Tutorial on high-performance load balancing in Windows and Linux environments

1. NGINX Architecture & Working Principle

/Worker process model

• Master: Only responsible for loading configuration and managing Workers (smooth reloading, signal processing).
• Worker: Event-driven, generally the same number of CPU cores or slightly more, throughepoll/kqueueEfficiently handle massive concurrent connections.

2. Non-blocking I/O & asynchronous events

• Each Worker can manage tens of thousands of connections within one thread, greatly reducing context switching.

3. Modular design

Subsystems such as HTTP, Stream, Mail, etc. can compile modules on demand:

• ngx_http_upstream_module(Load balancing)
• ngx_http_stub_status_module(Status Monitoring)
• Third-party health check module, Lua script extension, etc.

2. Windows platform deployment

Applicable scenarios: Internal testing of the enterprise, collaboration with IIS (passedARR / AJPProtocols, etc.) may facilitate rapid deployment in Windows environment.

2.1 Download and decompression

1. access(en) Download page
2. GetStable versionZIP packages, e.g.nginx-1.24.。
3. Unzip toC:\nginx\Table of contents.

2.2 Register as a Windows service (optional)

useNSSM（Non-Sucking Service Manager）：

# Assuming it has been placed in C:\tools\nssm\C:\tools\nssm\nssm install nginx "C:\nginx\"
# Start the servicenet start nginx

2.3 Configuration file location

• Main configuration:C:\nginx\conf\
• log:C:\nginx\logs\ /

2.4 Common Commands

# start upC:\nginx\

# Smooth reload (reread configuration)C:\nginx\ -s reload

# stopC:\nginx\ -s quit

Tips: Under Windows,-s reloadPerhaps not as stable as Linux, it can achieve zero Downtime by combining restarting of NSSM services.

3. Linux platform deployment

Applicable scenarios: The first choice for production environment. This article takes Ubuntu as an example, and CentOS/RHEL is the same.

3.1 Installation (package management/compilation and installation)

• 3.1.1 Package Management Installation

# Ubuntu / Debian
sudo apt update
sudo apt install -y nginx

# CentOS / RHEL
sudo yum install -y epel-release
sudo yum install -y nginx

• 3.1.2 Source code compilation (custom module)

# Installation dependenciessudo apt install -y build-essential libpcre3 libpcre3-dev zlib1g zlib1g-dev \
                     libssl-dev

# Download the source code and compile itwget /download/nginx-1.24.
tar zxvf nginx-1.24. && cd nginx-1.24.0

./configure \
  --prefix=/usr/local/nginx \
  --with-http_ssl_module \
  --with-http_v2_module \
  --with-http_stub_status_module \
  --with-stream \
  --with-stream_ssl_module
make && sudo make install

3.2 Service Management

Systemd

sudo systemctl enable nginx
sudo systemctl start nginx
sudo systemctl reload nginx
sudo systemctl status nginx

Configuration Directory

• /etc/nginx/: Main configuration
• /etc/nginx//*.conf: Virtual Host/Load Balancing Snippet
• /usr/share/nginx/html/: Default static service root

4. Core load balancing configuration

The following examples are all placed inhttp { … }In the section.

4.1 Basics Round-Robin (polling)

upstream backend {
    server :80;
    server :80;
    server :80;
}

server {
    listen 80;
    location / {
        proxy_pass http://backend;
    }
}

• characteristic: Default polling, new connections are distributed in sequence; peer servers do not require additional instructions.

4.2 Least-Connected (minimum connection)

upstream backend {
    least_conn;
    server ;
    server ;
    server ;
}

• Scene: The request time is large, or the performance of some nodes is different, and dynamic balance is more fair.

4.3 IP-Hash (based on client IP)

upstream backend {
    ip_hash;
    server ;
    server ;
    server ;
}

• Session keeping: The same client IP always hits the same backend, suitable for stateful applications (Session, shopping cart, etc.).

4.4 Weight

upstream backend {
    server  weight=4;
    server  weight=1;
    server  weight=1;
}

• significance: Of every 6 requests, srv1 will bear 4 and 1 other. Can be used in conjunction with any scheduling algorithm (polling, least_conn, ip_hash).

4.5 Alternate node & forced offline

upstream backend {
    server ;
    server  backup;   # Use only if the main group is unavailable    server  down;     # Permanent offline (manual maintenance)}

5. Health checks and failure recovery

5.1 Passive health check (out of the box)

• max_fails: Number of consecutive failures
• fail_timeout: How long does it take to disable after failure

upstream backend {
    server  max_fails=2 fail_timeout=15s;
    server ;
}

process：

• If 2 requests timeout/disconnect, srv1 is marked as Unhealthy;
• Disable 15s;
• After 15 seconds, the first new request is activated and it will be restored if successful.

5.2 Active health check (NGINX Plus/third-party module)

• NGINX Plus:built-inhealth_checkwith visualization API.
• Open Source: Compilablenginx-upstream-check-module, or use Lua script to activate the intervalhttpDetection.

6. Performance Tuning

6.1 Connection and Buffer

# Set up at http{} top levelworker_connections  10240;      # Maximum number of connections per Workerkeepalive_timeout   65s;        # Client long connection timeoutkeepalive_requests  100;        # Maximum number of requests for single connections

• proxy_buffers: Adjust the upstream response buffer
• proxy_busy_buffers_size: Large response scenario optimization

6.2 Timeout configuration

server {
    proxy_connect_timeout  3s;   # Establishing TCP connection timeout    proxy_send_timeout     10s;  # Send request to the backend timeout    proxy_read_timeout     30s;  # Receive backend response timeout}

6.3 Kernel & Network Tuning

• somaxconn

sysctl -w =65535

• tcp_tw_reuse/tcp_fin_timeout: Accelerate TIME_WAIT recycling
• ulimit -n: Increase file descriptor upper limit

7. Monitoring & Logging

7.1 Stub Status

existserver{}Section opens:

location /nginx_status {
    stub_status on;
    allow  127.0.0.1;
    deny   all;
}

Output example:

Active connections: 291 
server accepts handled requests
 15394 15394 54123 
Reading: 0 Writing: 1 Waiting: 290

7.2 Log format customization

log_format main '$remote_addr - $remote_user [$time_local] '
                '"$request" $status $body_bytes_sent '
                '"$http_referer" "$http_user_agent" '
                'rt=$request_time ua="$upstream_addr" us="$upstream_response_time"';
access_log /var/log/nginx/ main;

• rt/usCan be used for requesting time-consuming analysis and back-end performance monitoring.

7.3 Prometheus Integration

• usenginx-prometheus-exporter；
• Combined with Grafana to build a real-time large screen.

8. Common troubleshooting

9. Advanced & Expand

• Dynamic DNS resolution：

server  resolve; + resolver

• gRPC & HTTP/2：

http2 on;
grpc_pass grpc://backend;

• cache：proxy_cache_path + proxy_cache
• Safety: IP whitelist, WAF (ModSecurity), TLS uninstallation and hardware acceleration

Summarize

This article systematically covers:

1. NGINX architecture and event model
2. Windows and Linux dual-platform installation and service
3. Core scheduling algorithm: polling, minimum connection, IP hashing, weighting, standby/downline
4. Health Checkup (Passive & Active)
5. Performance and kernel tuning
6. Monitoring, log collection and visualization
7. Common troubleshooting and advanced functions

Through the above practices, you can quickly build a highly available, scalable HTTP load balancing layer in your own development environment or production cluster. In the future, it can combine Service Mesh, WAF, security audit and grayscale release to achieve more advanced traffic control and operation and maintenance automation.

The above is personal experience. I hope you can give you a reference and I hope you can support me more.