Calculate data statistics in bash, including pencentile, min, max, average, standard deviation. Suitable for inspecting system latencies distribution.

stats.sh

#!/bin/bash
# Calculate data statistics, including pencentile, min, max, average, standard deviation.
# Derived from https://gist.github.com/lewisd32/4be2605400acf0bb562d
#
# Usage & Output
# $ cat nums.txt | ./stats.sh
# total 100
# p50 49
# p90 89
# p95 94
# min 0
# max 99
# avg 49.5
# stddev 29.0115

# stdin should be integers, one per line.

function random_string () {
    length=${1:-6}
    s=`tr -dc A-Za-z0-9 </dev/urandom | head -c $length`
    echo $s
}

tmp="/tmp/stats_`random_string 6`"
total=$(sort -n | tee "$tmp" | wc -l)
echo "total $total"

percentiles=(50 90 95)
for percentile in ${percentiles[@]}; do
    # (n + 99) / 100 with integers is effectively ceil(n/100) with floats
    pos=$(((total * percentile + 99) / 100))
    printf "p${percentile} "
    sed -n ${pos}p "$tmp"
done

awk '
(NR==1){min=$0; max=$0}
(NR>=2){
    if ($0 > max) max = $0;
    if ($0 < min) min = $0;
}
{
    sum+=$0;
    a[NR]=$0
}
END {
    avg=sum/NR;
    for(i in a) y+=(a[i]-avg)^2;
    print "min", min;
    print "max", max;
    print "avg", avg;
    print "stddev", sqrt(y/(NR-1));
}' "$tmp"

rm "$tmp"