Enter a request volume in any unit — per second, per day, per month — and see it in all the others, with the peak you should plan for and the bandwidth it implies.
Runs entirely in your browser — nothing you enter here is sent to a server.
Traffic is quoted in whatever unit makes the point. Product reports speak of "five million requests a day", a provider bills per million per month, a load test is configured in requests per second, and a rate limit is written per minute. They describe the same thing and are hard to compare in your head: five million a day is only 58 per second.
The conversion is division by the number of seconds in the period, but the average is rarely what you should build for. Traffic follows the working day and the calendar; most sites see two to four times their daily average during the busiest hour, and far more during a campaign or an incident-driven retry storm.
Convert monthly or daily business volume to peak requests per second — the figure servers, pools and autoscaling are sized against.
Set the test rate from production numbers instead of a guess, and test at the peak, not the average.
Check whether a plan's "1 million calls a month" covers your usage: it is an average of only 0.38 requests per second.
Per-request and per-gigabyte pricing both follow from monthly volume and average response size.
Take the busiest hour of a typical week and divide its request rate by the weekly average. Business applications used in one time zone often come out between 3 and 5; global consumer services are flatter, around 1.5 to 2. If you have no data yet, 3 is a reasonable starting assumption — and remember that the worst peaks are not organic traffic but everything retrying at once after an outage.
Divide by 86,400, the number of seconds in a day. One million requests a day is about 11.6 requests per second on average.
About 0.38 on average — a month has roughly 2.63 million seconds. At a peak factor of 3 that is still only a little over one request per second.
None in practice. Requests per second and queries per second are used interchangeably; QPS is more common for databases and search, RPS for web services.
Peak, with headroom on top. A system sized for the average is overloaded every busy hour, which is exactly when failures are most visible.
Upzilla checks your sites, servers and APIs around the clock and alerts you the moment something breaks. 10 monitors free, no card required.