How Many Resources Does a VPS Need: How to Measure Load and Avoid Overpaying

Choosing a plan usually happens blindly: people pick something with headroom and spend years paying for idle cores, or they cut costs and then deal with crashes during peak hours. Yet the right configuration can actually be calculated: load is measured with a handful of commands, and the result translates into concrete numbers for RAM and CPU cores.

Let us go through it step by step: where to start, what tools to use, and what signs tell you it is time to upgrade.

Where to Start If the Project Does Not Exist Yet

If the site has not launched yet, there is no need to guess — start with the entry-level plan. You can upgrade at any time without reinstalling anything, and there is no reason to overpay from day one.

Practical starting points: a landing page or blog needs 1 vCPU and 2 GB of RAM; a corporate site or small store needs 2 vCPU and 4 GB; a store with hundreds of products, 1C integration, and background jobs needs 4 vCPU and 8 GB. A Telegram bot, VPN, or scraper runs fine on the smallest plan.

In the NodexGo panel this looks like: click Create Server, choose a location — your real ping to each one is shown right there — then pick a plan and OS image, set a server name, and choose a billing period of 1, 3, 6, or 12 months. Within a minute the IP address and root password will appear in the panel and arrive by email.

What to Use for Measurement: Five Commands

Connect to the server and install a set of utilities — a fresh system usually does not have them:

ssh root@server-IP-address
apt update && apt install -y htop sysstat

RAM is the primary resource that causes servers to crash. Check how much is in use and how much has spilled into swap:

free -h

If available is consistently below 15 percent of total RAM and swap keeps growing, memory is already insufficient. This is not a future problem — it is happening right now.

CPU: what matters is not the instantaneous load but the average over 1, 5, and 15 minutes:

uptime

The three numbers at the end are the load average. The rule is simple: the value should not consistently exceed the number of cores. For 2 vCPU, a load average of 2.0 is the limit; 4.0 means the server is choking.

A live view of processes — to see exactly what is consuming resources:

htop

Disk: both space and speed matter. A full disk will bring down a database just as reliably as insufficient RAM:

df -h
iostat -x 5 3

In the iostat output, watch the %util column: values near 100 percent mean the disk has become a bottleneck.

Testing Whether the Server Can Handle a Traffic Spike

A synthetic test reveals the ceiling before real visitors hit it. Install the utility and run 500 requests across 20 parallel threads:

apt install -y apache2-utils
ab -n 500 -c 20 https://example.com/

Look at Requests per second and Failed requests. If errors appear or response time multiplies, that is your real ceiling. Only test your own sites — running this against someone else's server will look like an attack.

How to Know It Is Time to Upgrade

RAM runs out and the system starts killing processes. Check directly:

dmesg -T | grep -i 'out of memory' | tail -5

Lines containing Out of memory mean the kernel has already been killing your processes — the database, PHP, or a container. This is the clearest signal there is.

Load average is consistently above the number of cores during working hours; site response time has increased with no code changes; the disk is more than 85 percent full; database logs show errors about insufficient connections or memory.

Note: a slow site does not always mean an underpowered server. First check caching, heavy SQL queries, and uncompressed images — fixing those often delivers a bigger improvement than doubling the plan.

What to Do When You Have Hit the Limit

Inside the server card in the panel there is an Increase Resources button: the plan upgrades to the next tier, the disk and IP address are preserved, and nothing needs to be migrated or reinstalled. You pay only the difference for the remaining billing period.

The opposite situation also occurs: a server running at 5 percent load for months. In that case it makes sense to review the configuration and stop paying for unused capacity.

What Else Affects the Choice Beyond the Numbers

Location. Ping to your audience affects speed more than an extra core. In the server creation wizard, the ping to each location is measured directly from your browser — use that as your guide, not the country name.

Disk type. NVMe versus a standard SSD makes a several-fold difference on database operations, and most websites and admin panels are bottlenecked by the database.

Traffic. For a VPN, file storage, or media server, the traffic allowance in the plan matters more than cores and RAM.

Headroom for peaks. If you run sales or send mass mailings, plan for peak load rather than average load — and keep at least one and a half times the memory you normally need.

In Brief

Start with the entry-level plan, install htop and sysstat, and check free -h, uptime, and df -h once a week. Upgrade when RAM spills into swap, load average is consistently above the number of cores, or the disk is 85 percent full. The Increase Resources button handles it in a minute with no reinstall required.