How much RAM does a VPS actually need?

RAM by workload

Prices exclude 18% GST. The rest of this explains where those numbers come from.

What actually uses the memory

On a typical LAMP or LEMP box:

• The OS — 200–400 MB for a minimal Ubuntu or AlmaLinux with no desktop.
• MySQL / MariaDB — 500 MB to 2 GB. The InnoDB buffer pool is the big one, and it should ideally hold your whole working data set.
• PHP-FPM — 40–120 MB per worker, and a WordPress site with a page builder sits at the top of that range. Ten workers at 100 MB is a gigabyte.
• The web server — Nginx is 50–100 MB, Apache with mod_php considerably more, LiteSpeed in between.
• Redis, if used — whatever you cap it at, typically 256 MB to 2 GB.
• cPanel, if installed — 1 GB before you host anything. Budget for it.

Add those up for your stack, then add 25–30% headroom. Memory that is full at idle has nothing left for your busiest hour.

Why running out is worse than being slow

A CPU-starved server is slow. A memory-starved server fails. Linux swaps to disk first — orders of magnitude slower — and when that is exhausted the out-of-memory killer picks the largest process and kills it. That is almost always MySQL, and the visible symptom is "Error establishing a database connection" until something restarts it.

This is why the honest advice is to size RAM generously and CPU tightly. CPU pressure degrades; RAM pressure breaks.

Rules of thumb that hold up

• WordPress, cached: ~1 GB per busy site, including its share of MySQL.
• WooCommerce: double it. Cart, checkout and account pages cannot be page-cached, so they hit PHP and MySQL on every request.
• Minecraft: ~1 GB per ten vanilla players, plus whatever the modpack asks for, rounded up.
• Tally on Cloud: Windows Server needs 4 GB before Tally starts; 8 GB covers one or two concurrent users comfortably.
• Node and Python APIs: measure. They vary more than anything else on this list.

Check what you are using now

# free memory, human readable
free -h

# the ten biggest processes by memory
ps aux --sort=-%mem | head -11

# is it swapping?
vmstat 1 5

In free -h, look at available, not free — Linux uses spare memory for disk cache and hands it back on demand, so a low "free" figure is normal and healthy. In vmstat, non-zero si/so columns mean active swapping, which means buy more memory.

Do not over-buy either

A 32 GB server running a 3 GB workload is money spent on nothing. Our plans resize in place — vCPU, RAM and disk grow with a single reboot, and nothing is reinstalled — so the right strategy is to start one size down and move up when the numbers say to. If it is CPU rather than RAM that runs short, a VDS with dedicated cores is the next step, and dedicated servers beyond that.

If you tell us the workload on +91 75994 50220, we will size it with you. We would rather sell you the ₹674 (US$7.66) plan that fits than the ₹2,696 (US$30.64) one that does not.

The full ladder is on our KVM VPS hosting page, and Best VPS hosting in India compares it with other providers. Machine-learning training and inference that need a GPU are a different sizing exercise: see GPU server hosting in India.

Sources (checked 13 November 2026)

• man7.org: free(1): Linux manual page
• MariaDB Knowledge Base: InnoDB Buffer Pool
• WordPress.org: Requirements