Re: [Detnet-dp-dt] high level questions on detnet dataplane

["jouni.nospam" <[e-mail]>]

Hi,

> On Feb 22, 2017, at 3:17 AM, Carlos Jesús Bernardos Cano <[e-mail]> wrote:
> 
> Hi Loa,
> 
> On Wed, 2017-02-22 at 18:32 +0800, Loa Andersson wrote:
>> Folks,
>> 
>> two questions, the first because I've be taling it for granted, the 
>> second brought up bu two indenpendent people I talked about DetNet
>> with.
>> 
>> - is a design criteria that the existing MPLS control plane protocols
>>    shall be able to use with the DetNet dataplane?
>> 
>> - I've heard the opinion (from two different directions) that doing
>>    replication/elimination other than at the egress/ingress will make
>> the
>>    solution to complicated. Do we have a good motiation why we need
>> it?
> 
> I also have this question in mind. I've checked draft-ietf-detnet-dp-
> alt-00 and found this:
> 
> "
> 4.5.  #5 Flow duplication and merging
> 
> [...]
> 
>    The solution alternative has to provide means for end systems, relay
>    and edge nodes to be able to duplicate packets into duplicate flows,
>    and later merge the flows into one for duplicate elimination.  The
>    duplication and merging may take place at multiple points in the
>    network in order to ensure that one (or more) equipment failure
>    event(s) still leave at least one path intact for a deterministic
>    networking flow.  The goal is again to enable hitless 1+1 protection
>    in a way that no packet gets lost or there is no ramp up time when
>    either one of the paths fails for one reason or another.
> "
> 
> Not completely clear to me why we need to allow duplication and merging
> at multiple points. 

One reason is to be equally good as the IEEE 802.1 counter part.

> I have also another question: are we considering somehow the
> requirements that some detnet applications (e.g., crosshauling) might
> require bidirectional flows to be symmetric?

There are already mechanisms that you can use to create bidirectional tunnels at the transport network level. How the control plane comes up with a DetNet flow pinning in a way that it only has bidirectional flows has not been worked on yet - it is a separate work.  The data plane will/must allow that, though.

- JOuni


> 
> Thanks,
> 
> Carlos
> 
>> 
>> /Loa