Internal load balancer

In addition to balancing requests from users to front-end servers, you can use Azure Load Balancer to distribute traffic from front-end servers evenly among back-end servers.

In your healthcare organization, front-end servers call business logic services hosted on a middle tier. You want to ensure that the middle tier is as scalable and resilient as the front end. You want to use a load balancer to distribute requests from the front-end servers evenly among the middle-tier servers. This way, you can scale out the middle-tier servers to achieve the highest capacity possible. You also ensure that the middle tier is resilient to failure. When a server fails, the load balancer automatically reroutes traffic to another server.

Here, you learn how to use load balancers to distribute internal traffic.

Configure an internal load balancer

In the healthcare-portal scenario, a web tier handles requests from users. The web tier connects to databases to retrieve data for users. The database tier is also deployed on two virtual machines. To allow the front-end web portal to continue to serve client requests if a database server fails, you can set up an internal load balancer to distribute traffic to the database servers.

You can configure an internal load balancer in almost the same way as an external load balancer, but with these differences:

• When you create the load balancer, select Internal for the Type value. When you select this setting, the load balancer's front-end IP address isn't exposed to the internet.
• Assign a private IP address instead of a public IP address for the load balancer's front end.
• Place the load balancer in the protected virtual network that contains the virtual machines you want to handle the requests.

The internal load balancer should be visible only to the web tier. All the virtual machines that host the databases are in one subnet. You can use an internal load balancer to distribute traffic to those virtual machines.

Choose the distribution mode

In the healthcare portal, the application tier is stateless, so you don't need to use source IP affinity. You can use the default distribution mode of a five-tuple hash. This mode offers the greatest scalability and resilience. The load balancer routes traffic to any healthy server.