Re: Questions on draft-ietf-ccamp-gmpls-recovery-analysis-05.txt

[dimitri papadimitriou <dpapadimitriou@psg.com>]

adrian, see below

Adrian Farrel wrote:

> Hi,
> Posting this question on behalf of Lei Lei who has had some difficulty 
> subscribing to the list.
> Thanks,
> Adrian
> ===
> The section 8.4.1 of draft-ietf-ccamp-gmpls-recovery-analysis-05.txt 
> discusses how to realize recovery resource sharing using GMPLS protocols.
> It draws the conclusion that no further extensions of GMPLS routing 
> protocols are needed to perform this task, since the use of Maximum 
> Reservable Bandiwidth, the Unreserved Bandwidth and the Maximum LSP 
> Bandwidth is sufficient for the purpose. However, I do not quite follow 
> this idea.
> I've read the [KODIALAM1],  [KODIALAM2] and [GLI] papers, which is 
> referred to in the draft, even before I read this draft. I have no 
> problem in understanding the ideas described in those papers, In those 
> papers, the Maximum TE Link Bandwidth is divided into three parts: the 
> capacity committed for working LSPs, and capacity reserved for backup 
> LSPs, and the unallocated capacity. The Partial  Information routing 
> algorithm needs to know the above three values in order to compute the 
> optimal path. The Full Information routing algorithm in [GLI] needs 
> still more information. However, section 8.4.1 of 
> draft-ietf-ccamp-gmpls-recovery-analysis-05.txt
> claims that only the value of Maximum Reservable Bandiwidth, the 
> Unreserved Bandwidth and the Maximum LSP Bandwidth need to be avaiable 
> for recovery resource sharing, as shown in the figure below. But I do 
> not understand how.
> 1. I do not understand the relationships between the  Max Reservable
>  Bandwidth, Maximum LSP Bandwidth and the Unreserved Bandwidth.

the relationships have been defined in various GMPLS routing documents, 
the profile proposed is by definition informational and shows a way to 
use these parameters such as to obtain the stated objective

all the below questions are specific to local policy rules and subject 
to initial configuration (except the nominal link capacity which is 
determined by the interface capability)

>   1.a Why should the difference between Max Reservable Bandwidth and the
>       Maximum LSP Bandwidth be the Maximum Shareable Bandwidth?
>   1.b Why should  the difference between the Max Reservable Bandwidth and
>       the Maximum TE Link Bandwidth be the amount of bandwidth currently
>       allocated for shared recovery?
>       The Max Reservable Bandwidth should be  larger than the Maximum TE
>       Link Bandwidth, then how to set the exact value of Max Reservable
>       Bandwidth?
>       Is it decided by policy or can be calculated by some equations?
>       Is it a constant value during network operation or change over time?
>   1.c Why should the difference between the Maximum TE Link Bandwidth and
>       the Maximum LSP Bandwidth be the actual bandwidth committed for
>       shared recovery instead of the sum of the bandwidth committed for
>       working LSPs and the actual bandwidth committed for shared recovery?
> 2. What is the routing algorithm supposed to be used in this method?
>  I suppose the working LSP can be calculated with SPF algorithm on the
>  network topology where the links whose Maximum LSP Bandwidth smaller
>  than the request bandwidth are pruned? But how to calculate the optimum
>  backup path? Is it similar with the Partial Information approach? If it
>  is, I fail to see how this could be done.

additional information needed for the partial information model is to 
find a way to determine the (shared) bandwidth reserved on each link by 
protecting LSPs that have already been set up (soft-provisioning) - so 
instead here of requiring such additional information, the setup 
discussed in this section is proposing to maximize the ratio between 
over-provisioned bandwidth and actual bandwidth that can be committed by 
assuming over-provisioning is used for shared recovery purposes; the 
considerations do not detail the exact algorithm but assess that the 
available information is enough achieve this computation

>                                - ---------- Max Reservable Bandwidth
>                               |  -----  ^
>                               |R -----  |
>                               |  -----  |
>                                - -----  |max_R
>                                  -----  |
>  --------  TE link Capacity    - ------ | - Maximum TE Link Bandwidth
>  -----                        |r -----  v
>  -----     <------ b ------>   - ---------- Maximum LSP Bandwidth
>  -----                           -----
>  -----                           -----
>  -----                           -----
>  -----                           -----
>  -----                           ----- <--- Minimum LSP Bandwidth
>  -------- 0                      ---------- 0
>
> Best regards
> Lei Lei
>
>
> .