Re: [Pce] Comments on draft-chen-pce-forward-search-p2p-path-computation-00.txt

JP Vasseur <jpv@cisco.com> Mon, 28 March 2011 20:36 UTC

Mime-Version: 1.0 (Apple Message framework v1082)
Content-Type: text/plain; charset="us-ascii"
From: JP Vasseur <jpv@cisco.com>
In-Reply-To: <5316A0AB3C851246A7CA5758973207D40493D394@dfweml503-mbx.china.huawei.com>
Date: Mon, 28 Mar 2011 22:38:07 +0200
Content-Transfer-Encoding: quoted-printable
Message-Id: <D5E3215B-1F72-4036-9E75-0EB8F78F6CDD@cisco.com>
References: <5316A0AB3C851246A7CA5758973207D40493D394@dfweml503-mbx.china.huawei.com>
To: Huaimo Chen <huaimo.chen@huawei.com>
Cc: "pce@ietf.org" <pce@ietf.org>
Subject: Re: [Pce] Comments on draft-chen-pce-forward-search-p2p-path-computation-00.txt
Precedence: list

Hi,

On Mar 28, 2011, at 10:32 PM, Huaimo Chen wrote:

> Hi JP,
> 
>    Thanks for your comments!
>    My answers are inline below.
> 
> Best Regards,
> Huaimo
> -----Original Message-----
> From: JP Vasseur [mailto:jpv@cisco.com] 
> Sent: Monday, March 28, 2011 3:36 AM
> To: Huaimo Chen
> Cc: pce@ietf.org
> Subject: Comments on draft-chen-pce-forward-search-p2p-path-computation-00.txt
> 
> Few comments:
> 1) What makes you think that you cannot enforce the set of domains that you want to explore with BRPC 
> using the IRO object specified in RFC5441?
> 
> Huaimo: This seems that there are some misunderstandings. I did not say that we could not enforce the sequence of domains or the set of domains that we want BRPC to explore using the IRO object. What I say is that the sequence of domains needs to be provided to BRPC in advance. 
> 

Not quite either, you could rely on BGP and furthermore this is orthogonal to the backward or
forward path computation aspect

> 2) If you have a mesh of domains and you try to compute the shortest inter-domain LSP according to some
> metric, you just cannot afford not to explore all of them (unless you make use of some super-TED comprising 
> resources information on all domains). This is true whether you use of forward or backward path computation ?
> 
> Huaimo: My understanding is that BRPC will explore all the possible sequence of domains, compute the shortest path for each sequence of domains and select the shortest one.

Not really, you could restrict the search or explore all of them.

> The forward search method works differently. It searches the shortest path in the way same as the normal CSPF does. It searches the special topology once and gets the shortest path. It will not explore all the possible sequence of domains.
> 
> I am trying to understand your claim about why forward computation would solve these problems ?
> 
> Huaimo: In the sense of that the forward search computation does not need any sequence of domains in advance, it guarantees that the path found is shortest, and it is efficient even for a topology of meshed domains.

But if you want to find the shortest constrained path and we have a mesh of domains, you still need to explore them all.

Thanks.

JP.

> 
> Thanks.
> 
> JP.

[Pce] Comments on draft-chen-pce-forward-search-p… JP Vasseur
Re: [Pce] Comments on draft-chen-pce-forward-sear… JP Vasseur
Re: [Pce] Comments on draft-chen-pce-forward-sear… Huaimo Chen
Re: [Pce] Comments on draft-chen-pce-forward-sear… Huaimo Chen