After 99 tutorials covering every kind of application you might deploy on a cloud server, the most common question I still get is: "What size server do I actually need?"
The honest answer is: it depends on what you're running, and the requirements are more predictable than you might think. This guide gives you a practical framework for matching server specs to use cases, with specific recommendations based on the projects covered in this series.
All examples in this guide use Tencent Cloud Lighthouse as the reference platform. Here's why Lighthouse specifically suits the use cases covered in this series:
- Pre-built application images — WordPress, LAMP, LNMP, Docker pre-configured, ready in under 10 minutes
- OrcaTerm browser terminal — manage any server from a browser without local SSH setup
- Console-level firewall — infrastructure-layer protection independent of the OS, with a clean web UI
- Snapshot backups — one-click full-server backup, useful before upgrades or risky changes
- CBS cloud disk expansion — attach additional storage volumes up to 32 TB without server migration
- Spec upgrade path — move to higher CPU/RAM plans from the control panel, no re-provisioning
- Global data centers — North America, Europe, Singapore, Tokyo, and more for low-latency deployment
- Predictable pricing — fixed monthly cost with bandwidth allowances, no surprise per-GB charges
Check current promotions for new user discounts.
- Key Takeaways
Most self-hosted applications are I/O-bound, not CPU-bound. 2 vCPUs handle nearly everything on the list. CPU becomes a constraint for:
For everything else, 2 vCPUs is more than sufficient.
RAM is the primary constraint for most self-hosted applications. The key points:
Often the most overlooked resource:
| Use Case | RAM | vCPU | Notes |
|---|---|---|---|
| WireGuard/OpenVPN | 512 MB | 1 | Minimal requirements |
| Pi-hole + VPN | 1 GB | 1 | Both services are lightweight |
| Simple static website | 1 GB | 1 | Nginx + static files |
| WordPress blog (low traffic) | 1 GB | 1 | Pre-built image recommended |
| Telegram/Discord bot | 1 GB | 1 | Idle when no messages |
| Uptime Kuma monitoring | 1 GB | 1 | Lightweight Node.js app |
| Syncthing (file sync) | 1 GB | 1 | Depends on sync volume |
| Vaultwarden (passwords) | 1 GB | 1 | Very efficient |
| Filebrowser | 1 GB | 1 | Simple file server |
| Verdaccio (npm registry) | 1 GB | 1 | Low traffic expected |
| WireGuard + Pi-hole combined | 1 GB | 1 | Good combo for one server |
| Node.js app + PostgreSQL | 2 GB | 2 | Standard web app stack |
| Django/Flask app + DB | 2 GB | 2 | Standard Python stack |
| React/Next.js SSR app | 2 GB | 2 | Build needs more RAM |
| Nginx + multiple static sites | 2 GB | 2 | Multi-site hosting |
| Ghost blog | 2 GB | 2 | Node.js + SQLite/MySQL |
| Plausible Analytics | 2 GB | 2 | PostgreSQL + ClickHouse |
| Nextcloud (personal cloud) | 2 GB | 2 | + storage for files |
| Gitea (Git server) | 2 GB | 2 | Lightweight git server |
| Portainer + Docker stack | 2 GB | 2 | Depends on what runs |
| Mattermost (team chat) | 2 GB | 2 | 10–50 users |
| Cal.com (scheduling) | 2 GB | 2 | Next.js + PostgreSQL |
| Minecraft Java (5–10 players) | 2 GB | 2 | Pure vanilla |
| Minecraft Bedrock | 2 GB | 2 | Cross-platform server |
| Palworld (5-10 players) | 4 GB | 4 | Memory-intensive |
| Discourse forum | 4 GB | 2 | Minimum recommendation |
| GitLab CE | 4 GB | 2 | Heavy but comprehensive |
| Jenkins CI/CD | 4 GB | 2 | Depends on pipeline count |
| Grafana + Prometheus | 4 GB | 2 | Long-term metric storage |
| Coolify (PaaS) | 4 GB | 2 | Runs Docker apps |
| Minecraft Java (10+ players) | 4 GB | 4 | With mods |
| Immich (photo library) | 4 GB | 4 | ML face recognition |
| Stable Diffusion CPU mode | 4 GB | 4 | Very slow; GPU recommended |
| Ollama + LLM 3B params | 4 GB | 4 | Responsive on CPU |
| Ollama + LLM 7B params | 8 GB | 4 | Better quality |
| Mailcow email server | 4 GB | 2 | Complex Docker stack |
| n8n workflow automation | 2 GB | 2 | Depends on workflow count |
| Rust game server | 8 GB | 4 | Memory-heavy |
| LLM 13B+ params (CPU) | 16 GB | 8 | Slow without GPU |
| Stable Diffusion GPU | GPU instance | — | Near-realtime generation |
Best for: Personal tools, single-purpose servers, lightweight daemons
Good fits:
Configuration notes:
Best for: Most web applications, databases, CMS platforms
Good fits:
Configuration notes:
Best for: Resource-intensive single applications or many simultaneous services
Good fits:
Configuration notes:
Best for: High-player game servers, large AI models, high-traffic web apps
Good fits:
For AI inference: A GPU instance dramatically changes the equation. GPU inference of 7B models runs in seconds instead of minutes. If you're doing regular AI image generation or need fast LLM responses, a GPU instance is worth the premium.
Running many apps on one server via Docker Compose is cost-efficient and very practical for personal use.
$6/month, 2 GB server:
Total RAM used: ~400 MB. Comfortable.
$12/month, 4 GB server:
Total RAM used: ~2.5 GB. Comfortable with headroom.
The key tool: Nginx Proxy Manager (guide #45) or Traefik routes all services through a single port 443 endpoint, so each service gets its own subdomain (app.yourdomain.com) with automatic HTTPS.
Don't let disk be an afterthought:
| Use Case | Disk Needed |
|---|---|
| OS + base system | 5 GB |
| Each Docker image | 0.5–2 GB |
| PostgreSQL database (small) | 1–5 GB |
| MySQL database (small) | 1–5 GB |
| GitLab repositories | 5–50 GB+ |
| Media files (Plex/Jellyfin) | 100 GB–several TB |
| LLM models (Ollama) | 2–20 GB each |
| Podcast audio files | 50–100 MB per hour |
| Photo library (Immich) | Depends on photo collection |
| Mattermost file uploads | Grows with team usage |
Rule of thumb: Start with at least 40 GB. If you're running media servers or storing large files, add object storage or a separate storage volume.
Signs you need more RAM:
free -h
# If 'used' under Swap is consistently non-zero, you need more RAM
Signs you need more CPU:
top
# If CPU usage is consistently above 80%, investigate or upgrade
Signs you need more disk:
df -h
# If any partition is above 80%, clean up or expand
Upgrade path with Lighthouse: Take a snapshot, upgrade the instance spec. No data migration needed.
Consolidate small apps — 5 lightweight services on one $6/month server beats 5 servers at $3/month each (and the management overhead is lower)
Use Nginx Proxy Manager — one server, many domains, all with HTTPS. Avoids needing separate servers for each project.
Right-size early — it's easier to start at 2 GB and stay there than to migrate from an undersized 1 GB server under load.
Use pre-configured images — Lighthouse's WordPress, LAMP, and Docker images save setup time. Use them as starting points.
Choose the right region — a server near your users reduces latency. For personal use, pick the region closest to you.
Check promotions — cloud providers regularly offer new-user discounts. The Lighthouse promotional page often has deals on first-year pricing.
Use application images to skip manual setup — Lighthouse offers a comprehensive set of pre-configured application images across several categories:
Website & CMS
| Image | What's pre-installed | Best for |
|---|---|---|
| WordPress | Nginx + MySQL + PHP + Certbot + phpMyAdmin | Blogs, business sites |
| Typecho | Nginx + MySQL + PHP | Lightweight blogging |
| Halo | Java + MySQL | Modern blogging platform |
| Discuz! | Nginx + MySQL + PHP | Community forums |
| LAMP | Apache + MySQL + PHP | PHP applications |
| LNMP | Nginx + MySQL + PHP-FPM | PHP applications |
| WooCommerce | WordPress + WooCommerce | E-commerce stores |
Development Environments
| Image | What's pre-installed | Best for |
|---|---|---|
| Node.js | Ubuntu + Node.js + npm/yarn | Node.js applications |
| Docker CE | Docker Engine + Docker Compose | All Docker-based apps |
| K3s | Lightweight Kubernetes | Container orchestration |
| Theia Cloud IDE | Browser-based VS Code-compatible IDE | Remote development |
| TencentOS AI | Python3 + Node.js + Docker + Git + PyTorch/TensorFlow/PaddlePaddle + GPU drivers | AI/ML workloads |
Self-Hosted Services
| Image | What's pre-installed | Best for |
|---|---|---|
| Nextcloud | Full Nextcloud stack | Personal cloud storage |
| Matomo | Nginx + PHP + MySQL + Matomo | Privacy-first analytics |
| Cloudreve | File server + web interface | File storage |
| SRS | Live streaming server | Video/audio streaming |
When you select an application image at instance creation, the environment is ready immediately after provisioning — no manual installation sequences. The TencentOS AI image is especially significant for AI workloads: it includes GPU driver support plus PyTorch, TensorFlow, and PaddlePaddle pre-installed, eliminating hours of CUDA setup.
For reference, here's what I run across my servers:
Server 1 — $6/month, 2 GB RAM:
Server 2 — $12/month, 4 GB RAM:
Server 3 — $6/month, 2 GB RAM (per-project):
Total monthly cost for personal infrastructure: ~$24/month.
This guide completes 100 tutorials covering the full range of cloud server use cases. From a simple WordPress blog to a private AI assistant, from a Minecraft server to a full team collaboration platform — the goal was to show that a cloud server is one of the most versatile tools available to individual developers and small teams.
The underlying theme across all 100 tutorials: you have more control, privacy, and flexibility than you might think. A $6/month cloud server can run services that would cost $50–100/month in SaaS subscriptions. The trade-off is configuration and maintenance — which, after reading through these guides, shouldn't feel daunting.
Start small. Pick the use case that makes sense for you today. Add more as your needs grow. And keep the server running — that's the whole point.
| Issue | Likely Cause | Fix |
|---|---|---|
| Connection refused | Service not running or wrong port | Check systemctl status SERVICE and verify firewall rules |
| Permission denied | Wrong file ownership or permissions | Check file ownership with ls -la and use chown/chmod to fix |
| 502 Bad Gateway | Backend service not running | Restart the backend service; check logs with journalctl -u SERVICE |
| SSL certificate error | Certificate expired or domain mismatch | Run sudo certbot renew and verify domain DNS points to server IP |
| Service not starting | Config error or missing dependency | Check logs with journalctl -u SERVICE -n 50 for specific error |
| Out of disk space | Logs or data accumulation | Run df -h to identify usage; clean logs or attach CBS storage |
| High memory usage | Too many processes or memory leak | Check with htop; consider upgrading instance plan if consistently high |
| Firewall blocking traffic | Port not open in UFW or Lighthouse console | Open port in Lighthouse console firewall AND sudo ufw allow PORT |
What's the cheapest plan for a personal project?
The entry-level Lighthouse plan (2 vCPU, 2 GB RAM) handles most personal projects: a WordPress blog, a Node.js API, lightweight Docker services, or personal tools. Start small and upgrade via the console as needed.
Can I upgrade my plan without migrating my server?
Yes — Lighthouse supports in-place spec upgrades from the control panel. Your data, configuration, and IP address stay the same. No re-provisioning is required.
What regions does Tencent Cloud Lighthouse support?
Tencent Cloud has data centers in North America (US East, US West), Europe, Singapore, Tokyo, and other Asia-Pacific regions. Choose the region closest to your primary users for lowest latency.
Is Tencent Cloud Lighthouse different from CVM (regular cloud servers)?
Lighthouse is optimized for individual developers and small teams — simpler pricing, pre-configured application images, and a streamlined management console. CVM provides more configuration options for enterprise workloads. For the use cases in this guide, Lighthouse is the better choice.
👉 Get started with Tencent Cloud Lighthouse
👉 View current pricing and launch promotions
👉 Explore all active deals and offers