group_9_song_distribution-weak-scalability.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from pyspark.sql import SparkSession\n",
    "import time\n",
    "\n",
    "# Start stopwatch\n",
    "startTime = time.time()\n",
    "\n",
    "# Alternate the amount of cores for performance testing\n",
    "\n",
    "# New API\n",
    "spark_session = SparkSession\\\n",
    "        .builder\\\n",
    "        .master(\"spark://192.168.1.153:7077\") \\\n",
    "        .appName(\"group_09_scale_test\")\\\n",
    "        .config(\"spark.dynamicAllocation.enabled\", True)\\\n",
    "        .config(\"spark.shuffle.service.enabled\", True)\\\n",
    "        .config(\"spark.dynamicAllocation.executorIdleTimeout\",\"30s\")\\\n",
    "        .config(\"spark.executor.cores\", 8)\\\n",
    "        .config(\"spark.cores.max\",\"16\")\\\n",
    "        .config(\"spark.executor.instances\",\"2\")\\\n",
    "        .getOrCreate()\n",
    "      \n",
    "        \n",
    "# Old API (RDD)\n",
    "spark_context = spark_session.sparkContext"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import h5py\n",
    "import sys\n",
    "import io\n",
    "\n",
    "# Duplicate the dataset to test for scalability\n",
    "rdd_1 = spark_context.binaryFiles(\"hdfs://192.168.1.153:9000/millionsongs/data/*/*/*/*\")\n",
    "rdd_2 = spark_context.binaryFiles(\"hdfs://192.168.1.153:9000/millionsongs/data/*/*/*/*\")\n",
    "\n",
    "rdd = rdd_1.union(rdd_2) # 2x data\n",
    "rdd = rdd.union(rdd)     # 4x data\n",
    "rdd = rdd.union(rdd)     # 8x data\n",
    "rdd = rdd.union(rdd)     # 16x data\n",
    "     # read a dataset and return it as a Python list #\n",
    "\n",
    "# Function that reads the h5 format and extracts the desired feature, in this case duration\n",
    "def f(x):\n",
    "    # x[0] = filename\n",
    "    # x[1] = binary content\n",
    "    with h5py.File(io.BytesIO(x[1])) as f:\n",
    "        # drill down with a path\n",
    "        #return list(f.keys())\n",
    "        g = f['analysis']['songs']['duration']\n",
    "        # g is a 'dataset'\n",
    "        return list(g)\n",
    "\n",
    "# Apply function to the data set\n",
    "rdd = rdd.map(f).cache()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Function that discretizes the duration into half minutes\n",
    "def time_conversion(duration):\n",
    "    duration_in_half_minutes = duration[0]/30\n",
    "    discretitized_duration = round(duration_in_half_minutes)\n",
    "    discretized_duration_in_minutes = discretitized_duration/2\n",
    "    return discretized_duration_in_minutes\n",
    "\n",
    "# Map function to the durations in the RDD\n",
    "converted_rdd = rdd.map(time_conversion)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Time Elapsed 7.5683817585309345\n"
     ]
    }
   ],
   "source": [
    "# Map reduce by duration length\n",
    "result = converted_rdd.map(lambda time: (time, 1)).reduceByKey(lambda a, b: a + b).sortBy(lambda x: x[0], True)\n",
    "\n",
    "# Stop stopwatch and print time elapsed\n",
    "endTime = time.time()\n",
    "print(\"Time Elapsed {}\".format((endTime-startTime)/60))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot the song duration distribution\n",
    "import matplotlib.pyplot as plt\n",
    "plt.bar(*zip(*result.collect()))\n",
    "plt.show()"
   ]
  },
  {
   "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": "3.6.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}