Free Server Request Rate calculator with clear step-by-step results.
Average request rate is total requests divided by the seconds in the period, but capacity is never sized on the average. Real traffic is peaky, and a peak-to-average ratio of 2-4 is common for consumer-facing services with a daily cycle. Little's law then converts the peak rate into concurrency: requests in flight = arrival rate × average time in the system, which is the number that determines how many workers, connections or threads you need.
Request rate
Requests per second = total requests ÷ (period hours × 3600)
Concurrency (Little's law)
Concurrent requests = peak rate × average response time
Traffic modelled as a flat average with a peak multiplier is a planning approximation. Size production capacity against measured percentiles and load-test results.
Because pools are sized in slots, not in requests per second. At 100 req/s with a 250 ms response you need about 25 concurrent slots; the same rate with a 2-second response needs 200, and the pool exhausts long before CPU does.
Measure it if you can. Where you cannot, 2-3 is typical for a service with a single daily peak, 5-10 for event-driven or campaign-driven traffic, and higher still for anything triggered by a scheduled job or a broadcast.
No — include them in the total request count. Retry storms are a common cause of the peak ratio being far worse than expected, since failures generate additional load precisely when the system is struggling.