Re: [CCAMP] draft-ali-ccamp-lsp-inquiry-00

"Matt Hartley (mhartley)" <mhartley@cisco.com> Wed, 07 August 2013 16:06 UTC

Return-Path: <mhartley@cisco.com>
X-Original-To: ccamp@ietfa.amsl.com
Delivered-To: ccamp@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id B062E21F9EB6 for <ccamp@ietfa.amsl.com>; Wed, 7 Aug 2013 09:06:30 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -10.598
X-Spam-Level:
X-Spam-Status: No, score=-10.598 tagged_above=-999 required=5 tests=[BAYES_00=-2.599, HTML_MESSAGE=0.001, RCVD_IN_DNSWL_HI=-8]
Received: from mail.ietf.org ([12.22.58.30]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id yG8JRDYPHUlt for <ccamp@ietfa.amsl.com>; Wed, 7 Aug 2013 09:06:16 -0700 (PDT)
Received: from rcdn-iport-9.cisco.com (rcdn-iport-9.cisco.com [173.37.86.80]) by ietfa.amsl.com (Postfix) with ESMTP id ECCB321F9E88 for <ccamp@ietf.org>; Wed, 7 Aug 2013 09:06:15 -0700 (PDT)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=cisco.com; i=@cisco.com; l=61731; q=dns/txt; s=iport; t=1375891576; x=1377101176; h=from:to:cc:subject:date:message-id:references: in-reply-to:mime-version; bh=TlaFiOu9yuS9xllQpZaqmhs//4PhdGUsCJnjBtTuD8U=; b=GQ4UmDmO/NyX7hhEUFme/tZerjco4RkhG1hZTuigqBMx549Esl+5Nk4o MS2IMA+8KDF03KcRbr7EEW6h1sykLFY1T7bgKrFr2IXy4Can883fHG9ps Wk3UUABe2gQHu98rKmIrjzU0zeDfpbn+u+zp4YWeQxBZAmhj28gks6SZM 4=;
X-IronPort-Anti-Spam-Filtered: true
X-IronPort-Anti-Spam-Result: AhwFAMBvAlKtJV2d/2dsb2JhbABbgkJENVC+RoEcFnSCJAEBAQQaE0UHEAIBCBEDAQEBCxYBBgcyFAkIAgQBDQUIEYd3uDaOVAqBCw0TEQYBAoMYdAOIc6A9gxeBaAkXIg
X-IronPort-AV: E=Sophos; i="4.89,834,1367971200"; d="scan'208,217"; a="241621873"
Received: from rcdn-core-6.cisco.com ([173.37.93.157]) by rcdn-iport-9.cisco.com with ESMTP; 07 Aug 2013 16:06:14 +0000
Received: from xhc-rcd-x13.cisco.com (xhc-rcd-x13.cisco.com [173.37.183.87]) by rcdn-core-6.cisco.com (8.14.5/8.14.5) with ESMTP id r77G6EN6017092 (version=TLSv1/SSLv3 cipher=AES128-SHA bits=128 verify=FAIL); Wed, 7 Aug 2013 16:06:14 GMT
Received: from xmb-rcd-x03.cisco.com ([169.254.7.202]) by xhc-rcd-x13.cisco.com ([173.37.183.87]) with mapi id 14.02.0318.004; Wed, 7 Aug 2013 11:06:14 -0500
From: "Matt Hartley (mhartley)" <mhartley@cisco.com>
To: "Zafar Ali (zali)" <zali@cisco.com>, Gert Grammel <ggrammel@juniper.net>, Daniele Ceccarelli <daniele.ceccarelli@ericsson.com>, Fatai Zhang <zhangfatai@huawei.com>, "Margaria, Cyril (Coriant - DE/Munich)" <cyril.margaria@coriant.com>, Khuzema Pithewan <kpithewan@infinera.com>, "CCAMP (ccamp@ietf.org)" <ccamp@ietf.org>
Thread-Topic: [CCAMP] draft-ali-ccamp-lsp-inquiry-00
Thread-Index: Ac6NL9ROMXULLvcwR2aMbEKex7ZPbAAG5uQwAAdbOAAAAUWKEAAHaZGAAA2DHKAABMFSAAABVtHgAAKruoAAAb6xQACeQ0EQAAiR5UAABYt+gACGkgXQADOIbAAAAED5UA==
Date: Wed, 7 Aug 2013 16:06:13 +0000
Message-ID: <9D50FCE7413E3D4EA5E42331115FB5BC105A7491@xmb-rcd-x03.cisco.com>
References: <3b003053a8884105a66387e1bb6b3521@BN1PR05MB041.namprd05.prod.outlook.com> <B6585D85A128FD47857D0FD58D8120D30E9FDF8D@xmb-rcd-x14.cisco.com>
In-Reply-To: <B6585D85A128FD47857D0FD58D8120D30E9FDF8D@xmb-rcd-x14.cisco.com>
Accept-Language: en-US
Content-Language: en-US
X-MS-Has-Attach:
X-MS-TNEF-Correlator:
x-originating-ip: [161.44.212.251]
Content-Type: multipart/alternative; boundary="_000_9D50FCE7413E3D4EA5E42331115FB5BC105A7491xmbrcdx03ciscoc_"
MIME-Version: 1.0
Subject: Re: [CCAMP] draft-ali-ccamp-lsp-inquiry-00
X-BeenThere: ccamp@ietf.org
X-Mailman-Version: 2.1.12
Precedence: list
List-Id: Discussion list for the CCAMP working group <ccamp.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ccamp>, <mailto:ccamp-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/ccamp>
List-Post: <mailto:ccamp@ietf.org>
List-Help: <mailto:ccamp-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ccamp>, <mailto:ccamp-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 07 Aug 2013 16:06:31 -0000

Gert, Fatai,

Further to Zafar's reply...

You're correct that a UNI-C could, in principle, make repeated inquiries that will never result in a path being found. However, hopefully a network operator will not configure a set of constraints on a tunnel that have no chance of being met by the core network and will thus result in indefinite inquiries... and if they do, that problem can be fixed by removing the unreasonable constraints.

So, yes, this is a tool that could be mis-used or abused, with adverse consequences if that happens. But I don't think this is sufficient reason to say that a useful tool should be denied to folks who will use it properly.

Cheers

Matt

Hi Gert-

Please see in-line.

Thanks

Regards ... Zafar

From: Gert Grammel <ggrammel@juniper.net<mailto:ggrammel@juniper.net>>
Date: Tuesday, August 6, 2013 11:40 AM
To: zali <zali@cisco.com<mailto:zali@cisco.com>>, Daniele Ceccarelli <daniele.ceccarelli@ericsson.com<mailto:daniele.ceccarelli@ericsson.com>>, "Margaria, Cyril (Coriant - DE/Munich)" <cyril.margaria@coriant.com<mailto:cyril.margaria@coriant.com>>, Khuzema Pithewan <kpithewan@infinera.com<mailto:kpithewan@infinera.com>>, "ccamp@ietf.org<mailto:ccamp@ietf.org>" <ccamp@ietf.org<mailto:ccamp@ietf.org>>
Subject: RE: [CCAMP] draft-ali-ccamp-lsp-inquiry-00

Hi Zhafar,

how would that policy enforced?

By UNI-C node having a policy from operator on when it should run inquiry procedure. Usually, during maintenance window.

In the end every topology information has to be poked out of the server network and how frequent/updated this information is, is not visible to the client.

No, as this is a UNI.

When you say: <snip> If a connection cannot be restore using existing constraints, UNI-C may change the constraints depending on what is acceptable to the client layer. <snip>
How is the client able to figure out which constraint is eventually able to succeed?

By trying a list of paths that releases constraints from toughest to easiest. We are talking about a list with typically 2-4 entries.

And if it doesn't what is the client about to do?


Release the path constraints - until it finds a path or constraints cannot be released any further.

In other words, how does the client network able to understand that it is beating a dead horse (a server network that is full), and how does it understand that it actually is no more?


Obviously we are talking about trying a very limited list of path options (2-4, typically).

-Gert

From: Zafar Ali (zali) [mailto:zali@cisco.com]
Sent: 04 August 2013 00:00
To: Gert Grammel; Daniele Ceccarelli; Margaria, Cyril (Coriant - DE/Munich); Khuzema Pithewan; CCAMP (ccamp@ietf.org<mailto:ccamp@ietf.org>)
Subject: Re: [CCAMP] draft-ali-ccamp-lsp-inquiry-00

Hi Gert:

Firstly, the inquiry (or poking) is much controlled by Policy. It is not that client network will probe server network blindly. Usually this operation will only be done during maintenance window and/ or in a well controlled fashion.

Secondly, routing is not only about the objective function (changes). As client layer understands and dedicates the requirement for the client connection (SRLG and other routing constraints, metric bound, OF, etc.), client layer is in better position to dedicate what path would be acceptable should an unprotected tunnel connection goes down. If a connection cannot be restore using existing constraints, UNI-C may change the constraints depending on what is acceptable to the client layer. Please also refer to Matt's email on this.

Thanks

Regards ... Zafar

From: Gert Grammel <ggrammel@juniper.net<mailto:ggrammel@juniper.net>>
Date: Saturday, August 3, 2013 4:28 PM
To: Daniele Ceccarelli <daniele.ceccarelli@ericsson.com<mailto:daniele.ceccarelli@ericsson.com>>, "Margaria, Cyril (Coriant - DE/Munich)" <cyril.margaria@coriant.com<mailto:cyril.margaria@coriant.com>>, zali <zali@cisco.com<mailto:zali@cisco.com>>, Khuzema Pithewan <kpithewan@infinera.com<mailto:kpithewan@infinera.com>>, "ccamp@ietf.org<mailto:ccamp@ietf.org>" <ccamp@ietf.org<mailto:ccamp@ietf.org>>
Subject: RE: [CCAMP] draft-ali-ccamp-lsp-inquiry-00

Hi,

I don't think it is a good design to let clients blindly poke into the server domain. If an objective function is met, why is the client still poking the server? Is it for a different objective function? Then what is the criteria to stop poking around?

Gert

From:ccamp-bounces@ietf.org<mailto:ccamp-bounces@ietf.org> [mailto:ccamp-bounces@ietf.org] On Behalf Of Daniele Ceccarelli
Sent: 03 August 2013 18:28
To: Margaria, Cyril (Coriant - DE/Munich); Zafar Ali (zali); Khuzema Pithewan; CCAMP (ccamp@ietf.org<mailto:ccamp@ietf.org>)
Subject: Re: [CCAMP] draft-ali-ccamp-lsp-inquiry-00

Hi Cyril,

Your idea is a bright workaround but I would not bind the capability of setting up inquiry LSPs to the capability of supporting priorities and supporting the capability of changing priorities as I suppose that, if you decide to commit the resources of the inquiry LSP, I guess you'd like to inherit the priorities of the LSP you're re-optimizing.

Re the Admin status object VS the LSP_attributes I think the choice depends on the relationship between the existing LSP and the inquiry LSP. If the inquiry LSP is a new one I would suggest to use the Admin status (it is not so different from e.g. the exercise status), while if the two LSP are linked, maybe the LSP_attributes is more appropriate.

My 2 cents
Daniele

PS. I have the suspect that we turning RSVP-TE into a management protocol between a client and a server networks...

From:ccamp-bounces@ietf.org<mailto:ccamp-bounces@ietf.org> [mailto:ccamp-bounces@ietf.org] On Behalf Of Margaria, Cyril (Coriant - DE/Munich)
Sent: mercoledì 31 luglio 2013 13:24
To: Zafar Ali (zali); Khuzema Pithewan; CCAMP (ccamp@ietf.org<mailto:ccamp@ietf.org>)
Subject: Re: [CCAMP] draft-ali-ccamp-lsp-inquiry-00

Hi,


The LSP inquiry without resource locking (at a given priority then) mechanism is present not reserve the resource for other LSPs (of same or higher priority) or preempt lower priority.

Could this be achieved by setting the LSP with a lower holding priority, signal is as pre-planned AND indicate that this should not preempt other LSPs.
When this is instantiated a new LSP (or the same ) can be signaled with SE and explicit ERO with the changed flag.


Mit freundlichen Grüßen / Best Regards
Cyril Margaria
From: Zafar Ali (zali) [mailto:zali@cisco.com]
Sent: Wednesday, July 31, 2013 12:54 PM
To: Margaria, Cyril (Coriant - DE/Munich); Khuzema Pithewan; CCAMP (ccamp@ietf.org<mailto:ccamp@ietf.org>)
Subject: Re: [CCAMP] draft-ali-ccamp-lsp-inquiry-00

Hi Margaria-

Resource reservation (in control plane) are always associated with a given priority. The inquiry LSP should be signaled using the same setup and hold priority as the currently active LSP. Changing priority of inquiry LSP to 7 (lowest) will cause incorrect blocking for the inquiry LSP (as resource may be available at the priority of the LSP but may not be available at the 7 (lowest) priority.

Thanks

Regards ... Zafar

From: <Margaria>, "Cyril (Coriant - DE/Munich)" <cyril.margaria@coriant.com<mailto:cyril.margaria@coriant.com>>
Date: Wednesday, July 31, 2013 6:46 AM
To: zali <zali@cisco.com<mailto:zali@cisco.com>>, Khuzema Pithewan <kpithewan@infinera.com<mailto:kpithewan@infinera.com>>, "ccamp@ietf.org<mailto:ccamp@ietf.org>" <ccamp@ietf.org<mailto:ccamp@ietf.org>>
Subject: RE: [CCAMP] draft-ali-ccamp-lsp-inquiry-00

Hi,

For the resource reservation aspect, this seems related to priorities, LSP inquiry without reservation could be defined as LSP with setup, holding priority 8 (or 255). This would in addition allow for reporting when the resource are gone (preempted)

Mit freundlichen Grüßen / Best Regards
Cyril Margaria
From:ccamp-bounces@ietf.org<mailto:ccamp-bounces@ietf.org> [mailto:ccamp-bounces@ietf.org] On Behalf Of Zafar Ali (zali)
Sent: Wednesday, July 31, 2013 11:00 AM
To: Khuzema Pithewan; CCAMP (ccamp@ietf.org<mailto:ccamp@ietf.org>)
Subject: Re: [CCAMP] draft-ali-ccamp-lsp-inquiry-00

Hi Khuzema:

Please see in-line.

Thanks

Regards ... Zafar

From: Khuzema Pithewan <kpithewan@infinera.com<mailto:kpithewan@infinera.com>>
Date: Wednesday, July 31, 2013 3:45 AM
To: zali <zali@cisco.com<mailto:zali@cisco.com>>, "ccamp@ietf.org<mailto:ccamp@ietf.org>" <ccamp@ietf.org<mailto:ccamp@ietf.org>>
Subject: RE: [CCAMP] draft-ali-ccamp-lsp-inquiry-00

Hi Zafar,

The point I am making here is.. the 2 approaches.. Admin Status and LSP_Attributes, are exactly same in terms of object re-use and both of them defines new bits for enhanced functionality. The LSP_Attribute approach has additional overhead of managing a separate control LSP, which is not desirable.


The inquire/ potential reopt LSP is likely not to follow path of the currently active LSP. Hence this cannot be implemented by just adding some Admin Status bit on the current LSP. One need to signal a separate LSP.

Thanks
Khuzema

From: Zafar Ali (zali) [mailto:zali@cisco.com]
Sent: Wednesday, July 31, 2013 2:17 AM
To: Khuzema Pithewan; CCAMP (ccamp@ietf.org<mailto:ccamp@ietf.org>)
Subject: Re: [CCAMP] draft-ali-ccamp-lsp-inquiry-00

Hi Khuzema:

For signaling inquiry LSP with resource locking, we are using the Pre-Planned LSP flag as-is as defined in RFC6001. Given this, we are defining a new flag when inquiry LSP needs to be signal without resource locking.

Thanks

Regards ... Zafar

From: Khuzema Pithewan <kpithewan@infinera.com<mailto:kpithewan@infinera.com>>
Date: Tuesday, July 30, 2013 5:45 PM
To: zali <zali@cisco.com<mailto:zali@cisco.com>>, "ccamp@ietf.org<mailto:ccamp@ietf.org>" <ccamp@ietf.org<mailto:ccamp@ietf.org>>
Subject: RE: [CCAMP] draft-ali-ccamp-lsp-inquiry-00

Well.. not really.

You are defining new bits for LSP_ATTRIBUTES for resource locking... aren't you?

Instead of doing that, you can define bits in ADMIN_STATUS and save new LSP life cycle management, which would be quite cumbersome.

Regards
Khuzema

From: Zafar Ali (zali) [mailto:zali@cisco.com]
Sent: Tuesday, July 30, 2013 10:08 PM
To: Khuzema Pithewan; CCAMP (ccamp@ietf.org<mailto:ccamp@ietf.org>)
Subject: Re: [CCAMP] draft-ali-ccamp-lsp-inquiry-00

Khuzema:

The point is to reuse what already exists. The Pre-Planned LSP flag in the Attribute Flags TLV of LSP_ATTRIBUTES object is already defined in [RFC5420] and is a glove fit.

Thanks

Regards ... Zafar

From: Khuzema Pithewan <kpithewan@infinera.com<mailto:kpithewan@infinera.com>>
Date: Tuesday, July 30, 2013 1:40 PM
To: Khuzema Pithewan <kpithewan@infinera.com<mailto:kpithewan@infinera.com>>, "ccamp@ietf.org<mailto:ccamp@ietf.org>" <ccamp@ietf.org<mailto:ccamp@ietf.org>>
Subject: Re: [CCAMP] draft-ali-ccamp-lsp-inquiry-00

Another point I spoke about in the meeting ..

Why can't we extend Admin Status object to signal resource locking, checking for re-optimization. Since this operation is typically done in maintenance window by Admin, it may make sense to use Admin Status Object. Moreover, we have lots of bits available/undefined in Admin Status object.

This will save network element to manage life of additional LSP and control plane failure related issues attached to the additional LSP.

Khuzema