Re: [Detnet-dp-dt] detnet LSPs and PWs

["Andrew G. Malis" <[e-mail]>]

Stewart,

In at least one implementation I’m familiar with, lack of common shared
memory between parallel microprogrammed forwarding engines for the same
physical port was the issue. I can see lack of intercard or interport
communications also being a problem.

We added the control word and sequence number to Ethernet PWs later on, and
I believe most implementations today do use it to reorder and comply with
the Ethernet spec.

Cheers,
Andy



On Mon, Feb 13, 2017 at 8:02 AM, Stewart Bryant <[e-mail]>
wrote:

>
>
> On 13/02/2017 11:58, Andrew G. Malis wrote:
>
>
> On Mon, Feb 13, 2017 at 5:55 AM, Stewart Bryant <[e-mail]>
> wrote:
>
>> Now for the elephant in the corner of all of the schemes I have seen. If
>> you have multiple paths to an X-PE, packets will likely arrive on different
>> line cards. Sequence number co-ordination amongst different line cards, and
>> at high speed even amongst different ports on the same line card is a hard
>> problem. Indeed depending on the pipeline design on the line card, ANY
>> sequence number processing can be hard. You could mitigate this (at the
>> cost of availability) by requiring a common ingress port at any detnet
>> X-PE. This would normally require an RSVP-TE or SR underlay.
>>
> I would like to second Stewart’s comment above. Back in the early days of
> PWs, some vendors implemented PWs without a sequence number because they
> couldn’t do sequence number processing across multiple parallel forwarding
> engines even on the same physical port, as the forwarding engines didn’t
> have common shared memory. Of course, their customers had to not use ECMP,
> at least for those PWs. I think most vendors have gotten past that these
> days, but you can never tell. But keep these possible limitations in mind
> as you design Detnet.
>
> Cheers,
> Andy
>
>
> Andy it is not common shared memory but the pipelining and the intercard
> communication that stops it working.
>
> As you know we prohibit ECMP on most PWs via the control word, with the
> exception of the Ethernet PW where it is optional and in place for historic
> reasons. However an Ethernet PW can re-order and as such does not fully
> comply with the Ethernet spec.
>
> The problem is still live - which is why we need the packet marking scheme
> for packet loss measurement.
>
> - Stewart
>