Spark

Caching and Persistence- By default, RDDs are recomputed each time you run an action on them. This can be expensive (in time) if you need to use a dataset more than once. Spark allows you to control what is cached in memory. [code lang=”scala”]val logs: RDD[String] = sc.textFile("/log.txt") val logsWithErrors = logs.filter(_.contains("ERROR”)).persist() val firstnrecords = logsWithErrors.take(10)[/code] Here, we cache logswithErrors in memory. After firstnrecords is computed, Spark will store the contents of firstnrecords for faster access in future operations if we would like to reuse it. [code lang=”scala”]val numErrors = logsWithErrors.count() //faster result[/code] Now, computing the count on logsWithErrors is much faster. There are many ways to c...

Transformations and Actions – Spark defines transformations and actions on RDDs. Transformations – Return new RDDs as results. They are lazy, Their result RDD is not immediately computed. Actions – Compute a result based on an RDD and either returned or saved to an external storage system (e.g., HDFS). They are eager, their result is immediately computed. Laziness/eagerness is how we can limit network communication using the programming model. Example – Consider the following simple example: val input: List[String] = List(“hi”,”this”,”is”,”example”) val words = sc.parallelize(input) val lengths = words.map(_.length) Nothing happened on the cluster at this point, execution of map(a transformation) is deferred. We need t...

RDDs can be created in two ways: 1)Transforming an existing RDD. 2)From a SparkContext or SparkSession object. – Transforming an existing RDD: When map called on List, it returns a new List. Similarly, many higher-order functions defined on RDD returns a new RDD. – From a SparkContext (or SparkSession) object: The SparkContext object (renamed SparkSession) can be thought of as your handle to the Spark cluster. It represents the connection between the Spark cluster and your running application. It defines a handful of methods which can be used to create and populate a new RDD: a)parallelize: convert a local Scala collection to an RDD. ex:- val rdd= sc.parallelize(Seq("1","2","3")) b)textFile: read a text file from HDFS or a local file system and return an RDD of String. ex:-val ...

Spark implements a distributed data parallel model called Resilient Distributed Datasets(RDDs). Given some large dataset that can’t fit into memory on a single node. ->Chunk up the data(Diagrams needs to be added) ->Distribute it over the cluster of machines. ->From there, think of your distributed data like a single collection. RDDs are Spark’s Distributed collections. It seems a lot like immutable sequential or parallel Scala collections. [code]abstract class RDD[T]{ def map[U](f: T => U): RDD[U] = … def flatMap[U](f: T => TraversableOnce[U]): RDD[U] = .. def filter(f; T => Boolean): RDD[T] = … def reduce(f: (T, T) => T): T = … }[/code] Most operations on RDDs, like Scala’s immutable List. and Scala’s parallel collections, ar...

Hadoop/MapReduce- Hadoop is a widely-used large-scale batch data processing framework. It’s an open source implementation of Google’s MapReduce. MapReduce was ground-breaking because it provided: -> simple API (simple map and reduce steps) -> fault tolerance Fault tolerance is what made it possible for Hadoop/MapReduce to scale to 100s or 1000s of nodes at all. Hadoop/MapReduce + Fault Tolerance Why is this important? For 100s or 1000s of old commodity machines. likelihood of at least one node failing is very high midway through a job. Thus, Hadoop/MapReduce’s ability to recover from node failure enabled: -> computations on unthinkably large data sets to succeed to completion. Fault tolerance + simple API At Google, MapReduce made it possible for an average Googl...