Technology · Free tool
Server Capacity Calculator
Before a launch, a sale or a seasonal peak, someone has to say how many application workers or instances to run. This calculator gives a first estimate from three numbers: peak requests per second, average handling time per request, and the headroom you want on top. It suits engineering leads and finance partners sizing a budget for infrastructure.
- OutputWorkers to run
- No sign-up
- Copy or print the result
- How it works ↓
Your entries stay in this browser. Use Fill example to see a finished page first.
What this technology tool gives you
The output shows workers to run as a whole number, plus the exact load with headroom before rounding. The note spells out the assumption behind it. Use it as a starting point in a capacity review, then confirm it with a load test before the peak arrives.
What you enter
- Peak requests per second
- Average handling time per request, in milliseconds
- Headroom, as a percentage above peak load
Worked example
The example uses a peak of 20 requests per second, 50 milliseconds average handling and 25% headroom. The load is 20 × 0.05 = 1 busy worker, and with headroom 1 × 1.25 = 1.25. The output shows Exact load with headroom: 1.25 and Workers to run: 2, because the tool rounds up.
How do you calculate the number of workers needed?
Workers = peak requests per second × (handling milliseconds ÷ 1,000) × (1 + headroom ÷ 100). The exact figure is rounded to two decimals, and workers to run rounds it up to the next whole number.
The assumption is that each worker handles one request at a time. Peak requests per second multiplied by seconds per request gives the number of requests in progress at once, which is the number of busy workers. Headroom adds a buffer on top of that.
Tips before you use the result
- Use a high-percentile handling time, such as p95, rather than a pure average if slow requests cluster at peak.
- If a worker handles several requests concurrently, divide the result by that concurrency; the tool assumes one at a time.
- Add at least one more worker than the result if you need to survive losing an instance or a zone.
- Measure peak requests per second over short intervals; a per-minute average hides the spikes that matter.
Frequently asked questions
How many servers do I need for my traffic?
A first estimate is peak requests per second multiplied by average handling time in seconds, which gives concurrent requests, plus headroom. This calculator applies that formula and rounds up to whole workers. Real capacity depends on concurrency per worker, memory and dependencies, so confirm with a load test.
What headroom percentage should I use?
The tool accepts any percentage and does not prescribe one. The example uses 25%. Teams often choose more for unpredictable traffic or services that are slow to scale, and less when autoscaling reacts quickly. Pick a number you can justify from past peaks.
Why does the calculator round up?
You cannot run a fraction of a worker, and rounding down would leave the peak under-served. So an exact load of 1.25 becomes 2 workers to run. The exact figure is still shown so you can see how close you are to the next step.
Does the tool save my capacity inputs?
Yes, within this browser. The three numbers are kept locally with no account and no upload, so they reload next time you open the page on this device. Copy output gives the result and the assumption as plain text for a ticket or planning doc.
More Technology tools
Technology guides on Kurums
Nasdaq Falls 5.5% as Midterms Raise AI Regulation Fears: Notes for Technology Leaders
The Nasdaq fell 5.5% into correction territory on 5 November 2026 as investors priced in tougher AI data-centre rules. What technology leaders should do next.
The Midterm Chill on AI Data Centers: What a Democratic Congress Means for Capacity Planning
Markets read the 2026 midterms as a brake on AI data-center buildout. Here is what is known, what is not, and how technology teams should plan capacity.
Markets Bet Democratic Sweep Will Slow AI Spending, November 4, 2026: Notes for Technology
Following Democratic majorities in both chambers after the November 3 midterms, technology shares rose on November 4 as markets priced a potential slowdown in AI data-center spending.
Roadmap, Decoded: What It Actually Means
A roadmap is a plan that shows what a product or technology team intends to work on over the coming months and why. It sets direction and priorities, grouping work into themes or time horizons, so everyone from engineers to sales knows what is coming, what is later and what is not planned at all.
AMD’s Data Center Revenue Nearly Doubles: What the Q3 Beat Means for Tech Leaders
AMD Q3 2026 data center revenue rose 97% to $8.5B. What it signals for AI infrastructure planning, supply and vendor strategy.
Supermarket Freezers Hacked Across Europe: What Happened and What Businesses Must Fix
Shops in seven European countries reported freezers warming after a suspected cyber-attack. What is known, why it matters, and a checklist for retailers.
Results depend on the figures you enter; check them against your own records before you rely on them. Last updated: October 2026 · Kurums editorial team.