Hello everyone, I would like to hear from you about MPLS and VXLAN.
Well today I have a network with some scattered CCRs, I do a lot of lan2lan service for my customers, I do it today using MPLS and VPLS for that.
Well, after reading and researching a little, I saw that maybe, correct me if I’m wrong, but from what I saw VXLAN for lan2lan would only be better than MPLS and VPLS.
But I saw that it adds an overhead of 50bytes, so my doubt would be like this, let’s say I have to serve 20 clients in lan2lan, that I will have to close 20 VXLAN tunnels from one router to another only, let’s say, this is an example.would it be an overhead of 20x50bytes?
This is in the case of L2MTU correct? Today I use 1500 MTU and 1896 L2MU interfaces, could someone give me an explanation, thanks.
hello,
let’s say I have to serve 20 clients in lan2lan, that I will have to close 20 VXLAN tunnels from one router to another only, let’s say, this is an example.would it be an overhead of 20x50bytes?
I’m not too well versed in vxlan since I had retired long enough to forget the industry.
but, afaik, i think you have been a bit confused to think vxlan as new approach to site to site vpn - as replacement to mpls/vpls as pure transport layer mechanism.
i would think of vxlan as a tool to interconnect dispersed data centers to be as a single connected cluster (as a single connected subnet).
while mpls/vpls - although they bear the name l2/l3vpn - were mainly focused in transport layer.
cmiiw.
hope this helps.
I worked for an org that deployed VXLAN/EVPN underlay for L2 over L3 transport and for providing L2 adjacency over different racks/sites. We used Juniper/Cumulus/Nokia, and we used 9216 L2 MTU and L3 MTU as 9000, network-wide.
So we never had MTU problems, all hosts can talk to each other at 9000 MTU. And performance wise, it didn’t matter as the underlying hardware supported line-rate forwarding for VXLAN/EVPN or MPLS/VPLS if we wanted to use the latter.
In MikroTik, neither are offloaded to ASIC.
Yet…
in MikroTik, neither are offloaded to ASIC.
speaking of which.. has anyone tried to port routeros to an onie-based bare metal switches?
like those from netberg, edge-core?