Stretching VLAN

[Anwar Safian]

Hi expert

There are lot of information about how could we stretch VLAN including L2TPv3, VPLS, OTV etc.

However, why do we need to stretch the VLAN? Why dont we just use normal layer 3 routing to send traffic?

Any example/case of using stretching VLAN?

[Francesco Molino]

Hi

It depends on your architecture and application designs/redundancy.

For example, in datacenters, some customers will have a VM running on 1 site and if this site goes down, they want the duplicate vm on site 2 to take over using same IP. In that case you'll need to have a 'Layer 2' continuity no matter the technology (l2tpv3, otv, vxlan....)

This is 1 of many use cases.

[Anwar Safian]

Hi Francesco

Thanks for your reply.

However I wonder why we need same IP address at difference site?

Why dont use same IP address? What is advantage of using same IP address at difference site? 

[Francesco Molino]

Because for some applications, you can change the IP address anf if you want to have cold redundancy, you'll need to restart a copy of that application or VM on the second site with the same config to make sure people can still access it.

Is that more clear?

[David Castro F.]

Hello Amwar,

It all depends of what the customer has and how the customer access things, for example:

1. Scenario: You can 3 call managers, 2 on Side A and 1 on Side B, and if the customer access that using domain, then they could use GSLB and with DNS be able to access the application even when Side A goes down, now very different if they need to access an A record/FQDN or individual IP on an APP within a VM. Probably and I say probably the customer could still access that IP or A record through B side and the B side having an MPLS to the A side, which is not that ideal and causes throughput and probably the MPLS getting exhausted of traffic.

Recommendation on this scenario: If the access is through DNS that covers all the IPs, then you dont really need the IP mobility/Vmotion. But if the customer access individual FQDNs and IPs, you may need a lot of redundant ways to get to the A side even when it is down. Or make it easier by doing vmotion during a DOWN situation, this is not meant to be a permanent solution of course, it is a solution on a DOWN situation or if the traffic is not being load balanced correctly through your DCs, then moving your workloads will look for doing the balance of that.

Keep in mind that also there times where you could move your workloads (VMs) to a public cloud, and for that movement, you are not changing IPs, you need to keep the same IPs for this purpose.

Regards,

David castro,