RE: Size of CR in CRH

[Andrew Alston <[e-mail]>]

Just a few comments on this.

Regarding the SID Size, and basing this on our rather extensive SR-MPLS deployments.  What we have found is that with many many hardware devices limiting the maximum SID size in MPLS to 64k we were running into some operational complexities on our side - nothing that could not be solved. 

It was not the straight up number of devices that was the problem - is was the fact that when running a single consolidated network across multiple countries, what we wanted was to assign a block of SID's to each country out of the same SRGB.  This allows each country operation to know exactly what block they are allocating out of and avoids potential problems of trying to manage all SID allocation from group level.  

This also allowed us to align our SRMS mappings correctly with the address allocations within the specific country.

However, this then creates a situation where you are effectively running into the same problem as address de-aggregation, and because you want those blocks to allow for future expansion etc, you can burn through 64k pretty damn fast when operating across 15 countries.

Now, if I look at CRH - we see a similar thing - where a large network wants to allocate blocks of CRH SID's on a per country basis, yet you are running a single consistent overarching block across the network, the 64k limit is actually a lot smaller than one would imagine.  I would however say that a 16bit CRH SID has a lot of application for smaller networks.

While it is very certainly possible in our scenario to make things work with a 16bit CRH SID, its our view that it would substantially increase complexity from an operational perspective.  Hence, while I certainly support the dual SID sizes, for me, if we had to drop one of them, it would be my preference to stay with the 32bit SID.  As I said though, if the WG does decide either pre or post adoption of the draft to go with one rather than two, I'm not hugely religious about which at t his point, though I see the need for something larger than 16bit becoming very apparent in the future.

Just by way of illustration - if we were to run networks across the entire African continent - which I hope to see us doing one day - as a single network, and we split the 64k block up per country, that only leaves 1213 CRH SID's per country - we use more than that in the SR-MPLS domain in at least 3 countries today.

Andrew