Differences
This shows you the differences between two versions of the page.
Both sides previous revision Previous revision Next revision | Previous revision Next revision Both sides next revision | ||
spark:spark-introduction [2014/10/03 12:56] straka |
spark:spark-introduction [2014/11/11 09:06] straka |
||
---|---|---|---|
Line 5: | Line 5: | ||
===== Running Spark Shell in Python ===== | ===== Running Spark Shell in Python ===== | ||
- | To run interactive Python shell in local Spark mode, run (on your local workstation or on cluster) | + | To run interactive Python shell in local Spark mode, run (on your local workstation or on cluster |
- | | + | |
- | The IPYSPARK=1 parameter instructs Spark to use '' | + | The IPYTHON=1 parameter instructs Spark to use '' |
- | After a local Spark executor is started, the Python shell starts. | + | After a local Spark executor is started, the Python shell starts. |
+ | the prompt line, the SparkUI address is listed in the following format: | ||
14/10/03 10:54:35 INFO SparkUI: Started SparkUI at http:// | 14/10/03 10:54:35 INFO SparkUI: Started SparkUI at http:// | ||
+ | The SparkUI is an HTML interface which displays the state of the application -- if a distributed computation is taking place, how many workers are part of it, how many tasks are left to be processed, any error logs, also cached datasets and their properties (cached on disk / memory, their size) are displayed. | ||
==== Running Spark Shell in Scala ==== | ==== Running Spark Shell in Scala ==== | ||
To run interactive Scala shell in local Spark mode, run (on your local workstation or on cluster) | To run interactive Scala shell in local Spark mode, run (on your local workstation or on cluster) | ||
- | | + | |
Once again, the SparkUI address is listed several lines above the shell prompt line. | Once again, the SparkUI address is listed several lines above the shell prompt line. | ||
===== Word Count Example ===== | ===== Word Count Example ===== | ||
- | The central object of Spark is RDD -- resilient distributed dataset. It contains ordered sequence of items. | + | |
+ | The central object of Spark framework | ||
+ | |||
+ | We start by simple word count example. We load the RDD from text file, every line of the input file becoming an element of RDD. We then split every line into words, count every word occurrence and sort the words by the occurrences. Try the following in the opened Python shell: | ||
+ | <file python> | ||
+ | wiki = sc.textFile("/ | ||
+ | words = wiki.flatMap(lambda line: line.split()) | ||
+ | counts = words.map(lambda word: (word, 1)).reduceByKey(lambda c1,c2: c1+c2) | ||
+ | sorted = counts.sortBy(lambda (word, | ||
+ | sorted.saveAsTextFile(' | ||
+ | |||
+ | # Alternatively, | ||
+ | (sc.textFile("/ | ||
+ | | ||
+ | | ||
+ | | ||
+ | | ||
+ | | ||
+ | </ | ||
+ | The output of ' | ||
+ | |||
+ | Note that ' | ||
+ | |||
+ | The Scala versions is quite similar: | ||
+ | <file scala> | ||
+ | val wiki = sc.textFile("/ | ||
+ | val words = wiki.flatMap(line => line.split(" | ||
+ | val counts = words.map(word => (word, | ||
+ | val sorted = counts.sortBy({case (word, count) => count}, ascending=false) | ||
+ | sorted.saveAsTextFile(' | ||
+ | |||
+ | // Alternatively without variables and using placeholders in lambda parameters: | ||
+ | (sc.textFile("/ | ||
+ | | ||
+ | | ||
+ | | ||
+ | | ||
+ | </ | ||
+ | |||
+ | |||
+ | ===== K-Means Example ===== | ||
+ | An example implementing [[http:// | ||
+ | <file python> | ||
+ | import numpy as np | ||
+ | |||
+ | def closestPoint(point, | ||
+ | return min((np.sum((point - centers[i]) ** 2), i) for i in range(len(centers)))[1] | ||
+ | |||
+ | lines = sc.textFile("/ | ||
+ | data = lines.map(lambda line: map(float, line.split())).cache() | ||
+ | |||
+ | K = 50 | ||
+ | epsilon = 1e-3 | ||
+ | |||
+ | centers = data.takeSample(False, | ||
+ | for i in range(5): | ||
+ | old_centers = sc.broadcast(centers) | ||
+ | centers = (data | ||
+ | # For each point, find its closest center index. | ||
+ | | ||
+ | # Sum points and counts in each cluster. | ||
+ | | ||
+ | # Sort by cluster index. | ||
+ | | ||
+ | # Compute the new centers by averaging points in clusters. | ||
+ | | ||
+ | | ||
+ | # If the change in center positions is less than epsilon, stop. | ||
+ | centers_change = sum(np.sqrt(np.sum((a - b)**2)) for (a, b) in zip(centers, | ||
+ | old_centers.unpersist() | ||
+ | if centers_change < epsilon: | ||
+ | break | ||
+ | |||
+ | print "Final centers: " + str(centers) | ||
+ | </ | ||
+ | The implementation starts by loading the data and caching them in memory using '' | ||
+ | |||
+ | Note that explicit broadcasting used for '' | ||
+ | |||
+ | For illustration, | ||
+ | <file scala> | ||
+ | import breeze.linalg.Vector | ||
+ | |||
+ | type Vector = breeze.linalg.Vector[Double] | ||
+ | type Vectors = Array[Vector] | ||
+ | |||
+ | def closestPoint(point : Vector, centers : Vectors) : Double = | ||
+ | centers.map(center => (center-point).norm(2)).zipWithIndex.min._2 | ||
+ | |||
+ | val lines = sc.textFile("/ | ||
+ | val data = lines.map(line => Vector(line.split(" | ||
+ | |||
+ | val K = 50 | ||
+ | val epsilon = 1e-3 | ||
+ | |||
+ | var i = 0 | ||
+ | var centers_change = Double.PositiveInfinity | ||
+ | var centers = data.takeSample(false, | ||
+ | while (i < 10 && centers_change > epsilon) { | ||
+ | val old_centers = sc.broadcast(centers) | ||
+ | centers = (data | ||
+ | // For each point, find its closes center index. | ||
+ | | ||
+ | // Sum points and counts in each cluster. | ||
+ | | ||
+ | // Sort by cluster index. | ||
+ | | ||
+ | // Compute the new centers by averaging corresponding points. | ||
+ | | ||
+ | | ||
+ | |||
+ | // Compute change in center positions. | ||
+ | centers_change = (centers zip old_centers.value).map({case (a,b) => (a-b).norm(2)}).sum | ||
+ | old_centers.unpersist() | ||
+ | i += 1 | ||
+ | } | ||
+ | |||
+ | print(centers.deep) | ||
+ | </ |