Choosing the right VPS tier: CPU, RAM and disk based on real traffic
How to map daily visits, php-fpm processes and database size to a concrete VPS tier, with commands to verify actual usage before changing plans.
UptimeMag editorial team · 22 November 2026 · 3 min read

Anyone running a website often faces the same question: is my VPS too small, too large, or just right? The answer doesn't lie in "daily visits" read from Google Analytics, but in three measurable numbers on the server: how many php-fpm processes are running concurrently, how much RAM each one takes up, and how much the database weighs. This guide connects those numbers to a concrete CPU/RAM/disk tier.
The number that really matters: RAM per php-fpm process
The parameter that determines the minimum RAM tier is pm.max_children, which specifies how many PHP processes can run together. The reference formula, cited across multiple independent technical guides, is: "max_children = (total RAM - reserved) / per-worker RAM". The PHP budget isn't the total RAM: you must subtract what's used by the operating system, web server, OPcache and local database.
To measure how much RAM each php-fpm process actually consumes on your existing server, before deciding whether to upgrade, run:
ps --no-headers -o "rss,cmd" -C php-fpm | awk '{ sum+=$1 } END { printf ("%dMB", sum/1024) }'
This command returns how much RAM each PHP-FPM process uses and the total used by all processes in MB. Dividing the available RAM by the average value per process gives you the maximum number of children the VPS can handle without going into swap.
A safety margin is essential: multiple sources recommend keeping a margin and allocating up to 80% of total RAM to PHP-FPM, so on 2 GB of RAM max_children should be around 32 instead of the full theoretical number.
Check actual usage before upgrading
Before upgrading, verify whether your current VPS is genuinely under pressure using two commands that server administrators run routinely:
htop— displays CPU per core in real time, RAM used and swap. If swap stays above zero continuously, RAM is insufficient.vmstat 1— showssi/socolumns (swap in/out) andwa(I/O wait). Ifwaremains high, the bottleneck is disk, not CPU.
Another reliable signal comes from the php-fpm status panel: by enabling pm.status_path you can check the request queue. If the "listen queue" stays above zero continuously, the server is already at the limit of current capacity and the tier needs to be raised, not just reconfigured.
Three practical profiles
| Profile | Concurrent php-fpm processes active | Recommended minimum RAM | Notes |
|---|---|---|---|
| Blog with a few thousand visits/day | 5–15 | 2–4 GB | Load is almost entirely read-heavy; page caching drastically reduces active PHP processes |
| Small WooCommerce shop | 15–30 | 4–8 GB | Shopping cart and checkout generate heavier, longer-running PHP processes than simple static articles |
| Multi-user application or admin dashboard | 30–60+ | 8–16 GB | Multiple sessions, more frequent database queries, less room for caching |
These ranges come from applying the RAM-per-process formula to typical values reported in technical tuning guides (PHP process consuming 50–120 MB depending on CMS and active plugins), not from memory: every site should still be measured with ps and htop before deciding.
Cost of the tier you choose
For those starting from scratch or needing to upgrade the VPS, a useful pricing reference is Hetzner Cloud, which publishes prices on its official page (hetzner.com/cloud). Cost-Optimized plans start at €5.99 per month, Regular Performance plans with shared resources at €5.49 per month, while General Purpose plans with dedicated vCPU — described by the provider for high-traffic applications with consistently high CPU — start at €36.99 per month, prices excluding VAT as of 22 November 2026. The provider describes shared resource plans as suitable for personal websites and blogs, small databases, or web servers with low to moderate traffic, whilst dedicated plans are recommended for business applications with high production loads, such as high-traffic sites and large databases. For a blog, the entry-level Cost-Optimized plan is sufficient; for a multi-user application it's worth starting with a dedicated vCPU plan, because shared CPU slows down precisely during spikes in concurrent queries.
In summary, the right tier isn't chosen by looking at visits but by measuring active processes, RAM per process and disk I/O on the server you already have: only after those measurements does it make sense to look at the price list.
Written with the help of artificial intelligence and checked by the editors (EU AI Act, art. 50).