An Overview of Apache Spark

Ilya GulmanCTO, TwingoJune, 2014

About Twingo

• A Big Data Company

• Established in 2006 by Golan Nahum

• 25 Employees

• Reseller and expert integrator in HP VERTICA

• Reseller and integrator in MAPR

• Reseller and expert integrator in MICROSTRATEGY

• Deep knowedge in Phyton and Linux

• We did more than 20 Big Data successful Projects

• Expertise in SAAS /OEM BIG DATA solutions

Agenda

• What is Spark?

• The Difference with Spark

• SQL on Spark

• Combining the power

• Real-World Use Cases

• Resources

What is Spark?

Fast and general MapReduce-like engine for large-scale data processing

• Fast In memory data storage for very fast interactive queriesUp to 100 times faster then Hadoop

• General - Unified platform that can combine:SQL, Machine Learning , Streaming , Graph & Complex analytics

• Ease of useCan be developed in Java, Scala or Python

• Integrated with Hadoop Can read from HDFS, HBase, Cassandra, and any Hadoop data source.

What is Spark ?

Spark is the Most Active Open Source Project in Big Data

Proj

ect

cont

ribu

tors

in p

ast

year

0

20

40

60

80

100

120

140

GiraphStorm

Tez

The Spark Community

Unified Platform

Shark (SQL)

Spark Streaming (Streaming)

MLlib(Machine learning)

Spark (General execution engine)

GraphX (Graph computation)

Continued innovation bringing new functionality, e.g.:• Java 8 (Closures, Lamba Expressions)• Spark SQL (SQL on Spark, not just Hive)• BlinkDB (Approximate Queries)• SparkR (R wrapper for Spark)

Supported Languages

• Java• Scala• Python• SQL

Data Sources• Local Files

– file:///opt/httpd/logs/access_log• S3• Hadoop Distributed Filesystem

– Regular files, sequence files, any other Hadoop InputFormat

• Hbase • Can also read from any other Hadoop data

source.

The Difference with Spark

Easy and Fast Big Data

• Easy to Develop– Rich APIs in Java, Scala,

Python– Interactive shell

• Fast to Run– General execution graphs– In-memory storage

2-5× less code Up to 10× faster on disk,100× in memory

Resilient Distributed Datasets (RDD)

• Spark revolves around RDDs• Fault-tolerant collection of elements that can be

operated on in parallel– Parallelized Collection: Scala collection which is run in

parallel– Hadoop Dataset: records of files supported by

Hadoop

http://www.cs.berkeley.edu/~matei/papers/2012/nsdi_spark.pdf

RDD Operations• Transformations

– Creation of a new dataset from an existing• map, filter, distinct, union, sample, groupByKey, join, etc…

• Actions– Return a value after running a computation

• collect, count, first, takeSample, foreach, etc…

Check the documentation for a complete listhttp://spark.apache.org/docs/latest/scala-programming-guide.html#rdd-operations

RDD Code Examplefile = spark.textFile("hdfs://...")

errors = file.filter(lambda line: "ERROR" in line)

# Count all the errorserrors.count()

# Count errors mentioning MySQLerrors.filter(lambda line: "MySQL" in line).count()

# Fetch the MySQL errors as an array of stringserrors.filter(lambda line: "MySQL" in line).collect()

RDD Persistence / Caching

• Variety of storage levels– memory_only (default), memory_and_disk, etc…

• API Calls– persist(StorageLevel)– cache() – shorthand for persist(StorageLevel.MEMORY_ONLY)

• Considerations– Read from disk vs. recompute (memory_and_disk)– Total memory storage size (memory_only_ser)– Replicate to second node for faster fault recovery

(memory_only_2)• Think about this option if supporting a web application

http://spark.apache.org/docs/latest/scala-programming-guide.html#rdd-persistence

Cache Scaling Matters

04080 69 58 41 30 12

% of working set in cache

Exec

utio

n ti

me

(s)

RDD Fault RecoveryRDDs track lineage information that can be used to efficiently recompute lost data

msgs = textFile.filter(lambda s: s.startsWith(“ERROR”)) .map(lambda s: s.split(“\t”)[2])

HDFS File Filtered RDD Mapped RDDfilter

(func = startsWith(…))map

(func = split(...))

Interactive Shell• Iterative Development

– Cache those RDDs– Open the shell and ask questions

• We have all wished we could do this with MapReduce– Compile / save your code for scheduled jobs later

• Scala – spark-shell• Python – pyspark

SQL on Spark

Before Spark - Hive

• Puts structure/schema onto HDFS data• Compiles HiveQL queries into MapReduce

jobs• Very popular: 90+% of Facebook Hadoop

jobs generated by Hive• Initially developed by Facebook

But.. Hive is slow

• Takes 20+ seconds even for simple queries

• "A good day is when I can run 6 Hive queries” – @mtraverso

SQL over Spark

Shark – SQL over Spark

• Hive-compatible (HiveQL, UDFs, metadata)– Works in existing Hive warehouses without changing

queries or data!• Augments Hive

– In-memory tables and columnar memory store• Fast execution engine

– Uses Spark as the underlying execution engine– Low-latency, interactive queries– Scale-out and tolerates worker failures

Shark – SQL over Spark

• Hive-compatible (HiveQL, UDFs, metadata)– Works in existing Hive warehouses without changing

queries or data!• Augments Hive

– In-memory tables and columnar memory store• Fast execution engine

– Uses Spark as the underlying execution engine– Low-latency, interactive queries– Scale-out and tolerates worker failures

Machine Learning

Machine Learning - MLlib• K-Means• L1 and L2-regularized Linear Regression• L1 and L2-regularized Logistic Regression• Alternating Least Squares• Naive Bayes• Stochastic Gradient Descent

* As of May 14, 2014** Don’t be surprised if you see the Mahout library converting to Spark soon

Streaming

Comparison to Storm• Higher throughput than Storm

– Spark Streaming: 670k records/sec/node– Storm: 115k records/sec/node– Commercial systems: 100-500k records/sec/node

100 10000

10

20

30 WordCount

Spark

Storm

Record Size (bytes)

Thro

ughp

ut p

er

node

(MB/

s)100 1000

0

20

40

60 Grep

Spark

Storm

Record Size (bytes)

Thro

ughp

ut p

er

node

(MB/

s)

Combining the power

Combining the power• Use Machine Learningg result as table.

GENERATE KMeans(tweet_locations)

SAVE AS TABLE tweet_clusters;

• Combine SQL, ML, and streaming (Scala)val points = sc.runSql[Double, Double](

“select latitude, longitude from historic_tweets”)

val model = KMeans.train(points, 10)

sc.twitterStream(...)

.map(t => (model.closestCenter(t.location), 1))

.reduceByWindow(“5s”, _ + _)

Real-World Use Cases

Spark at Yahoo!

• Fast Machine Learning Personalized news pages

Spark at Yahoo!• Hive on Spark (Shark)

Using existing BI tools to view and query advertising analytic data collected in Hadoop.Any tool that plugs into Hive, like Tableau, automatically works with Shark.

Spark at• One of the largest streaming video companies

on the Internet

• 4+ billion video feeds per month (second only to YouTube)

• CONVIVA uses Spark Streaming to learn network conditions in real time

Resources

Remember• If you want to use a new technology you must learn that

new technology

• For those who have been using Hadoop for a while, at one time you had to learn all about MapReduce and how to manage and tune it

• To get the most out of a new technology you need to learn that technology, this includes tuning– There are switches you can use to optimize your work

Configuration

http://spark.apache.org/docs/latest/

Most Important• Application Configuration

http://spark.apache.org/docs/latest/configuration.html

• Standalone Cluster Configurationhttp://spark.apache.org/docs/latest/spark-standalone.html

• Tuning Guidehttp://spark.apache.org/docs/latest/tuning.html

Resources

• Pig on Spark– http://apache-spark-user-list.1001560.n3.nabble.com/Pig-on-Spark-td2367.html– https://github.com/aniket486/pig– https://github.com/twitter/pig/tree/spork– http://docs.sigmoidanalytics.com/index.php/Setting_up_spork_with_spark_0.8.1– https://github.com/sigmoidanalytics/pig/tree/spork-hadoopasm-fix

• Latest on Spark– http://databricks.com/categories/spark/– http://www.spark-stack.org/

Thank You

www.twingo.co.il

Optional - More Examples

SparkContext sc = new SparkContext(master, appName, [sparkHome], [jars]);JavaRDD<String> file = sc.textFile("hdfs://...");JavaRDD<String> counts = file.flatMap(line -> Arrays.asList(line.split(" ")))

.mapToPair(w -> new Tuple2<String, Integer>(w, 1))

.reduceByKey((x, y) -> x + y);counts.saveAsTextFile("hdfs://...");

val sc = new SparkContext(master, appName, [sparkHome], [jars])val file = sc.textFile("hdfs://...")val counts = file.flatMap(line => line.split(" "))

.map(word => (word, 1))

.reduceByKey(_ + _)counts.saveAsTextFile("hdfs://...")

Word Count

• Java MapReduce (~15 lines of code)• Java Spark (~ 7 lines of code)• Scala and Python (4 lines of code)

– interactive shell: skip line 1 and replace the last line with counts.collect()• Java8 (4 lines of code)

Network Word Count – Streaming

// Create the context with a 1 second batch sizeval ssc = new StreamingContext(args(0), "NetworkWordCount", Seconds(1),System.getenv("SPARK_HOME"), StreamingContext.jarOfClass(this.getClass))

// Create a NetworkInputDStream on target host:port and count the// words in input stream of \n delimited text (eg. generated by 'nc')val lines = ssc.socketTextStream("localhost", 9999, StorageLevel.MEMORY_ONLY_SER)

val words = lines.flatMap(_.split(" "))

val wordCounts = words.map(x => (x, 1)).reduceByKey(_ + _)

wordCounts.print()

ssc.start()

Deploying Spark – Cluster Manager Types• Standalone mode

– Comes bundled (EC2 capable)• YARN• Mesos