Systems Capacity Calculators

Compare growth classes (n, n log n, n^2, 2^n, n!) at a given n: operation counts, run time, largest n a time budget allows.

Units needed for a peak load at a target utilisation plus N+k redundancy, utilisation after failures, and growth projection.

Erlang B and C: trunks for a blocking target, or agents for a service level such as 80 percent answered in 20 seconds.

Add up serial stage latencies against a target (share per stage, slack, bottleneck) and compute tail amplification from fan-out.

Little's law L = rate x time in system: solve for concurrency, throughput or latency given any two; sizes pools.

M/M/1 and M/M/c queue: utilisation, probability of waiting (Erlang C), queue length, wait and response time, tail odds.

RAID 0/1/5/6/10 and ZFS RAID-Z usable capacity, efficiency, fault tolerance, write penalty and rebuild read-error risk.

SLA availability (99.9 percent, three nines) to allowed downtime per day, month and year; series and parallel composition.

Requests per second times payload size to Mbps with peak factor, protocol overhead and link headroom; or IOPS to MB/s.