Website sluggish under 500 RPS load? Moving to a dedicated server is a logical step, but incorrect configuration negates the advantages. We often see servers where expensive hardware sits idle due to suboptimal Nginx and PostgreSQL settings. Our team, with ten years of experience, configures dedicated servers end-to-end: from hardware RAID to monitoring. We'll assess your project for free and choose a configuration that can handle 10,000 RPS. Get an engineer's consultation — contact us.
Why Proper Server Configuration is Critical
Even a powerful server with 128 GB RAM and NVMe drives can show poor performance if system parameters are not tuned. A typical mistake is default file descriptor limits (1024) under thousands of connections. Or disabled opcache for PHP applications. We fix such issues during setup.
Recommended Configurations for Different Loads
| Load | CPU | RAM | Disks |
|---|---|---|---|
| Up to 1000 RPS | 4 vCPU | 16 GB | 2× 240 GB SSD |
| 1000–10000 RPS | 8 vCPU | 64 GB | 2× 480 GB NVMe |
| >10000 RPS | 2× Xeon (16 cores) | 128 GB | 2× 960 GB NVMe |
For projects with PostgreSQL, we recommend at least 64 GB RAM.
How to Choose a RAID Level?
RAID 1 provides mirroring — if a disk fails, the system continues working. RAID 10 (or RAID 0 for data) gives maximum read/write speed. For the root partition we use RAID 1, for data — RAID 10 if more than two disks are available. Configuration is done via mdadm:
# Check disks lsblk fdisk -l # Create RAID 1 for root partition mdadm --create /dev/md0 --level=1 --raid-devices=2 /dev/sda /dev/sdb mkfs.ext4 /dev/md0 # RAID 10 for data (4 disks) mdadm --create /dev/md1 --level=10 --raid-devices=4 \ /dev/sdc /dev/sdd /dev/sde /dev/sdf More about RAID can be read on Wikipedia.
Comparison of RAID Levels for a Web Server
| Level | Fault Tolerance | Write Speed | Disk Utilization |
|---|---|---|---|
| RAID 0 | None | ✦✦✦✦✦ | 100% |
| RAID 1 | 1 disk | ✦✦✦ | 50% |
| RAID 5 | 1 disk | ✦✦✦ | 67–94% |
| RAID 10 | up to 50% of disks | ✦✦✦✦✦ | 50% |
RAID 10 is 3 times faster in write speed than RAID 5, so we choose it for data.
Example configuration for 10,000 RPS
For a project with 10,000 RPS load we recommend:
- CPU: 2× Intel Xeon E5-2670 (16 cores / 32 threads)
- RAM: 128 GB DDR4 ECC
- SSD: 2× NVMe 960 GB (RAID 1 for OS, RAID 0 for data)
- Network: 1 Gbps Unmetered
System and Stack Optimization
System Parameters
We optimize system parameters: increase file descriptor limits, tune network stack, disable swap. Typical settings:
# /etc/sysctl.conf and /etc/security/limits.conf net.core.somaxconn = 65536 net.core.netdev_max_backlog = 65536 net.ipv4.tcp_max_syn_backlog = 65536 net.ipv4.tcp_fin_timeout = 15 net.ipv4.tcp_tw_reuse = 1 net.ipv4.ip_local_port_range = 10240 65535 fs.file-max = 2097152 vm.swappiness = 10 * soft nofile 1048576 * hard nofile 1048576 root soft nofile 1048576 Nginx
# /etc/nginx/nginx.conf worker_processes 8; events { worker_connections 4096; use epoll; multi_accept on; } http { keepalive_timeout 65; keepalive_requests 1000; sendfile on; tcp_nopush on; tcp_nodelay on; include /etc/nginx/mime.types; default_type application/octet-stream; } PHP-FPM and OPcache
; /etc/php/8.3/fpm/pool.d/www.conf pm = dynamic pm.max_children = 50 pm.start_servers = 10 pm.min_spare_servers = 5 pm.max_spare_servers = 20 pm.max_requests = 500 [opcache] opcache.enable = 1 opcache.memory_consumption = 512 opcache.max_accelerated_files = 20000 opcache.validate_timestamps = 0 PostgreSQL
PostgreSQL tuning is critical: shared_buffers = 25% RAM, effective_cache_size = 75% RAM, random_page_cost = 1.1 for SSD. Example for 128 GB RAM:
shared_buffers = 32GB effective_cache_size = 96GB maintenance_work_mem = 2GB work_mem = 128MB wal_buffers = 64MB checkpoint_completion_target = 0.9 max_wal_size = 4GB random_page_cost = 1.1 effective_io_concurrency = 200 Server Setup Process
- Audit of current configuration — if the server is already in use, we analyze metrics and logs.
- OS installation (Ubuntu LTS or Debian stable) with RAID and ext4 filesystem.
- Web stack installation and optimization — Nginx, PHP-FPM, PostgreSQL.
- Security setup — firewall (ufw), fail2ban, SSL certificates (Let's Encrypt).
- Monitoring integration — Prometheus + Grafana, Zabbix for alerts, Sentry for errors.
- Backup and deployment automation (Ansible / GitLab CI).
- Handover of documentation and training for your team.
Timeline and Cost
Setting up a dedicated server from scratch (OS, RAID, stack, SSL, monitoring) takes 3–5 days. Optimization for a specific load takes another 2–3 days. Cost is calculated individually based on complexity and required stack. Incorrect configuration can cost up to 40% performance. Proper configuration pays for itself in 2–3 months by reducing latency and increasing throughput.
Why Choose Us
Over 10 years of experience in server configuration. Certified Linux and PostgreSQL engineers. 30-day warranty on completed work. Completed over 50 projects in migration and infrastructure optimization. Typical mistakes we fix: swap not configured (swap on SSD kills cell life), default limits (leading to 'Too many open files' errors), wrong RAID level (RAID 5 on SSD gives poor write speed), lack of monitoring (problems discovered only on failure). Order an audit of your server — we will identify bottlenecks and suggest an optimal configuration. Contact us for a free consultation.