Created Notebook on Chaining

RDD_Chaining_Execution.ipynb

@@ -0,0 +1,198 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Chaining\n",
+    "We can **chain** transformations and aaction to create a computation **pipeline**\n",
+    "Suppose we want to compute the sum of the squares\n",
+    "$$ \\sum_{i=1}^n x_i^2 $$\n",
+    "where the elements $x_i$ are stored in an RDD."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Start the `SparkContext`"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import numpy as np\n",
+    "from pyspark import SparkContext\n",
+    "sc = SparkContext(master=\"local[4]\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "7, 6, 4, 7, 9, 0, 3, 7, 6, 2, 7, 7, 1, 5, 3, 0, 3, 9, 2, 4, 4, 9, 5, 8, 9, 8, 3, 9, 3, 5, 1, 1, 9, 9, 0, 0, 5, 9, 2, 1, 1, 6, 9, 6, 3, 0, 4, 1, 3, 4, 1, 6, 3, 9, 1, 3, 7, 7, 1, 3, 3, 8, 6, 5, 5, 8, 3, 0, 6, 2, 7, 7, 7, 2, 0, 3, 7, 4, 4, 4, 7, 1, 9, 2, 7, 8, 8, 4, 7, 1, 9, 9, 9, 6, 5, 2, 7, 7, 3, 3, 0, 0, 9, 9, 7, 3, 6, 5, 1, 5, 9, 4, 8, 3, 2, 6, 7, 4, 8, 6, 8, 7, 5, 2, 5, 5, 9, 0, 6, 4, 7, 8, 2, 6, 6, 0, 7, 7, 1, 7, 6, 0, 8, 8, 1, 8, 0, 3, 9, 1, 1, 8, 4, 6, 0, 1, 4, 8, 0, 0, 6, 4, 4, 4, 8, 4, 1, 0, 0, 1, 5, 1, 1, 6, 6, 4, 7, 8, 2, 1, 6, 6, 1, 7, 9, 0, 3, 9, 3, 6, 7, 1, 5, 2, 3, 9, 1, 0, 9, 6, 3, 7, 9, 4, 2, 0, 5, 0, 8, 2, 8, 5, 8, 8, 1, 0, 3, 9, 2, 3, 6, 0, 0, 7, 6, 6, 5, 4, 2, 4, 8, 0, 4, 2, 6, 7, 8, 5, 9, 7, 0, 2, 0, 6, 1, 0, 0, 6, 3, 9, 1, 8, 7, 9, 5, 9, 0, 2, 9, 3, 7, 3, 3, 4, 8, 3, 5, 6, 5, 4, 4, 3, 9, 0, 0, 4, 1, 9, 9, 7, 9, 7, 0, 9, 8, 2, 6, 1, 4, 3, 3, 7, 8, 1, 1, 9, 1, 8, 4, 1, 3, 0, 3, 3, 1, 2, 2, 5, 9, 9, 1, 9, 4, 1, 4, 7, 8, 5, 8, 3, 9, 4, 6, 0, 5, 3, 6, 1, 6, 3, 8, 7, 3, 9, 0, 5, 9, 8, 5, 0, 7, 2, 8, 7, 6, 5, 2, 9, 3, 5, 6, 3, 4, 9, 3, 4, 3, 9, 9, 0, 5, 0, 2, 5, 7, 6, 7, 6, 7, 7, 6, 6, 0, 3, 5, 3, 7, 5, 9, 6, 3, 9, 2, 5, 1, 5, 7, 0, 5, 8, 5, 9, 0, 8, 0, 2, 5, 5, 1, 2, 9, 3, 1, 7, 2, 2, 6, 2, 9, 4, 0, 7, 9, 8, 9, 4, 2, 0, 7, 5, 5, 4, 2, 4, 8, 0, 3, 8, 0, 2, 3, 8, 5, 1, 5, 9, 4, 6, 8, 5, 8, 4, 0, 0, 0, 6, 1, 1, 8, 8, 5, 9, 0, 5, 9, 2, 8, 9, 4, 2, 4, 6, 2, 2, 6, 4, 4, 8, 1, 9, 6, 0, 7, 0, 5, 9, 1, 0, 6, 1, 6, 8, 7, 1, 8, 4, 8, 7, 1, 0, 8, 4, 8, 1, 9, 9, 1, 5, 4, 6, 7, 4, 4, 0, 1, 3, 0, 0, 1, 8, 8, 4, 5, 5, 4, 4, 8, 2, 0, 5, 1, 8, 5, 2, 0, 9, 8, 8, 7, 1, 0, 8, 6, 8, 3, 3, 6, 7, 6, 6, 6, 6, 9, 8, 6, 8, 5, 8, 6, 9, 2, 1, 0, 6, 8, 7, 6, 5, 8, 3, 3, 4, 3, 7, 9, 3, 8, 7, 8, 7, 5, 0, 3, 4, 7, 6, 5, 8, 9, 9, 5, 4, 0, 8, 4, 7, 5, 8, 7, 1, 4, 2, 6, 5, 8, 5, 7, 9, 8, 6, 0, 0, 8, 6, 1, 0, 0, 6, 4, 6, 1, 7, 7, 9, 0, 8, 7, 8, 0, 8, 8, 6, 3, 3, 8, 7, 3, 2, 1, 7, 7, 5, 7, 8, 3, 1, 7, 2, 7, 8, 5, 6, 3, 7, 8, 0, 6, 8, 7, 7, 3, 7, 0, 7, 7, 8, 5, 6, 4, 2, 0, 8, 7, 6, 3, 2, 3, 9, 4, 2, 3, 1, 1, 0, 1, 9, 3, 2, 6, 4, 8, 7, 0, 4, 2, 1, 2, 5, 4, 8, 6, 2, 2, 7, 7, 8, 6, 8, 2, 8, 9, 4, 7, 2, 1, 5, 5, 1, 9, 2, 1, 1, 4, 2, 2, 1, 6, 8, 2, 4, 2, 3, 7, 8, 9, 8, 6, 1, 1, 1, 9, 3, 6, 8, 2, 1, 2, 4, 1, 9, 4, 9, 6, 1, 0, 0, 1, 3, 8, 1, 3, 9, 8, 9, 5, 4, 9, 1, 6, 6, 3, 1, 2, 7, 5, 4, 7, 7, 4, 7, 1, 0, 0, 3, 4, 5, 4, 4, 5, 4, 1, 5, 2, 8, 1, 8, 0, 0, 2, 5, 5, 2, 2, 0, 3, 4, 3, 6, 0, 2, 6, 7, 7, 6, 7, 9, 6, 3, 6, 2, 4, 5, 0, 6, 7, 6, 4, 6, 5, 1, 1, 6, 9, 7, 6, 7, 6, 5, 2, 2, 7, 3, 7, 7, 2, 7, 1, 1, 2, 1, 9, 3, 8, 4, 5, 1, 7, 0, 3, 6, 5, 1, 2, 1, 8, 8, 0, 7, 2, 4, 6, 6, 8, 0, 5, 4, 6, 4, 2, 3, 2, 2, 8, 5, 2, 8, 2, 8, 2, 2, 4, 0, 4, 2, 9, 7, 9, 1, 2, 0, 4, 4, 9, 6, 3, 5, 3, 7, 1, 2, 6, 0, 7, 8, 7, 8, 1, 6, 9, 4, 1, 5, 5, 9, 3, 6, 9, 5, 1, 7, 3, 0, 8, 7, 5, 5, 2, 4, 2, 0, 9, 6, 0, 1, 5, 9, 2, 7, 1, 8, 3, 2, 9, 8, 6, 9, 4, 5, 0, 0, 5, 7, 0, 7, 0, 9, 1, 4, 7, 1, 7, 8, 6, 3, 8, 1, 1, 1, 7, 1, 6, 4, 3, 7, 7, 4, 1, 0, 5, 2, 1, 8, 4, 7, 2, 8, 1, 4, 6, 8, 8, 5, 2, 6, 2, 9, 7, 1, 6, 2, "
+     ]
+    }
+   ],
+   "source": [
+    "B=sc.parallelize(np.random.randint(0,10,size=1000))\n",
+    "lst = B.collect()\n",
+    "for i in lst: \n",
+    "    print(i,end=', ')"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Sequential syntax for chaining\n",
+    "Perform assignment after each computation"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "CPU times: user 15.2 ms, sys: 11.9 ms, total: 27.1 ms\n",
+      "Wall time: 1.01 s\n"
+     ]
+    }
+   ],
+   "source": [
+    "%%time\n",
+    "Squares=B.map(lambda x:x*x)\n",
+    "summation = Squares.reduce(lambda x,y:x+y)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "29395\n"
+     ]
+    }
+   ],
+   "source": [
+    "print(summation)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Cascaded syntax for chaining\n",
+    "Combine computations into a single cascaded command"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "CPU times: user 13.9 ms, sys: 13 ms, total: 26.9 ms\n",
+      "Wall time: 304 ms\n"
+     ]
+    },
+    {
+     "data": {
+      "text/plain": [
+       "29395"
+      ]
+     },
+     "execution_count": 6,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "%%time\n",
+    "B.map(lambda x:x*x).reduce(lambda x,y:x+y)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Both syntaxes mean exactly the same thing\n",
+    "The only difference:\n",
+    "* In the sequential syntax the intermediate RDD has a name `Squares`\n",
+    "* In the cascaded syntax the intermediate RDD is *anonymous*\n",
+    "\n",
+    "The execution is identical!\n",
+    "\n",
+    "### Sequential execution\n",
+    "The standard way that the map and reduce are executed is\n",
+    "* perform the map\n",
+    "* store the resulting RDD in memory\n",
+    "* perform the reduce\n",
+    "\n",
+    "### Disadvantages of Sequential execution\n",
+    "\n",
+    "1. Intermediate result (`Squares`) requires memory space.\n",
+    "2. Two scans of memory (of `B`, then of `Squares`) - double the cache-misses.\n",
+    "\n",
+    "### Pipelined execution\n",
+    "Perform the whole computation in a single pass. For each element of **`B`**\n",
+    "1. Compute the square\n",
+    "2. Enter the square as input to the `reduce` operation.\n",
+    "\n",
+    "### Advantages of Pipelined execution\n",
+    "\n",
+    "1. Less memory required - intermediate result is not stored.\n",
+    "2. Faster - only one pass through the Input RDD."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "sc.stop()"
+   ]
+  }
+ ],
+ "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": "3.6.6"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}