SLB Network Architecture Overview
From LB Wiki
If you remember nothing else about how load balancers fit into a network, remember this: It all boils down to two things, making sure that traffic hits the load balancer on the way in, and making sure it hits the load balancer on the way out.
One of the trickiest aspects of load balancing for people (networking and application alike) to wrap their heads around is how the load balancer actually fits into a given network infrastructure. What changes need to be done to the servers? What changes need to be done to the network?
No matter what method used to deploy a load balancer, two things need to happen:
- Traffic goes through the load balancer on the way in
- Traffic goes through the load balancer on the way out
The only exception to this is DSR (Direct Server Return), where the web server is tricked into doing the load balancer's work on the way out.
Traffic on the way in
Getting traffic to go through the load balancer on the way in is easy. The load balancer holds a VIP (Virtual IP) address, where traffic is pointed to (typically through DNS).
Figure 3: VIP ensures traffic passes through the load balancer on the way in
Traffic on the way out
This is the tricky part in putting the load balancer into the network. We need to make sure traffic hits the load balancer on the way out. This can be done with one of two primary methods:
- By being in the Layer-2 Path (also called bridge-path, bridge-mode, etc.)
- By being in the Layer-3 Path (also called route-path, route-mode, etc.)
