[c-nsp] EVPN Book/paper recommendation

adamv0025 at netconsultings.com adamv0025 at netconsultings.com
Sun Jul 15 05:10:13 EDT 2018


> Pete Lumbis
> Sent: Saturday, July 14, 2018 8:41 PM
> 
> Dinesh Dutt, the co-author of VxLAN wrote two books you can get for free*
> They are both focused on the datacenter, but the principals are the same
for
> both DC and non-DC use cases.
> 
> BGP in the datacenter: http://cumulusnetworks.com/bgp EVPN in the
> datacenter:
> https://cumulusnetworks.com/lp/evpn-data-center-
> oreilly/?utm_source=social+media&utm_term=EVPN&utm_campaign=2018
> +EVPN+in+the+data+center+eBook
> 
> * Behind a regwall. Disclaimer: I work for Cumulus
> 
//rant//
I'd love to ask Dinesh why VXLAN???  
-I mean what was wrong with MPLS labels for these DC folks?
If you think about it VXLAN is like VPN label directly on top of IP (just a
tenant separator).
But there's no concept of stacking VXLAN headers unlike in MPLS, -that is no
extensibility for other applications (e.g. TE = source routing = service
chaining, or other example would be micro services = VPNs),    
The EVPN CP for VXLAN is just another patch borrowed from SP folks, they
still need ACLs for micro services and god knows what they will come up with
for service chaining (PBR I guess) both utterly un-scalable solutions.
Looks like DC folks have the tendency of reinventing the wheel again and
again. 
All of these DC "solutions" could have been solved by simple well
established MPLS:  FabricPath, TRILL, LISP, VXLAN, NVGRE, OTV and Shortest
Path Bridging (SPB). 
Not mentioning all the crazy complexity where all these need to interface
with MPLS backbone at the DC boundaries (while preventing L2 loops).

Finally with the latest generation of nexus switches they seem to finally
realize that MPLS is the path.
//rant//

adam

netconsultings.com
::carrier-class solutions for the telecommunications industry::



More information about the cisco-nsp mailing list