08-28-2018 10:14 PM
Quick Synopsis
I have a customer that is currently running a L3 MPLS network across the United States. We are looking to convert this entire network to a L2 VPLS solution as its more cost effective for higher bandwidth throughput.
Current Layout
# of Sites 52
- 48 are branch sites with a 20-50Mbps link
- 3 sites are DCs with 150Mbps Links
- 1 Site is HQ with 100Mbps Link.
-Every site has a single router running a private AS that peers with the L3 MPLS Provider via ebgp.
-Each site also has a backup cable modem connected to the router that creates a GRE over IPSEC tunnel to the DCs with a padded AS in the event the primary L3 fiber link fails.
-Each site is running BFD at 300ms to the L3 MPLS BGP neighbor so a SIP call can fail over to the cable modem with less than 1 ping loss.
-AS-Override is NOT enabled in our provider VRF as every site has a unique private AS number
-the 2 DCs have a dedicated router that maintains all the P2P GRE tunnels via the backup cable modems at each site and announces a default route when their primary link fails and also injects that remote sites routes at the DCs and is then announced out to the L3 MPLS provider so other branch sites can continue to talk to the failed site.
New Design
Would like to have one L2 broadcast domain using a VPLS solution that each sites would connect to via a L3 interface (local site router). So the provider MAC table would only have 52 entries.
Questions
- How many sites is recommended to run in one broadcast domain? or how many has someone deployed successfully using a L3 interface at each site?
- What routing protocol assuming we would could expand to 75-100 sites over 36 months?
If I deployed OSPF should I worry about hitting a 2MB BUM filter the provider sets. (assume links could flap often)
if I deployed iBGP and used two route reflectors at the DCs I would have to use the same AS at each branch sites. I have concerns of setting up the failover GRE as it would also be peering iBGP to a DC router with the same AS. (loop issues)?
if I deployed eBGP how do I get around not peering with every other site and not have all traffic pass through a route reflector.
I would like every site to talk directly to each other and not have to pass though a DC. (No Hub and Spoke Model)
Any advise or suggestions would be much appreciated!
09-13-2018 09:22 PM
Hi Socom0810
Sorry for could not help you your question.
Could you share me how the cost decrease and we have more bandwidth if you use L2 VPLS?
Thank you
Quan
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide