Your endpoints still connect through classic ethernet to the leafs. In most practical implementations, we need redundancy between the endpoints and the leaf switches. So, we typically leverage vPC to provide us MC-LAG. vPC allows an endpoint connects to two leafs. The endpoint sees those switches as a single switch to which connected via a […]
Users connect to our data centers above the application layer. They are typically external to data centre resources. In this post, we will discuss how to connect our VXLAN-based data centre to an external Layer 3 network as well as the external connectivity to a Layer 2 network. Layer 3 could involve: the Internet, WAN, […]
From the previous posts, we learned that in order to differentiate between routes in the MP-BGP table, we include a 64-bit RD value along with the MAC address. In this post, we will discuss what else other than MAC address can propagate through BGP EVPN. Different route types carry BGP EVPN information. RFC 7432 defines […]
Integrated Route and Bridge (IRB) simply means first-hop routing. We have two different semantics for IRB: Asymmetric IRB: We have one L2VNI for each subnet we want to route between. Symmetric IRB: We have one L3VNI associated with the VRF which is used for all routed traffic Asymmetric IRB For now, forget the VXLAN and […]
In this section, we are going to discuss MP-BGP VXLAN with PIM to handle multi-destination traffic. So, your complete bipartite graph must support PIM. This is why in this post, we happened to choose loopback 254 on the spines for our PIM Anycast RP. VTEP V1 then sends out the packet toward the IP core. […]
In this post, I am going to explain how the location-identity mapping database populates using MP-BGP. Consider the image below where VTEP01 takes the original ARP L2 frame from the HOST01. VTEP01 learns IP-MAC binding of the end host. BGP Update sends the information to the Route-Reflector (iBGP), which in turn forwards this update message to […]
In the last post, we talk about F&L with Ingress Replication (AKA head-end replication) to handle multi-destination traffic. In this section, we are going to discuss F&L with PIM to handle multi-destination traffic. So, your complete bipartite graph must support PIM. This is why in this post, we happened to choose loopback 254 on the […]
In this post, I am going to explain how the location-identity mapping database populates using Flood and Learn method. Consider the image below where VTEP01 takes the original ARP L2 frame from the HOST01 and then adds VXLAN header which includes L2VNI 20100 (think of it as an overlay L2 interface between VTEPs). VTEP01 will forward […]
Underlay Transport for VXLAN In the first section of our VXLAN course we talked about overlay, underlay, and some terminologies with this technology. By now, you know what an underlay is. Within our datacenter context the underlay is our Clos network which includes the leafs and spines and their interconnections. To get rid of STP, […]
VXLAN Improvement to Datacenter Fabric In this section, I am going to talk about data center fabric journey and how Virtual eXtensible LAN emerged. I will start with STP, then vPC and PortChannel. I won’t talk about FabricPath. We will finally jump in to VXLAN in Clos networks. STP Drawbacks Let’s architect our Clos topologies […]
Elissa Garlin
Think of a news blog that’s filled with content hourly on the day of going live.