High-Concurrency LiteSpeed Web Server Tuning for WordPress in 2026

When a WordPress site bursts past 200k concurrent visitors, the default LiteSpeed settings become a bottleneck, causing queue buildup and skyrocketing TTFB. This guide walks you through the exact kernel, systemd, and LiteSpeed tweaks that let a 8‑core VPS sustain 10k RPS with sub‑50 ms latency, all while keeping memory footprints under control. Mera Blogger

Why Traditional LiteSpeed Defaults Fail at Scale

Out‑of‑the‑box LiteSpeed ships with conservative maxConnections (5000) and maxSSLConnections (2000). In a high‑traffic WordPress environment these limits trigger connection throttling, especially when LSCache is disabled for dynamic content. Additionally, the default PHP worker pool (maxChildren=35) cannot keep up with the PHP‑FPM demand generated by Gutenberg blocks, REST API calls, and WooCommerce cart updates.

Architectural Overview

We’ll layer three optimization domains:

• Kernel & System Tuning: sysctl, ulimit, and IRQ affinity.
• LiteSpeed Core Settings: connection limits, listener tuning, and LSCache policies.
• PHP‑FPM / LSAPI Workers: dynamic scaling, opcache, and per‑worker memory caps.

Benchmark Matrix

1. Kernel & System Tweaks

# /etc/sysctl.d/99-litespeed.conf
net.core.somaxconn = 65535
net.ipv4.tcp_syncookies = 1
net.ipv4.tcp_tw_reuse = 1
net.ipv4.tcp_fin_timeout = 15
net.ipv4.tcp_max_syn_backlog = 4096
net.ipv4.ip_local_port_range = 1024 65535
net.core.netdev_max_backlog = 5000
fs.file-max = 2000000
vm.swappiness = 1
vm.dirty_ratio = 15
vm.dirty_background_ratio = 5

Increase the open‑file limit for the LiteSpeed user:

# /etc/security/limits.d/litespeed.conf
litespeed   soft    nofile  200000
litespeed   hard    nofile  400000

2. LiteSpeed Core Configuration

All changes are made via the WebAdmin console (port 7080) or directly in /usr/local/lsws/conf/httpd_config.conf. After editing, run /usr/local/lsws/bin/lswsctrl restart.

listener Default {
    address                 *:80
    secure                  0
    maxConnections          25000
    maxSSLConnections       20000
    enableSSL               0
    map                     example.com $VH_ROOT
}
listener SSL {
    address                 *:443
    secure                  1
    maxSSLConnections       20000
    sslProtocol             TLSv1.3
    sslCipher               HIGH:!aNULL:!MD5
    map                     example.com $VH_ROOT
}
# Virtual Host tweaks
virtualhost example.com {
    docRoot                 $VH_ROOT/html
    enableGzip              1
    enableCache             1
    cachePolicy             LSCache
    cacheTTL                300
    phpIniOverride          /usr/local/lsws/conf/php.ini
}

LSCache Fine‑Tuning

Enable private caching for logged‑in users and set a 5‑minute TTL for dynamic fragments:

# /usr/local/lsws/conf/lscache.conf
CacheEnable public /
CacheEnable private /wp-admin/
CacheEnable private /wp-login.php
CacheTTL 300
CacheIgnoreHeaders Set-Cookie,Authorization

3. PHP‑FPM / LSAPI Worker Optimization

# /usr/local/lsws/conf/php.ini (partial)
opcache.enable=1
opcache.memory_consumption=256
opcache.max_accelerated_files=20000
opcache.validate_timestamps=0

# /usr/local/lsws/conf/lsphp.conf
lsapi:example {
    env                     LSAPI_MAX_CHILDREN=200
    env                     LSAPI_CHILDREN=200
    env                     LSAPI_MAX_REQ=1000
    env                     LSAPI_MAX_IDLE=30
    env                     LSAPI_MAX_SPARE=20
    env                     LSAPI_MIN_SPARE=10
    env                     LSAPI_MAX_SPARE=30
    env                     LSAPI_MAX_MEMORY=256M
}

4. Monitoring & Auto‑Scaling Hooks

Integrate lscpd metrics with Prometheus Node Exporter. Example alert for >80% worker queue depth:

# prometheus rule
- alert: LiteSpeedHighQueue
  expr: lscpd_worker_queue_length{instance="web01"} > 80
  for: 2m
  labels:
    severity: critical
  annotations:
    summary: "LiteSpeed worker queue high"
    description: "Queue length {{ $value }} exceeds threshold on {{ $labels.instance }}"

5. End‑to‑End Deployment Checklist

• Apply sysctl and limits files, then sysctl -p.
• Restart LiteSpeed and verify lswsctrl status shows 25k connections.
• Run ab -n 50000 -c 5000 https://example.com/ and confirm <90ms average latency.
• Enable Prometheus node exporter and set up Grafana dashboards for LSCache hit ratio.

FAQs