“OSPF is fast and automatic, but BGP gives you surgical policy control over multi-tenant traffic paths. In enterprise SDDC architecture, BGP is the clear winner.”

At IBM, when designing connectivity between VMware NSX Tier-0 Gateway Edges and physical Arista/Cisco ToR switches, client teams often debated whether to deploy OSPF or eBGP.

Protocol Comparison in Overlay Environments

While OSPF dynamically discovers neighbors using multicast (224.0.0.5), it lacks granular administrative distance, BGP community tagging, and AS-Path manipulation required for multi-tenant isolation.

• OSPF (Link-State): Excellent for flat, single-tenant internal routing, but floods LSAs across the entire area when virtual subnets flap.
• eBGP (Path-Vector): Explicit TCP unicast peering (port 179). Supports BGP Community attributes (NO_EXPORT, LOCAL_PREF) to control traffic steering per tenant pod.

[!NOTE] Use eBGP between NSX Tier-0 Gateways and physical L3 switches, reserving OSPF strictly for internal physical leaf-spine underlay routing.

! # NSX-T Tier-0 BGP Neighbor Policy
router bgp 65001
 bgp router-id 192.168.200.2
 neighbor 192.168.200.1 remote-as 65000
 neighbor 192.168.200.1 send-community standard
 neighbor 192.168.200.1 route-map TENANT-A-OUT outbound

route-map TENANT-A-OUT permit 10
 set community 65000:100

The Verdict

Key Takeaway

Standardize on eBGP for SDDC Overlay Gateway Peering.

Deploy eBGP at the SDDC border to leverage BGP community tags, AS-Path prepending, and deterministic policy control across multi-tenant environments.