Mastering Load Balancing: Choose Between LVS, Nginx, and HAProxy

Most internet systems today rely on server‑cluster technology, where identical services are deployed on multiple machines to form a unified service pool. Before a web‑server cluster, a load‑balancing server acts as the entry point, directing client requests to the most appropriate web server.

Cloud computing and distributed architectures essentially package backend servers as a single service, abstracting the underlying machines from the client.

LVS (Linux Virtual Server)

LVS is a kernel‑level load‑balancing solution that has been part of Linux since the 2.4 kernel. It operates at Layer 4 (the transport layer) and forwards packets by modifying IP or MAC addresses.

LVS Architecture

The LVS cluster consists of three layers:

Front‑end Load Balancer layer.

Middle Server Array layer.

Back‑end Shared Storage layer.

Load‑Balancing Mechanisms

LVS works in two primary modes. In NAT (DNAT/SNAT) mode it rewrites destination and source IP addresses so that responses appear to come from the virtual IP. In DR (Direct Routing) mode it only changes the destination MAC address, leaving IP addresses untouched, which yields higher performance.

Advantages

Strong load‑handling capability; operates at the transport layer with minimal CPU and memory consumption.

Simple configuration reduces the chance of human error.

Stable operation with built‑in dual‑machine hot‑standby (LVS + Keepalived).

Zero traffic generation on the balancer itself, preserving I/O performance under heavy load.

Broad applicability: can balance HTTP, databases, chat services, and any TCP/UDP application.

Disadvantages

Does not support regular‑expression routing; cannot perform content‑based (Layer 7) separation.

Complex deployment for large sites compared with Nginx or HAProxy.

Nginx

Nginx is a powerful, event‑driven web server and reverse proxy designed for high‑concurrency HTTP handling and load balancing.

Architecture

Nginx uses a master process and multiple worker processes. Workers are single‑threaded, share memory, and rely on epoll for non‑blocking I/O, enabling efficient handling of tens of thousands of concurrent connections.

Load‑Balancing Features

Nginx balances HTTP/HTTPS traffic at Layer 7 using the upstream directive. Supported algorithms include:

Round‑robin (default).

Weighted round‑robin.

IP hash (session‑affinity).

Fair (third‑party, based on response time).

URL hash (third‑party, useful for cache servers).

Pros

Cross‑platform, easy to configure.

Handles 50 k+ concurrent connections with modest memory usage.

Event‑driven design with epoll for high scalability.

Built‑in health checks, gzip compression, and bandwidth saving.

Stable reverse‑proxy with low latency.

Cons

Limited to HTTP, HTTPS, TCP, and mail protocols.

Health checks are port‑based only; session persistence requires workarounds such as ip_hash.

HAProxy

HAProxy supports both Layer 4 (TCP) and Layer 7 (HTTP) proxying and can serve virtual hosts. It offers a rich set of scheduling algorithms such as round‑robin, weighted round‑robin, source, RI, and cookie‑based methods.

HAProxy complements Nginx by providing stronger session‑persistence features and more flexible backend health checks, often delivering higher throughput in pure load‑balancing scenarios.