How much RAM, CPU and disk does a VPS need: a sizing table

How many resources do typical workloads need?

The figures below are starting points for one project on a Linux server without a desktop. Real needs depend on traffic, code and data size, so start with the suggested configuration and watch the actual load for the first week or two.

How much RAM does a server need?

RAM is the main limit of a VPS: when it runs out, the server does not slow down gracefully, it starts killing processes. Ubuntu or Debian without a desktop takes a few hundred megabytes; the rest goes to your applications. Add up what each service uses and put 25-30% on top for peaks and the file system cache:

• nginx serving static files - tens of megabytes;
• a bot or API in Python / Node.js - usually 100-500 MB per process;
• PHP-FPM for WordPress - 50-100 MB per worker, typically 5-20 workers;
• MySQL or PostgreSQL - as much as you give them for cache (innodb_buffer_pool_size, shared_buffers);
• a Minecraft server - whatever -Xmx says, plus 1-2 GB for the JVM itself and the OS.

Swap (paging space on disk) saves you from a sudden crash during a short spike but does not replace memory: if swap is in use all the time, the server needs a plan with more RAM. How to add 2 GB of swap is in the first VPS setup guide.

How many cores (vCPU) do you need?

Most web workloads are fine on 1-2 vCPU: websites and bots spend more time waiting for the network and the database than computing. Cores become the bottleneck for parallel or long-running computation: code builds, video and image compression, game servers, language model inference. A vCPU is a host CPU thread assigned to your virtual machine; at Tihost the hosts run the AMD Ryzen 9 5950X and 9950X, with the high per-core clock that matters for single-threaded work such as Minecraft and PHP.

How much disk space does a VPS need?

A Linux system with updates takes 3-5 GB; everything else is your data, logs and images. Plan 30-50% headroom above your current data and do not run the disk above 80-85% full: databases and Docker stop when space runs out. Three things eat disk space quietly:

• Docker images and logs - cap the logs and prune old images as shown in the Docker article;
• systemd and application logs in /var/log;
• local backups - keep them on another server, see the backup guide.

Disk type matters as much as size: databases and builds run noticeably faster on NVMe than on SATA SSD - the difference is covered in NVMe vs SSD. Every Tihost plan uses NVMe.

How do you tell a VPS is short on resources?

Start with four commands: memory, core count, load average and free disk space.

free -h
nproc
uptime
df -h /

• free -h - look at the available column, not free: Linux uses idle memory for cache and hands it back on demand. If available is below 10-15% of total and the Swap line keeps growing, memory is short.
• uptime - three load average values for 1, 5 and 15 minutes. If they stay above the core count from nproc for long, the CPU is not keeping up.
• df -h / - how full the root partition is; above 85% it is time to clean up or grow.

To see which process uses the memory and CPU, open htop: sort by MEM% or CPU% with F6. The st (steal) column in top shows how long the vCPU waited for the host; values constantly above 5-10% are worth raising with your provider's support.

sudo apt install -y htop
htop

The most reliable sign of a memory shortage is the OOM killer: the Linux kernel forcibly terminates a process that ran out of memory and logs it. If the commands below find anything, the service needs more RAM or a cap on its usage:

sudo dmesg -T | grep -i -E "out of memory|oom-kill"
sudo journalctl -k --since "7 days ago" | grep -i "out of memory"

When should you move to a bigger plan?

• The OOM killer fired at least once in a week, or swap is used constantly rather than only during peaks.
• Load average stays above the vCPU count at peak hours, and pages or bot replies visibly slow down.
• The disk is more than 80-85% full even after clearing logs and old images.
• A second project with its own database has moved onto the server.

If memory runs out while the CPU is idle, pick a configuration with more RAM and the same cores: for example, from 4 vCPU / 4 GB to 4 vCPU / 8 GB. If you hit the cores, go from 6 vCPU / 12 GB to 8 vCPU / 12 GB. Every line is covered in the Starter, Advanced and Power articles.

Tihost VPS configurations

The most affordable configuration is 1 vCPU / 2 GB at $4.00 per month; for LLMs and large databases there are 12 vCPU / 24 GB at $49.00 and 16 vCPU / 32 GB at $77.50 in Poland. A server is deployed in about 2 minutes, so starting small and ordering the next size once the load proves it is a sound strategy. The full comparison is in the plans overview.