{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Multi-Threading Performance Benchmark\n",
    "\n",
    "This notebook contains a performance comparison of different methods to process the NDVI calculations.\n",
    "\n",
    "The `%%timeit` cell magic runs the cell content multiple times and outputs statistics on those multiple runs, thereby reducing factors such as garbage collection pauses etc."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from multiprocessing import Pool, cpu_count\n",
    "from numpy import ma\n",
    "from pathlib import Path\n",
    "import rasterio as r"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of files: 27\n"
     ]
    }
   ],
   "source": [
    "test_files = list(Path('output/ndvi').glob('*.tif'))\n",
    "print(f'Number of files: {len(test_files)}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The function we test with:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def average(file_path):\n",
    "    with r.open(file_path) as src:\n",
    "        data = src.read(1, masked=True)\n",
    "        return file_path, ma.average(data)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## In a single process\n",
    "### Time to process a single file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "36.2 ms ± 42.6 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)\n"
     ]
    }
   ],
   "source": [
    "%%timeit\n",
    "average(test_files[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Time to process all files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "980 ms ± 7.38 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)\n"
     ]
    }
   ],
   "source": [
    "%%timeit\n",
    "averages = [avg for avg in map(average, test_files)]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Increasing the list size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4.86 s ± 10.5 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)\n"
     ]
    }
   ],
   "source": [
    "%%timeit\n",
    "averages = [avg for avg in map(average, test_files * 5)]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Time when using a worker pool\n",
    "\n",
    "Number of CPUs the multiprocessing pools can access:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "4"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cpu_count()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### On One element"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "277 ms ± 3.92 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)\n"
     ]
    }
   ],
   "source": [
    "%%timeit\n",
    "with Pool() as pool:\n",
    "    averages = [avg for avg in pool.map(average, test_files[:1])]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### On the complete list"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "630 ms ± 8.5 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)\n"
     ]
    }
   ],
   "source": [
    "%%timeit\n",
    "with Pool() as pool:\n",
    "    averages = [avg for avg in pool.map(average, test_files)]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Increasing the list size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2.1 s ± 20 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)\n"
     ]
    }
   ],
   "source": [
    "%%timeit\n",
    "with Pool() as pool:\n",
    "    averages = [avg for avg in pool.map(average, test_files * 5)]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Result\n",
    "\n",
    "As we can see when processing a single element, multiprocessing comes with an overhead.\n",
    "When the list to be processed is sufficiently large, we get a reduction in processing time of roughly 30%-50%, depending on list size.\n",
    "\n",
    "Averaging the masked array is a fairly simple operation that scales in $O(N)$ with the size of the input array.\n",
    "The time reduction should be even higher for more complex tasks."
   ]
  },
  {
   "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": "3.8.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}