A load balancer distributes incoming requests across multiple server instances to reduce bottlenecks and support availability.

Common algorithms

• Round robin: send requests in sequence across eligible servers.
• Weighted round robin: assign more traffic to servers with higher weights.
• Least connections: prefer fewer active connections, depending on implementation.
• IP hash: select a backend based on client IP information.

Health checks and benefits

• Health checks help remove unhealthy instances from rotation.
• Traffic can be distributed across instances.
• Rolling maintenance and deployments can be supported with sufficient healthy capacity.
• Redundancy can improve availability.

Design considerations

• Avoid relying on a single server for session state.
• Use meaningful readiness checks.
• Database and other dependencies must also have enough capacity.
• A load balancer itself may need redundancy.
• Retries can amplify overload if poorly configured.