masters-thesis/plots/plot_histograms.ipynb
Dennis af25b4b828 Initial commit of master's thesis
This is the version I submitted to RWTH Aachen University at November 9,
2018.
2018-11-12 12:56:59 +01:00

384 lines
13 KiB
Plaintext
Executable File

{
"cells": [
{
"cell_type": "markdown",
"metadata": {
"collapsed": true
},
"source": [
"# Description of Functionality\n",
"This script loads csv files with the following information:\n",
"* Time the client enqueues the packet (`enq`) or time the clien actually sends the packet (`send`)\n",
"* Time the client gets a Work Completion (`send_wc`)\n",
"* Time the server receives the packet (`recv`)\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Algorithm\n",
"\n",
"## Load settings from JSON file\n",
"\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"import os\n",
"from sys import argv\n",
"rootdir = argv[1]\n",
"\n",
"#############################\n",
"# FOR NOTEBOOK USE #\n",
"# SET DIRECTORY HERE #\n",
"# #\n",
"#rootdir = \"\"\n",
"# #\n",
"#############################\n",
"\n",
"print(\"Using root directory: {}\".format(rootdir))\n",
"\n",
"subdirs = sorted([ name for name in os.listdir('{}'.format(rootdir)) if os.path.isdir(os.path.join('{}'.format(rootdir), name)) ])\n",
"\n",
"print(\"Available subdirs: {}\".format(subdirs))"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"import json\n",
"from sys import exit\n",
"\n",
"try:\n",
" with open(\"{}/settings.json\".format(rootdir)) as json_file:\n",
" settings = json.load(json_file)\n",
"except:\n",
" print(\"Please define a correct JSON file!\")\n",
" exit()\n",
"\n",
"print(\"Succesfully loaded JSON file\")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Import\n",
"First, import the numpy library, initialize the arrays, and finally load the csv files. \n",
"\n",
"Because of the way the C script dumps the variables, the last character of the csv-file will be a comma and thus the last value of the `*_times` arrays will be `NaN`. Hence, the last value has to be eliminated."
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"import numpy as np\n",
"\n",
"# Initialize arrays\n",
"enq_send = []\n",
"recv = []\n",
"\n",
"if not settings['compare_tests']:\n",
" send_wc = []\n",
"\n",
"# Load all data and remove the last comma.\n",
"# This for loop distinguish between tests which measure the enqueue time and tests which\n",
"# measure the actual send time.\n",
"for i, subdir in enumerate(subdirs):\n",
" enq_send.append(np.genfromtxt('{}/{}/enq_send_times.csv'.format(rootdir, subdir), delimiter=','))\n",
" recv.append(np.genfromtxt('{}/{}/recv_times.csv'.format(rootdir, subdir), delimiter=','))\n",
" \n",
" # Remove last comma\n",
" enq_send[i] = np.delete(enq_send[i], -1)\n",
" recv[i] = np.delete(recv[i], -1)\n",
" \n",
" if not settings['compare_tests']:\n",
" send_wc.append(np.genfromtxt('{}/{}/send_wc_times.csv'.format(rootdir, subdir), delimiter=','))\n",
" \n",
" # Remove last comma\n",
" send_wc[i] = np.delete(send_wc[i], -1)\n",
"\n",
" #Print number of datapoints\n",
" print('Loaded {} + {} datapoints from {}.'.format(np.size(enq_send[i]), np.size(recv[i]), subdir))"
]
},
{
"cell_type": "markdown",
"metadata": {
"collapsed": true
},
"source": [
"## Process data\n",
"Now, the data is processed. First, a check for overflows have to be performed. The timestamps are determined with the function `clock_gettime(clockid_t clk_id, const struct timespec *tp)`. Both the `struct tp`, as well as the function are showed below\n",
"\n",
"```\n",
"struct timespec {\n",
" time_t tv_sec; /* seconds */\n",
" long tv_nsec; /* nanoseconds */\n",
"} tp;\n",
"\n",
"clock_gettime(CLOCK_MONOTONIC, &tp);\n",
"```\n",
"\n",
"The application only sends the `long tv_nsec` value, which goes from 999999999ns to 0ns. Since transmissions cannot take longer than 1 second, this overflow is resolved by adding 1000000000ns to the receive timestamps and the send confirmation timestamps, if they are smaller than the enqueue- or send timestamps.\n",
"\n",
"Subsequentely, the deltas between the enqueue- or send time and the receive time, and the delta between the enqueue- or send time and the send confirmation time are calculated."
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"#Initialize arrays\n",
"enq_send_recv_d = []\n",
"\n",
"if not settings['compare_tests']:\n",
" enq_send_send_wc_d = []\n",
"\n",
"#Resolve overflow issues and then calculate deltas\n",
"for i in range(0, len(subdirs)):\n",
" recv[i][recv[i] < enq_send[i]] += 1000000000\n",
" enq_send_recv_d.append(recv[i] - enq_send[i])\n",
" \n",
" if not settings['compare_tests']:\n",
" send_wc[i][send_wc[i] < enq_send[i]] += 1000000000\n",
" enq_send_send_wc_d.append(send_wc[i] - enq_send[i])"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Plotting\n",
"\n",
"The data will now be plotted."
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# Define \"find nearest\" function\n",
"def find_nearest(array, value):\n",
" array = np.asarray(array)\n",
" idx = (np.abs(array - value)).argmin()\n",
" return array[idx], idx"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"scrolled": false
},
"outputs": [],
"source": [
"import matplotlib.pyplot as plt\n",
"from matplotlib.font_manager import FontProperties\n",
"\n",
"x_limit = 10000\n",
"plots_saved = 0\n",
"\n",
"#Start creating plots\n",
"for i in range(0, len(subdirs)):\n",
" datasets = [] \n",
"\n",
" if settings['compare_tests']:\n",
" if i % 2 == 1:\n",
" continue\n",
" \n",
" datasets.append(enq_send_recv_d[i])\n",
" datasets.append(enq_send_recv_d[i+1])\n",
" else:\n",
" datasets.append(enq_send_recv_d[i])\n",
" datasets.append(enq_send_send_wc_d[i])\n",
"\n",
" medians = []\n",
" medians.append(np.median(datasets[0]))\n",
" medians.append(np.median(datasets[1]))\n",
" \n",
" # Determine correction, in case figure needs to be bigger\n",
" correction = 0\n",
" if abs(medians[1] - medians[0]) < 200:\n",
" correction = 0.2\n",
"\n",
" fig = plt.figure(num=None, figsize=(11, 2.7 + correction), dpi=500, facecolor='w', edgecolor='k')\n",
"\n",
" # Add plot and set title\n",
" ax = fig.add_subplot(111)\n",
"\n",
" # Set grid\n",
" ax.set_axisbelow(True)\n",
" ax.grid(True, linestyle='--')\n",
"\n",
" bins = np.arange(0, x_limit+1, 100.0)\n",
"\n",
" # Data in plot\n",
" # http://www.color-hex.com/color-palette/33602\n",
" ax.hist(datasets[1], label=settings['labels'][1], edgecolor='black', bins=bins, color='#00549f', zorder=1)\n",
" ax.axvline(medians[1], color='red', linestyle='-', linewidth=1, zorder=2, alpha=0.85)\n",
"\n",
" ax.hist(datasets[0], label=settings['labels'][0], edgecolor='black', bins=bins, color='#8ebae5', zorder=3, alpha=0.75)\n",
" ax.axvline(medians[0], color='red', linestyle='-', linewidth=1, zorder=4)\n",
"\n",
" # Set axis\n",
" plt.xlim([0,x_limit])\n",
" \n",
" # Calculate how many values are larger than the x_limit\n",
" errors = []\n",
" errors.append((np.size(datasets[0][datasets[0] > x_limit]) / np.size(datasets[0])) * 100)\n",
" errors.append((np.size(datasets[1][datasets[1] > x_limit]) / np.size(datasets[1])) * 100)\n",
" \n",
" errors[0] = round(errors[0], 4)\n",
" errors[1] = round(errors[1], 4)\n",
" \n",
" # Set ticks\n",
" ticks_unmodified = ticks = np.arange(0, x_limit+1, 1000.0)\n",
"\n",
" nearest = [None] * 2\n",
" nearest_idx = [None] * 2\n",
" \n",
" nearest[0], nearest_idx[0] = find_nearest(ticks, medians[0])\n",
" nearest[1], nearest_idx[1] = find_nearest(ticks, medians[1])\n",
" \n",
" if medians[0] < medians[1]:\n",
" ticks = np.append(ticks, medians[0])\n",
" ticks = np.append(ticks, medians[1])\n",
" else:\n",
" ticks = np.append(ticks, medians[1])\n",
" ticks = np.append(ticks, medians[0])\n",
"\n",
" # Explicitly set labels\n",
" labels = []\n",
" \n",
" for value in ticks:\n",
" if value == nearest[0] and np.abs(nearest[0] - medians[0]) < 200:\n",
" labels.append(\"\")\n",
" elif value == nearest[1] and np.abs(nearest[1] - medians[1]) < 200:\n",
" labels.append(\"\")\n",
" else:\n",
" labels.append(str(int(value)))\n",
"\n",
" # Set xticks\n",
" plt.xticks(ticks, labels, fontsize=10, family='monospace', rotation=30, horizontalalignment='right', rotation_mode=\"anchor\")\n",
" \n",
" # Color median values red\n",
" first_median_is_set = False\n",
" \n",
" for j, value in enumerate(ax.get_xticklabels()):\n",
" try:\n",
" if float(value.get_text()) == int(medians[0]) or float(value.get_text()) == int(medians[1]):\n",
" value.set_color('red')\n",
" \n",
" if abs(medians[0] - medians[1]) < 170 and first_median_is_set:\n",
" value.set_y(-0.07)\n",
"\n",
" nearest, nearest_idx = find_nearest(ticks_unmodified, float(value.get_text()))\n",
" \n",
" if abs(nearest - float(value.get_text())) < 350:\n",
" ax.get_xticklabels()[nearest_idx].set_y(-0.07)\n",
" \n",
" first_median_is_set = True\n",
" \n",
" except ValueError:\n",
" # We got some empty values. Ignore them\n",
" pass\n",
" \n",
" # Set yticks\n",
" plt.yticks(fontsize=10, family='monospace')\n",
"\n",
" #Labels\n",
" font_text = FontProperties()\n",
" font_text.set_size(9.5)\n",
" font_text.set_family('monospace')\n",
" \n",
" ax.set_xlabel('latencies [ns]', fontsize=10, family='monospace', labelpad = 4 - 2 * correction)\n",
" ax.set_ylabel('frequency', fontsize=10, family='monospace', labelpad = 6)\n",
"\n",
" test = settings['labels'][1] + '$\\mathtt{{> {}\\/ns: }}${: >7.4f}% (max: {:8} ns)\\n'.format(x_limit, errors[1], max(datasets[1]))\n",
" test += settings['labels'][0] + '$\\mathtt{{> {}\\/ns: }}${: >7.4f}% (max: {:8} ns)'.format(x_limit, errors[0], max(datasets[0]))\n",
" \n",
" # bbox accepts FancyBboxPatch prop dict\n",
" x_position_box = 0.99 if medians[1] < 6000 else 0.38\n",
" \n",
" ax.text(x_position_box, 0.95, test,\n",
" verticalalignment='top', horizontalalignment='right',\n",
" transform=ax.transAxes, zorder=5,\n",
" color='black', fontproperties = font_text,\n",
" bbox={'facecolor':'white', 'alpha':0.85, 'pad':0.30, 'boxstyle':'round',\n",
" 'edgecolor':'#dbdbdb'})\n",
"\n",
" # Show plot\n",
" plt.yscale('log')\n",
" plt.tight_layout()\n",
" \n",
" # Save plot\n",
" fig.savefig('{}/plot_{}.pdf'.format(rootdir, plots_saved), dpi=600, format='pdf')\n",
" plots_saved += 1\n",
" \n",
" if i == 0:\n",
" # Create and save legend\n",
" import pylab\n",
" \n",
" # create a second figure for the legend\n",
" figLegend = pylab.figure(figsize = settings['dimensions']['legend'])\n",
"\n",
" # produce a legend for the objects in the other figure\n",
" pylab.figlegend(*ax.get_legend_handles_labels(), loc = 'upper left',\n",
" prop={'family':'monospace', 'size':'8'}, ncol=2)\n",
" figLegend.savefig(\"{}/legend.pdf\".format(rootdir), format='pdf')"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.7.0"
}
},
"nbformat": 4,
"nbformat_minor": 2
}