Add web server benchmarking script (tsenart#451)

This script will automatically run vegeta against a target with different request
rates and graph the latency distribution and success rate at each request rate.

Usage:

```
echo GET http://localhost:8080/ | ./ramp-requests.py
```

Dependencies:

* vegeta
* Python 3
* Gnuplot

For more documentation, see https://github.com/tsenart/vegeta/wiki/Load-ramping

scripts/load-ramping/README.md

@@ -0,0 +1,17 @@
+# Load ramping
+
+This script will automatically run vegeta against a target with different request
+rates and graph the latency distribution and success rate at each request rate.
+
+Usage:
+
+```
+echo GET http://localhost:8080/ | python3 ramp-requests.py
+```
+
+Dependencies:
+
+* Python 3
+* Gnuplot
+
+For more documentation, see https://github.com/tsenart/vegeta/wiki/Load-ramping

scripts/load-ramping/ramp-requests.plt

@@ -0,0 +1,75 @@
+# Two plots: success rate plot on top, rate/latency distribution below
+set multiplot layout 2,1
+
+
+#
+# Shared config
+#
+
+# Scale (x/color)
+set autoscale xfix
+set logscale xycb 10
+
+# Grid
+set mxtics 10
+set mytics 10
+set tics scale 0.0000000001  # Tics themselves can't be styled indepedently, so use grid only
+set grid xtics ytics mxtics mytics lc rgb '#888888' lw 0.5 lt 1, lc rgb '#888888' lt 1 lw 0.1
+
+
+#
+# Top plot: success rate
+#
+
+# Manual positioning to align both plots
+set lmargin at screen 0.10
+set rmargin at screen 0.87
+set bmargin at screen 0.80
+set tmargin at screen 0.95
+
+# Scale (y only)
+set yrange [1.0:100.0]
+
+# Axes
+unset xlabel
+set xtics format ""
+set ylabel "Success"
+set ytics format "%.2f%%"
+
+# Plot (incl. fraction to percentage conversion)
+plot "results_success.txt" using 1:($2 * 100.0):(0.0) with line lw 3 lc rgb "red" title ""
+
+
+#
+# Bottom plot: rate vs latency
+#
+
+# Manual positioning to align both plots
+set lmargin at screen 0.10
+set rmargin at screen 0.87
+set tmargin at screen 0.75
+set bmargin at screen 0.15
+
+# Scale (y only)
+unset yrange
+set autoscale yfix
+
+# Axes
+set xlabel "Requests (per sec)"
+set xtics format "%.f"
+set ylabel "Latency" offset -1.5,0,0
+set ytics ( \
+    "1ns" 1.0e0, "10ns" 1.0e1, "100ns" 1.0e2, \
+    "1us" 1.0e3, "10us" 1.0e4, "100us" 1.0e5, \
+    "1ms" 1.0e6, "10ms" 1.0e7, "100ms" 1.0e8, \
+    "1s" 1.0e9, "10s" 1.0e10, "100s" 1.0e11 )
+
+# Color box
+set cblabel ""
+set cbrange[0.001:100.0]
+set format cb "%.9g%%"
+
+# Plot (incl. fraction to percentage conversion)
+set datafile separator " "
+set pm3d map corners2color c1
+splot "results_latency.txt" u 1:2:($3 * 100.0) with pm3d title ""

scripts/load-ramping/ramp-requests.py

@@ -0,0 +1,66 @@
+#!/usr/bin/env python3
+
+import json
+import os
+import subprocess
+import sys
+import time
+
+
+if '-h' in sys.argv or '--help' in sys.argv:
+    print('usage:', file=sys.stderr)
+    print('echo GET http://localhost:8080/ | %s' % sys.argv[0], file=sys.stderr)
+    sys.exit(1)
+
+target = sys.stdin.read().strip()
+
+
+# Log-spaced rates (each ca. +25% (+1dB) of the previous, covering 1/sec to 100k/sec)
+rates = [10.0 ** (i / 10.0) for i in range(50)]
+
+# Log-spaced buckets (each ca. +25% (+1dB) of the previous, covering <1us to >10s)
+buckets = [0] + [1e3 * 10.0 ** (i / 10.0) for i in range(71)]
+
+
+# Run vegeta attack
+for rate in rates:
+    filename='results_%i.bin' % (1000*rate)
+    if not os.path.exists(filename):
+        cmd = 'vegeta attack -duration 5s -rate %i/1000s -output %s' % (1000*rate, filename)
+        print(cmd, file=sys.stderr)
+        subprocess.run(cmd, shell=True, input=target, encoding='utf-8')
+        time.sleep(5)
+
+
+# Run vegeta report, and extract data for gnuplot
+with open('results_latency.txt', 'w') as out_latency, \
+     open('results_success.txt', 'w') as out_success:
+
+    for rate in rates:
+        cmd = 'vegeta report -type=json -buckets \'%s\' results_%i.bin' \
+            % ("[%s]" % ",".join("%ins" % bucket for bucket in buckets), 1000*rate)
+        print(cmd, file=sys.stderr)
+        result = json.loads(subprocess.check_output(cmd, shell=True))
+
+        # (Request rate, Response latency) -> (Fraction of responses)
+        for latency, count in result['buckets'].items():
+            latency_nsec = float(latency)
+            fraction = count / sum(result['buckets'].values()) * result['success']
+            print(rate, latency_nsec, fraction, file=out_latency)
+        print(file=out_latency)
+
+        # (Request rate) -> (Success rate)
+        print(rate, result['success'], file=out_success)
+
+print('# wrote results_latency.txt and results_success.txt', file=sys.stderr)
+
+
+# Visualize with gnuplot (PNG)
+cmd = 'gnuplot -e "set term png size 1280, 800" ramp-requests.plt > result.png'
+print(cmd, file=sys.stderr)
+subprocess.run(cmd, shell=True)
+
+# Visualize with gnuplot (default, likely a UI)
+cmd = 'gnuplot -persist ramp-requests.plt'
+print(cmd, file=sys.stderr)
+subprocess.run(cmd, shell=True)