 
> That is why I was asking what is there today and so far no one gave 
> precise answer. 

I did provide a pretty detailed answer about what is there in one 
implementation.


> You said links don't change delay characteristics which 
> by itself only applies to some link categories. Not satellite not even 
> 5G.Leave alone VPWS.
> 
> So I guess your conclusion is that this is hard problem and we should 
> not go there. But if so let's not pretend we are sending even min link 
> delay as dynamic value and redefine it as static approximated link delay.

no, I did not conclude anything. I was providing a reasoning of why min 
delay was chosen versus another, even more dynamic values.

Peter
> 
> Thx
> R.
> 
> 
> 
> 
> 
> 
> On Thu, Mar 4, 2021, 11:06 Peter Psenak <ppsenak@cisco.com 
> <mailto:ppsenak@cisco.com>> wrote:
> 
>     Robert,
> 
>     On 04/03/2021 10:50, Robert Raszuk wrote:
>      > Peter,
>      >
>      > Sorry but the draft does not have a disclaimer section which states:
>      >
>      > "Extensions defined below are only applicable to networks with
>     not even
>      > a single emulated circuit in the IGP."
>      >
>      >  > Obviously, if the min delay fluctuates wildly, one can not
>     achieve
>      > delay optimized
>      >  > forwarding no matter what.
>      >
>      > If VPWS provider's IP network reconverges once per day or per
>     hour I see
>      > nothing wrong with flooding new link delays and recomputing the
>     topology
>      > of the network running on top of such constructs.
> 
>     the question is how much sense would it make to try to optimize
>     based on
>     delay if it fluctuates every ms and we optimize once per day. But feel
>     free to give it a shot if you believe it's a good idea.
> 
>     Peter
> 
>      > My point is that
>      > changes are real, pretty unpredictable and in ms or 10s of ms.
>     And IMO
>      > the proposal on the table is specifically designed to catch and
>     address
>      > those cases - not just dismiss it as you can not use it in your
>     network
>      > - sorry mr customer. >
>      > Best,
>      > R.
>      >
>      >
>      >
>      >
>      >
>      >
>      > On Thu, Mar 4, 2021 at 10:41 AM Peter Psenak <ppsenak@cisco.com
>     <mailto:ppsenak@cisco.com>
>      > <mailto:ppsenak@cisco.com <mailto:ppsenak@cisco.com>>> wrote:
>      >
>      >     Robert,
>      >
>      >     On 04/03/2021 10:23, Robert Raszuk wrote:
>      >      > Peter,
>      >      >
>      >      > Completely disagree.
>      >
>      >     that's what you seem to be doing from the beginning of this
>      >     conversation ;)
>      >
>      >      >
>      >      > Real say enterprise networks are being build with emulated
>     circuits
>      >      > (example VPWS). In one company I was working at it was
>     about 80% of
>      >      > emulated links all over the world in their WAN. Yes for me
>     it was a
>      >      > shock as I did not realize how much this emulated links
>     took over
>      >     the
>      >      > world. In most geographies you even can not get any link
>     of less
>      >     then
>      >      > 10Gbps to be real any more. Only emulated option is on the
>     table.
>      >      >
>      >      > Emulated circuits run over someone's IP backbones. You can
>      >     understand
>      >      > the consequences of this. Not only link delay changes a
>     lot but
>      >     you run
>      >      > into very interesting set of issues.
>      >
>      >     look at it from the opposite direction. The provider of the
>     VPWS is the
>      >     one who can use this technology to guarantee the delay bound
>     of the
>      >     WPWs
>      >     service. And if it does, the user of the WPWs would not
>     experience the
>      >     wild variation in the min delay.
>      >
>      >     So you have to apply right set of tools at right location.
>      >     Obviously, if
>      >     the min delay fluctuates wildly, one can not achieve delay
>     optimized
>      >     forwarding no matter what. That does not mean that the
>     network will get
>      >     unstable.
>      >
>      >     Peter
>      >
>      >
>      >      >
>      >      > Maybe you think of the backbones of the 90s or 2000s where
>     dark
>      >     fiber or
>      >      > SDH or SONET were in use for interconnects.
>      >      >
>      >      > Well no more. IETF came with such brilliant ideas to
>     emulate L2
>      >     over L3
>      >      > and here we go.
>      >      >
>      >      > Reality is not what we wish it to be.
>      >      >
>      >      > Cheers,
>      >      > R.
>      >      >
>      >      >
>      >      > On Thu, Mar 4, 2021 at 10:07 AM Peter Psenak
>     <ppsenak@cisco.com <mailto:ppsenak@cisco.com>
>      >     <mailto:ppsenak@cisco.com <mailto:ppsenak@cisco.com>>
>      >      > <mailto:ppsenak@cisco.com <mailto:ppsenak@cisco.com>
>     <mailto:ppsenak@cisco.com <mailto:ppsenak@cisco.com>>>> wrote:
>      >      >
>      >      >     Robert,
>      >      >
>      >      >     On 03/03/2021 20:57, Robert Raszuk wrote:
>      >      >      > Peter,
>      >      >      >
>      >      >      >  >  that differ by few microsecond
>      >      >      >
>      >      >      > Really you normalize only single digit microseconds ???
>      >      >      >
>      >      >      > What if link delay changes in milliseconds scale ?
>     Do you
>      >     want to
>      >      >      > compute new topology every few milliseconds ?
>      >      >
>      >      >     let me repeat again.
>      >      >
>      >      >     Min delay is not something that changes every few
>     milliseconds
>      >      >     significantly. It's a semi static variable that
>     reflects the
>      >      >     property of
>      >      >     the underlying physical path. It only changes when the
>      >     physical path
>      >      >     properties changes - e.g. the optical path reroutes,
>     etc. We
>      >      >     deliberately picked Min delay for flex-algo purposes
>     for this
>      >     semi
>      >      >     static property.
>      >      >
>      >      >     The small, non significant changes can be filtered by the
>      >     normalization.
>      >      >
>      >      >     If the min delay changes significantly every few
>     milliseconds
>      >     there's
>      >      >     something wrong with the link itself - we have
>     standard dampening
>      >      >     mechanisms in LS protocols to deal with unstable links
>     that
>      >     would kick
>      >      >     in. Similar to what we do if the link flaps every few
>      >     milliseconds.
>      >      >
>      >      >      >
>      >      >      > Out of curiosity as this is not a secret -  What
>     are your
>      >     default
>      >      >     min
>      >      >      > delay normalization timers (if user does not overwrite
>      >     with their
>      >      >     own).
>      >      >
>      >      >     there is no timer needed for the normalization itself.
>      >      >
>      >      >     You are likely referring TWAMP computation interval
>     which is
>      >     30 sec,
>      >      >     with probes being sent every 3 seconds in our
>     implementation by
>      >      >     default,
>      >      >     if I'm not mistaken.
>      >      >
>      >      >     Normalization is applied to the value that come from
>     the above.
>      >      >
>      >      >     thanks,
>      >      >     Peter
>      >      >
>      >      >
>      >      >
>      >      >      > Likewise how Junos or Arista normalizes it today ?
>      >      >      >
>      >      >      > Thx,
>      >      >      > R.
>      >      >      >
>      >      >      > On Wed, Mar 3, 2021 at 7:41 PM Peter Psenak
>      >     <ppsenak@cisco.com <mailto:ppsenak@cisco.com>
>     <mailto:ppsenak@cisco.com <mailto:ppsenak@cisco.com>>
>      >      >     <mailto:ppsenak@cisco.com <mailto:ppsenak@cisco.com>
>     <mailto:ppsenak@cisco.com <mailto:ppsenak@cisco.com>>>
>      >      >      > <mailto:ppsenak@cisco.com
>     <mailto:ppsenak@cisco.com> <mailto:ppsenak@cisco.com
>     <mailto:ppsenak@cisco.com>>
>      >     <mailto:ppsenak@cisco.com <mailto:ppsenak@cisco.com>
>     <mailto:ppsenak@cisco.com <mailto:ppsenak@cisco.com>>>>> wrote:
>      >      >      >
>      >      >      >     Tony,
>      >      >      >
>      >      >      >     On 03/03/2021 19:14, Tony Li wrote:
>      >      >      >      >
>      >      >      >      > Peter,
>      >      >      >      >
>      >      >      >      >>> There are several link types in use that
>     exhibit
>      >     variable
>      >      >      >     delay: satellite links (e.g., Starlink), microwave
>      >     links, and
>      >      >      >     ancient link layers that deliver reliability
>     through
>      >      >     retransmission.
>      >      >      >      >>> Any of these (and probably a lot more) can
>     create a
>      >      >     noticeable
>      >      >      >     and measurable difference in TWAMP. That would be
>      >     reflected
>      >      >     in an FA
>      >      >      >     metric change. If you imagine a situation with
>      >     multiiple parallel
>      >      >      >     paths with nearly identical delays, you can easily
>      >     imagine an
>      >      >      >     oscillatory scenario.   IMHO, this is an
>     outstanding
>      >     concern
>      >      >     with FA.
>      >      >      >      >> yes, and that is what I referred to as "delay
>      >     normalization",
>      >      >      >     which can avoid that oscillation.
>      >      >      >      >
>      >      >      >      >
>      >      >      >      > It can also negate the benefits of the
>     feature. One
>      >     might well
>      >      >      >     imagine that Starlink would want to follow a
>     min-delay
>      >     path for
>      >      >      >     optimality.  If the delay variations are
>     “normalized”
>      >     out of
>      >      >      >     existence, then the benefits are lost.  The whole
>      >     point is to
>      >      >     track
>      >      >      >     the dynamics.
>      >      >      >
>      >      >      >     for all practical purposes that we use it for,
>     the two
>      >     values
>      >      >     of min
>      >      >      >     delay that differ by few microsecond can be
>     treated as
>      >     same
>      >      >     without any
>      >      >      >     loss of functionality. So it's about the required
>      >     normalization
>      >      >      >     interval
>      >      >      >     - something that can be controlled by the user.
>      >      >      >
>      >      >      >     thanks,
>      >      >      >     Peter
>      >      >      >
>      >      >      >
>      >      >      >
>      >      >      >      >
>      >      >      >      >
>      >      >      >      >>> Please note that I’m NOT recommending that we
>      >     back away.
>      >      >      >     Rather, we should seek to solve the
>     long-standing issue of
>      >      >      >     oscillatory routing.
>      >      >      >      >>
>      >      >      >      >> not that I disagree. History tells us that
>     the generic
>      >      >     case of
>      >      >      >     oscillation which is caused by the traffic
>     itself is a
>      >     hard
>      >      >     problem
>      >      >      >     to solve.
>      >      >      >      >
>      >      >      >      >
>      >      >      >      > Any oscillation is difficult to solve.  Positive
>      >     feedback
>      >      >      >     certainly can exacerbate the problem. But
>     solving hard
>      >      >     problems is
>      >      >      >     why we are here.
>      >      >      >      >
>      >      >      >      > Yours in control theory,
>      >      >      >      > Tony
>      >      >      >      >
>      >      >      >      >
>      >      >      >      >
>      >      >      >
>      >      >
>      >
>