Re: [Lsr] LSR WG Adoption Poll for "Flexible Algorithms: Bandwidth, Delay, Metrics and Constraints" - draft-hegde-lsr-flex-algo-bw-con-02

["Dongjie (Jimmy)" <[e-mail]>]

Hi Peter, 

Thanks for your reply, please see inline:

> -----Original Message-----
> From: Peter Psenak [mailto:[e-mail]]
> Sent: Friday, May 14, 2021 5:12 PM
> To: Dongjie (Jimmy) <[e-mail]>; Acee Lindem (acee)
> <acee=[e-mail]>; [e-mail]
> Cc: [e-mail]
> Subject: Re: [Lsr] LSR WG Adoption Poll for "Flexible Algorithms: Bandwidth,
> Delay, Metrics and Constraints" - draft-hegde-lsr-flex-algo-bw-con-02
> 
> Hi Jie,
> 
> please see the response for one of your questions inline:
> 
> On 14/05/2021 09:52, Dongjie (Jimmy) wrote:
> > Hi authors,
> >
> > I’ve read the latest version of this document and have the following
> > comments:
> >
> > 1.Is the generic metric type applicable to applications other than
> > Flex-Algo? If so, it is better to make this clear in the document, or
> > perhaps it may be defined separately from the Flex-Algo specific extensions?
> >
> > 2.The “Exclude Minimum Bandwidth” constraint is compared with the
> > maximum link bandwidth to exclude the links from the computation, it
> > would be helpful if there is some analysis about how much this can
> > help in traffic engineering, such as to reduce the congestion or
> > improve the link utilization. One simple example is, if multiple
> > Flex-Algos use this constraint to exclude the same set of links, this
> > may increase the possibility of congestion on the rest of the links?
> >
> > Perhaps a more general question is, what would be the benefit of
> > introducing bandwidth attribute into Flex-Algo based distributed path
> > computation?  It is known that bandwidth can be used in centralized
> > computation for efficient path placement and resource management, can
> > distributed computation with bandwidth constraint achieve the same, or
> > is there some advantages compared with centralized computation?
> >
> > 3.With the automatic metric calculation, it could introduce per
> > Flex-Algo link metric value, while the existing Flex-Algo only refers
> > to the metric of the link via metric type. Is this the expected behavior?
> > Will it be further extended to make other link attributes flex-algo
> > specific?
> 
> we need to distinguish between:
> 
> a) flex-algo application specific metric (applies to all flex-algos)
> b) flex-algo X specific metric

Yes, here I mean flex-algo number specific metric.

> 
> (a) already exists in the form of the ASLA advertisement for delay and TE-metric.
> Bandwidth metric will be no different.
> 
> (b) there has been no such thing as flex-algo X specific metric/attribute defined
> so far. And we are not defining it in this draft either. The draft defines sub-TLVs
> for automatic bandwidth metric calculation. It is the winning FAD for the
> Flex-Algorithm X that specifies whether the automatic bandwidth metric
> calculation is done or not and it would be rather complicated (certainly possible)
> to have the parameters for such calculation being advertised outside of the
> FAD.
> 
> You are right these being part of the FAD may result in the calculated bandwidth
> metric being different for each flex-algo on the same link.
> The intention was NOT to have the different per algo bandwidth metric value,
> rather it was the convenience of reusing the FAD that was the motivation for
> the existing encoding.

Understood. Perhaps some text to clarify this would be helpful, such as:

The reference bandwidth is considered a global parameter for the network, it is not expected to use different reference bandwidth for different flex-algos.

> 
> So to answer your question, we do NOT intend to make any link attributes per
> flex-algo number.

This answer is clear, thanks.

Best regards,
Jie

> 
> thanks,
> Peter
> 
> 
> 
> 
> >
> > 4.In the reference bandwidth method, the draft says it simplifies the
> > management in case the reference bandwidth needs to be changed. Since
> > the reference bandwidth applies to the metric calculation of all the
> > links in the flex-algo with the same proportion, it seems the change
> > of the reference bandwidth will not impact the result of the path
> > computation in the flex-algo. In which case the reference bandwidth
> > need to be changed?
> >
> > Best regards,
> >
> > Jie
> >
> > *From:*Lsr [mailto:[e-mail]] *On Behalf Of *Acee Lindem
> > (acee)
> > *Sent:* Thursday, May 13, 2021 5:09 AM
> > *To:* [e-mail]
> > *Cc:* [e-mail]
> > *Subject:* [Lsr] LSR WG Adoption Poll for "Flexible Algorithms:
> > Bandwidth, Delay, Metrics and Constraints" -
> > draft-hegde-lsr-flex-algo-bw-con-02
> >
> > Esteemed Members of the LSR WG,
> >
> > This begins a 2 week WG adoption call for the following draft:
> >
> > https://datatracker.ietf.org/doc/draft-hegde-lsr-flex-algo-bw-con/
> >
> > Please indicate your support or objection by May 27^th , 2021.
> >
> > Authors, please respond to the list indicating whether you are aware
> > of any IPR that applies to this draft.
> >
> > Thanks,
> >
> > Chris and Acee
> >