Five Core Load Balancing Algorithms Explained

Load balancing is a fundamental technology in distributed architecture and a must‑learn skill for advancing in backend engineering.

When a single machine cannot bear the traffic load, horizontal scaling adds multiple servers to share the load, dramatically reducing backend pressure and improving user performance. The challenge then becomes how to distribute client requests among these servers, which is solved by load‑balancing algorithms.

The article focuses on five core load‑balancing algorithms.

1. Round Robin

Round Robin assigns incoming requests to backend servers in a fixed sequential order, treating each server equally without considering current connections or load.

Suitable scenario: all application servers have identical hardware specifications.

2. Weighted Round Robin

Based on Round Robin, this method assigns weights to servers according to their hardware capabilities, directing more traffic to higher‑capacity machines and less to lower‑capacity ones.

Suitable scenario: give higher weight to machines with better configuration and lower load, and lower weight to weaker machines.

3. Random

The algorithm uses a random function to pick a server from the backend list, resulting in an even distribution of requests as traffic grows, similar to Round Robin.

4. Least Connections

This method tracks the number of active connections on each server and forwards new requests to the server with the fewest connections; it requires maintaining internal state and is less commonly recommended.

5. Source IP Hash

The client’s IP address is hashed, and the hash value modulo the number of servers determines the target server, ensuring that the same IP consistently maps to the same backend as long as the server list remains unchanged.

Suitable scenario: when request routing should be stable per client IP.

These five algorithms form the core toolbox for implementing effective load balancing in modern backend systems.