Understanding Load Balancing Algorithms and Hardware Solutions

This is the first article of a detailed series on load balancing, introducing load balancing algorithms and hardware load balancing. Some content is excerpted from reading notes.

Load Balancing Algorithms

Common load balancing algorithms include round‑robin, random, least connections, source address hash, and weighted methods.

Round‑Robin

Distributes all requests sequentially to each server, suitable when server hardware is identical.

Advantages: each server receives the same number of requests.

Disadvantages: server load may differ; not suitable for servers with different configurations.

Random

Requests are assigned randomly to servers.

Advantages: simple to use.

Disadvantages: not suitable for environments with heterogeneous server configurations.

Least Connections

Assigns requests to the server with the fewest active connections (the server currently handling the least load).

Advantages: dynamically distributes based on current server load.

Disadvantages: implementation is relatively complex and requires monitoring of connection counts.

Hash (Source Address Hash)

Computes a hash from the IP address to determine the target server.

Advantages: requests from the same IP during a session are forwarded to the same server, enabling session stickiness.

Disadvantages: if the target server fails, the session is lost.

Weighted

Builds on round‑robin, random, least connections, and hash algorithms by adding weights to influence server selection.

Advantages: adjusts request distribution according to server weight.

Disadvantages: relatively complex to use.

Hardware Load Balancing

Hardware load balancing is implemented using dedicated appliances, which are expensive and typically considered by large enterprises; leading products include F5 and A10.

When using hardware load balancers, consider the following aspects:

Functionality: comprehensive support for load balancing at all layers, full algorithm support, and global load balancing.

Performance: software load balancers struggle beyond 50,000 concurrent connections, while hardware can handle higher loads.

Stability: commercial hardware devices have undergone rigorous testing and large‑scale deployment, offering high stability.

Security: hardware devices also provide firewall and DDoS protection features.

Maintenance: provide good management interfaces, after‑sales service, and technical support.

Cost for large enterprises: F5 Big‑IP price ranges from 150k to 550k; A10 price ranges from 550k to 1M.

Drawbacks:

High price.

Poor scalability.

Summary

Generally, hardware load balancers require dual‑machine high availability, increasing cost.

Internet companies usually use open‑source software load balancers, so most applications adopt software load balancing; some also use hardware solutions.

For example, a certain internet company currently uses several F5 devices for global load balancing and Nginx for internal software load balancing.

Summary of This Share

The above introduced load balancing principles, classifications, algorithms, and hardware load balancing. In the next session, load balancing details (part three) will cover software load balancers (LVS, Nginx, Haproxy, Apache) characteristics, architecture, technology selection comparison, application load balancing issues, and solutions.