FREE · NETWORK & IT INFRASTRUCTURE
Website Server Sizing Calculator.
Estimate peak traffic, concurrent visitors, requests per second, bandwidth and the CPU and RAM a server needs from your monthly website visitors.
Traffic
2 = the busiest day has twice the traffic of an average day. Campaigns, launches and news spikes can reach 5–20×.
4.2 % means perfectly even traffic. 8–12 % is common for regional sites; global audiences are flatter.
Pages and caching
Everything a page downloads: HTML, scripts, styles, images, fonts.
WordPress page cache, Varnish, Nginx FastCGI cache or CDN HTML caching.
Application server
Measure with APM or a load test. Typical: 20–80 ms for light pages, 100–400 ms for heavy CMS or shop pages.
Time a worker stays busy, including database and API waits.
0 if the database runs elsewhere.
- Page views per month
- 300,000
- Peak page views per second
- 0.82
- Peak concurrent visitors
- 50 Little’s law: arrivals per hour × visit duration
- Peak HTTP requests per second
- 58 All requests, before CDN offload
- Uncached pages per second at origin
- 0.82
- Static requests per second at origin
- 56.7
- CPU cores needed
- 0.24 At 60 % utilisation with 50 % growth
- Concurrent workers (PHP-FPM, Node, Puma…)
- 2
- Memory needed
- 2.2 GB
- Peak bandwidth (visitors)
- 16.43 Mbps
- Peak bandwidth (origin server)
- 16.43 Mbps
- Monthly transfer (visitors)
- 750 GB
- Monthly transfer (origin)
- 750 GB
- Capacity of the recommended size
- About 638,044 visitors/month (CPU-bound)
- This is a planning estimate built from your assumptions, not a measurement. Confirm with a load test (k6, Locust, JMeter, Gatling) on a staging copy before buying.
- Peak load, not monthly visitors, decides server size: visitors arrive unevenly, so the peak inputs matter most.
- Months are averaged to 30.44 days. Data sizes are decimal (1 MB = 1,000,000 bytes).
How to use Website Server Sizing Calculator
Enter monthly visitors and pages per visit, then refine the peak, caching and application settings or start from a preset (blog, shop, static site, news spike, SaaS).
How it works
Monthly visitors are converted to the busiest minute using the peak-day, peak-hour and burst factors. Little’s law gives concurrent visitors and workers; CPU is uncached pages × CPU time ÷ target utilisation; bandwidth is page views × page weight.
Example
100,000 visitors a month, 3 pages each and 2.5 MB pages peak at about 0.8 pages per second and 16 Mbps: a 1 vCPU / 4 GB server with room to spare.
From monthly visitors to server size
Analytics report visitors per month, but servers fail at the busiest minute. A site with 100,000 monthly visitors averages fewer than one page view per second, yet a newsletter, a sale or a mention on social media can concentrate a day’s traffic into an hour. This calculator converts monthly numbers into the peak load that decides CPU, memory and bandwidth, and shows each step so you can challenge the assumptions.
How the peak is estimated
- Page views per month = visitors × pages per visit.
- Average day = monthly page views ÷ 30.44.
- Busiest day = average day × the peak-day factor (2 means twice an average day).
- Busiest hour = busiest day × the share of the day in its busiest hour. Perfectly flat traffic would be 4.2 %; 8–12 % is common for audiences in one region.
- Busiest minute = busiest-hour rate × the burst factor.
Concurrent visitors follow Little’s law: arrivals per hour multiplied by the average visit length. Concurrent server workers use the same law with the server response time instead of the visit length.
CPU, memory and bandwidth
Only pages that are actually rendered cost meaningful CPU. The tool multiplies uncached pages per second by the CPU time per page, adds a small cost for cached pages and static files, and divides by a target utilisation (60 % by default) so the server still answers quickly at the peak. Memory is the base system plus the database, if it runs on the same machine, plus one allowance per concurrent worker.
| Setting | What to enter |
|---|---|
| CPU time per uncached page | From APM traces or a load test. Light pages 20–80 ms; heavy CMS or shop pages 100–400 ms. |
| Full-page cache | Share of views served from WordPress page caching, Varnish, Nginx FastCGI cache or CDN HTML caching. |
| CDN offload | Share of images, scripts and styles served by a CDN or the browser cache instead of your server. |
| Memory per worker | The resident size of a PHP-FPM child, Node process or Puma worker, visible in top or htop. |
Bandwidth is page views multiplied by page weight. Many home pages weigh 2 to 3 MB once images and scripts are counted, which is why a CDN often matters more than a faster CPU.
Reading the result
The recommendation rounds up to common cloud sizes (1, 2, 4, 8, 16 vCPU…) and reports how many monthly visitors that size could handle with the same assumptions. Above 32 vCPU it suggests several 16 vCPU servers behind a load balancer plus one spare, which also gives you redundancy. Use the presets to compare a cached blog, an online shop, a static site, a news spike and a SaaS application.
Treat the figure as a starting point. Measure with a load-testing tool such as k6, Locust, JMeter or Gatling against a staging copy, then adjust the CPU time and memory inputs to match what you observe. To budget the result, continue with the cloud cost calculator; for the network link, the bandwidth calculator.
Questions about Website Server Sizing Calculator
How many visitors can a 2 GB server handle?
It depends on caching far more than on RAM. A cached blog can serve hundreds of thousands of visitors a month on a small server; an uncached shop with 300 ms pages may need several times more. Enter your own CPU time and cache rate to see the capacity.
Why size for peaks instead of averages?
Traffic is uneven. A server sized for the monthly average is overloaded during every busy hour, which is exactly when visitors and sales arrive.
Does this replace a load test?
No. It turns your assumptions into numbers. A load test measures how your application actually behaves.
Is Website Server Sizing Calculator free, and do I need an account?
Yes. This tool is free to use in your browser without an account. Processing takes place on your device.
What are the limits and what should I check?
A planning estimate from your assumptions, not a measurement; confirm with a load test. Months are 30.44 days and data sizes decimal. Database servers and multi-region designs are not sized separately. Everything is calculated in your browser; nothing is sent or probed on the network. Review the result before using it in your work.
Is my input sent to a server?
Your input is processed in your browser and is not uploaded by this tool. Files and pasted text are cleared when the page is refreshed. Normal website requests are still required to load the page and its libraries. Read the privacy explanation.