Hi,Robert:

Aijun Wang
China Telecom

> On Dec 2, 2021, at 17:55, Robert Raszuk <[email protected]> wrote:
> 
> 
> Peter,
>  
>> Pulse does not replace the overlay protocols functionality. They are the 
>> ultimate source of the trust.
> 
> The entire point of PULSE is to speed up the process of service and 
> connectivity restoration. You can't just back off from it now and say that it 
> is all up to overlay to discontinue using single path to a service. 
>  
>> Pulse is just an indication that something 
>> may be going wrong an overlay protocols may switch to alternate source 
>> if there are any.
> 
> Too many "may(s) ..."  here.
> 
>> > If I receive a PULSE stating that my PE is down why would I continue to 
>> > pump Gb/s data to such PE even if there is only single BGP path to it.
>> 
>> I don't see a problem. Without the pulse that is what will happen 
>> anyway, until the overlay detects the loss of the service.
> 
> That is not the point. 
> 
> If we are to standardize something in this space it is IMO much better for it 
> to be rock solid and useful. 
> 
> Not just a lightweight indication that PE may perhaps went down, but we are 
> not sure. 
>  
>> > Absolutely not. I cam have control plane issue on the ABRs while data 
>> > plane works fine. Needless to say while all PEs work fine as well.
>> 
>> if you have control plane issue on ABR you are busted anyway, we are not 
>> after that case.
> 
> Nope I am not busted if ABR's data plane still works fine and summary 
> attracts traffic to it. If control plane dies summary will be removed but my 
> end to end service will not be impacted at all (it will continue happily via 
> other ABRs to the area). 
[WAJ] The PE should only switchover the overlay service when it receives the 
“DOWN” message from all of its ABRs.
> 
> But if ingress PE implementation will stop the service based on the fake 
> PULSE while service layer has no clue about it that is pretty nasty.
> 
> Cheers,
> R.
> 
> PS. And there is also the risk of insider attack injecting those PULSES - but 
> I am sure if this topic continues solid security review will take care of it 
> :)  Yes I know - if you let anyone inject anything like this to your IGP or 
> BGP you have bigger problems but things happen ... sometimes by unplanned 
> accidents. 

[WAJ] Same as above explanation, then only one accident or injecting “DOWN” 
will not trigger the switchover of overlay service.

> 
> _______________________________________________
> Lsr mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/lsr
_______________________________________________
Lsr mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/lsr

Reply via email to