Melbourne Spark Meetup Dec 09 2015
4 likes536 views
1. The document discusses techniques for improving Apache Spark performance through mechanical sympathy, which means optimizing for hardware performance by considering factors like CPU cache usage and minimizing random memory access. 2. It provides examples of how to improve sorting, matrix multiplication, and thread synchronization by making them more cache-friendly and reducing cache misses and context switches. 3. The speaker demonstrates performance improvements from these techniques using Linux perf and flame graph profiling tools. Optimizations like Project Tungsten that customize Spark for the hardware are also discussed.
![Power of data. Simplicity of design. Speed of innovation.
IBM Spark
spark.tc
spark.tc
Power of data. Simplicity of design. Speed of innovation.
IBM Spark
CPU Cache Naïve Matrix Multiplication
// Dot product of each row & column vector
for (i <- 0 until numRowA)
for (j <- 0 until numColsB)
for (k <- 0 until numColsA)
res[ i ][ j ] += matA[ i ][ k ] * matB[ k ][ j ];
19
Bad: Row-wise traversal,
not using full CPU cache line,
ineffective pre-fetching](https://image.slidesharecdn.com/melbournesparkmeetupdec092015-151211033419/85/Melbourne-Spark-Meetup-Dec-09-2015-19-320.jpg)
![Power of data. Simplicity of design. Speed of innovation.
IBM Spark
spark.tc
spark.tc
Power of data. Simplicity of design. Speed of innovation.
IBM Spark
CPU Cache Friendly Matrix Multiplication
// Transpose B
for (i <- 0 until numRowsB)
for (j <- 0 until numColsB)
matBT[ i ][ j ] = matB[ j ][ i ];
20
Good: Full CPU Cache Line,
Effective Prefetching
OLD: matB [ k ][ j ];
// Modify algo for Transpose B
for (i <- 0 until numRowsA)
for (j <- 0 until numColsB)
for (k <- 0 until numColsA)
res[ i ][ j ] += matA[ i ][ k ] * matBT[ j ][ k ];](https://image.slidesharecdn.com/melbournesparkmeetupdec092015-151211033419/85/Melbourne-Spark-Meetup-Dec-09-2015-20-320.jpg)
)
…
def increment(leftIncrement: Int, rightIncrement: Int) : Long = {
var originalCounters: Counters = null
var updatedCounters: Counters = null
do {
originalCounters = getCounters()
updatedCounters = new Counters(originalCounters.left+ leftIncrement,
originalCounters.right+ rightIncrement)
}
// Retry lock-free, optimistic compareAndSet() until AtomicRef updates
while !(counters.compareAndSet(originalCounters, updatedCounters))
}
}
25
Lock Free!!](https://image.slidesharecdn.com/melbournesparkmeetupdec092015-151211033419/85/Melbourne-Spark-Meetup-Dec-09-2015-25-320.jpg)
![Power of data. Simplicity of design. Speed of innovation.
IBM Spark
spark.tc
spark.tc
Power of data. Simplicity of design. Speed of innovation.
IBM Spark
Custom Data Structs & Algos: Sorting
o.a.s.util.collection.unsafe.sort.
UnsafeSortDataFormat
UnsafeExternalSorter
UnsafeShuffleWriter
UnsafeInMemorySorter
RecordPointerAndKeyPrefix
33
Ptr
Key-Prefix
2x CPU Cache-line Friendly!
SortDataFormat<RecordPointerAndKeyPrefix, Long[ ]>
Note: Mixing multiple subclasses of SortDataFormat
simultaneously will prevent JIT inlining.
Supports merging compressed records
(if compression CODEC supports it, ie. LZF)
In-place external sorting of spilled BytesToBytes data
AlphaSort-based, 8-byte aligned sort key
In-place sorting of BytesToBytesMap data](https://image.slidesharecdn.com/melbournesparkmeetupdec092015-151211033419/85/Melbourne-Spark-Meetup-Dec-09-2015-33-320.jpg)
![Power of data. Simplicity of design. Speed of innovation.
IBM Spark
spark.tc
spark.tc
Power of data. Simplicity of design. Speed of innovation.
IBM Spark
Pushdowns
aka. Predicate or Filter Pushdowns
Predicate returns true or false for given function
Filters rows deep into the data source
Reduces number of rows returned
Data Source must implement PrunedFilteredScan
def buildScan(requiredColumns: Array[String],
filters: Array[Filter]): RDD[Row]
41](https://image.slidesharecdn.com/melbournesparkmeetupdec092015-151211033419/85/Melbourne-Spark-Meetup-Dec-09-2015-41-320.jpg)
![Power of data. Simplicity of design. Speed of innovation.
IBM Spark
spark.tc
spark.tc
Power of data. Simplicity of design. Speed of innovation.
IBM Spark
Parquet Data Source
Configuration
spark.sql.parquet.filterPushdown=true
spark.sql.parquet.mergeSchema=false (unless your schema is evolving)
spark.sql.parquet.cacheMetadata=true (requires sqlContext.refreshTable())
spark.sql.parquet.compression.codec=[uncompressed,snappy,gzip,lzo]
DataFrames
val gendersDF = sqlContext.read.format("parquet")
.load("file:/root/pipeline/datasets/dating/genders.parquet")
gendersDF.write.format("parquet").partitionBy("gender")
.save("file:/root/pipeline/datasets/dating/genders.parquet")
SQL
CREATE TABLE genders USING parquet
OPTIONS
(path "file:/root/pipeline/datasets/dating/genders.parquet")
52](https://image.slidesharecdn.com/melbournesparkmeetupdec092015-151211033419/85/Melbourne-Spark-Meetup-Dec-09-2015-52-320.jpg)
![Power of data. Simplicity of design. Speed of innovation.
IBM Spark
spark.tc
spark.tc
Power of data. Simplicity of design. Speed of innovation.
IBM Spark
Spark SQL API
Datasets type safe API (similar to RDDs) utilizing Tungsten
val ds = sqlContext.read.text("ratings.csv").as[String]
val df = ds.flatMap(_.split(",")).filter(_ != "").toDF() // RDD API, convert to DF
val agg = df.groupBy($"rating").agg(count("*") as "ct”).orderBy($"ct" desc)
Typed Aggregators used alongside UDFs and UDAFs
val simpleSum = new Aggregator[Int, Int, Int] with Serializable {
def zero: Int = 0
def reduce(b: Int, a: Int) = b + a
def merge(b1: Int, b2: Int) = b1 + b2
def finish(b: Int) = b
}.toColumn
val sum = Seq(1,2,3,4).toDS().select(simpleSum)
Query files directly without registerTempTable()
%sql SELECT * FROM json.`/datasets/movielens/ml-latest/movies.json`
67](https://image.slidesharecdn.com/melbournesparkmeetupdec092015-151211033419/85/Melbourne-Spark-Meetup-Dec-09-2015-67-320.jpg)
Melbourne Spark Meetup Dec 09 2015
- 1. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc After Dark 1.5+End-to-End, Real-time, Advanced Analytics: Reference Big Data Processing Pipeline Melbourne Spark Meetup Dec 09, 2015 Chris Fregly Principal Data Solutions Engineer advancedspark.com
- 2. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Who Am I? 2 Streaming Data Engineer Open Source Committer Data Solutions Engineer Apache Contributor Principal Data Solutions Engineer IBM Technology Center Founder Advanced Apache Meetup Author Advanced . Due 2016
- 3. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Advanced Apache Spark Meetup Meetup Metrics Top 5 Most Active Spark Meetup! 1700+ Members in just 4 mos!! 2100+ Downloads of Docker Image with many Meetup Demos!!! @ advancedspark.com Meetup Goals Code dive deep into Spark and related open source code bases Study integrations with Cassandra, ElasticSearch,Tachyon, S3, BlinkDB, Mesos, YARN, Kafka, R, etc Surface and share patterns and idioms of well-designed, distributed, big data processing systems
- 4. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Upcoming Meetups and Conferences London Spark Meetup (Oct 12th) Scotland Data Science Meetup (Oct 13th) Dublin Spark Meetup (Oct 15th) Barcelona Spark Meetup (Oct 20th) Madrid Big Data Meetup (Oct 22nd) Paris Spark Meetup (Oct 26th) Amsterdam Spark Summit (Oct 27th) Brussels Spark Meetup (Oct 30th) Zurich Big Data Meetup (Nov 2nd) Geneva Spark Meetup (Nov 5th) San Francisco Datapalooza (Nov 10th) San Francisco Advanced Spark (Nov 12th) 4 Oslo Big Data Hadoop Meetup (Nov 19th) Helsinki Spark Meetup (Nov 20th) Stockholm Spark Meetup (Nov 23rd) Copenhagen Spark Meetup (Nov 25th) Budapest Spark Meetup (Nov 26th) Istanbul Spark Meetup (Nov 28th) Singapore Pre-Strata Spark Meetup (Dec 1st) Sydney Spark Meetup (Dec 8th) Melbourne Spark Meetup (Dec 9th) San Francisco Advanced Spark (Dec 10th) San Francisco Univ of San Francisco (Dec 11th) Toronto Spark Meetup (Dec 14th)
- 5. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Topics of This Talk (15-20 mins each) Spark Streaming and Spark ML Kafka, Cassandra, ElasticSearch, Redis, Docker Spark Core Tuning and Profiling Spark SQL Tuning and Customizing 5
- 6. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc Live, Interactive Demo! 6
- 7. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Audience Participation Required! 7 You -> Audience Instructions Navigate to sparkafterdark.com Click 3 actresses & 3 actors Data -> Scientist This is Totally Anonymous!!
- 8. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Topics of This Talk (15-20 mins each) Spark Streaming and Spark ML Kafka, Cassandra, ElasticSearch, Redis, Docker Spark Core Tuning and Profiling Spark SQL Tuning and Customizing 8
- 9. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Mechanical Sympathy “Hardware and software working together.” - Martin Thompson http://mechanical-sympathy.blogspot.com “Whatever your data structure, my array will beat it.” - Scott Meyers Every C++ Book, basically 9
- 10. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Spark and Mechanical Sympathy 10 Project Tungsten (Spark 1.4-1.6+) GraySort Challenge (Spark 1.1-1.2) Minimize Memory & GC Maximize CPU Cache Saturate Network I/O Saturate Disk I/O
- 11. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark CPU Cache Line and Cache Sizes 11 My Laptop My SoftLayer BareMetal aka “LLC”
- 12. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark CPU Cache Sympathy (AlphaSort paper) Key (10 bytes) + Pointer (4 bytes) = 14 bytes 12 Key Ptr Pre-process & Pull Key From Record Ptr Key-Prefix 2x CPU Cache-line Friendly! Key-Prefix (4 bytes) + Pointer (4 bytes) = 8 bytes Key (10 bytes) + Pad (2 bytes) + Pointer (4 bytes) = 16 bytes Ptr Must Dereference to Compare Key Key Pointer (4 bytes) = 4 bytes Key Ptr Pad /Pad CPU Cache-line Friendly! Dereference full key only to resolve prefix duplicates
- 13. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Performance Comparison 13
- 14. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Sequential vs Random Memory Access 14
- 15. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc Demo! Performance Impact of Random Pointer vs. Sequential Key Sort 15
- 16. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Instrumenting and Monitoring CPU Use Linux perf command! 16 http://www.brendangregg.com/blog/2015-11-06/java-mixed-mode-flame-graphs.html
- 17. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Results of Random vs. Sequential Sort 17 Naïve Random Access Pointer Sort Cache Friendly Sequential Key/Pointer Sort Ptr Key Must Dereference to Compare Key Key Ptr Pre-process & Pull Key From Record -35% -90% -68% -26% % Change -55% perf stat –event L1-dcache-load-misses,L1-dcache-prefetch-misses,LLC-load-misses,LLC-prefetch-misses
- 18. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc Demo! Performance Impact of CPU Cache & Prefetch Misses 18
- 19. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark CPU Cache Naïve Matrix Multiplication // Dot product of each row & column vector for (i <- 0 until numRowA) for (j <- 0 until numColsB) for (k <- 0 until numColsA) res[ i ][ j ] += matA[ i ][ k ] * matB[ k ][ j ]; 19 Bad: Row-wise traversal, not using full CPU cache line, ineffective pre-fetching
- 20. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark CPU Cache Friendly Matrix Multiplication // Transpose B for (i <- 0 until numRowsB) for (j <- 0 until numColsB) matBT[ i ][ j ] = matB[ j ][ i ]; 20 Good: Full CPU Cache Line, Effective Prefetching OLD: matB [ k ][ j ]; // Modify algo for Transpose B for (i <- 0 until numRowsA) for (j <- 0 until numColsB) for (k <- 0 until numColsA) res[ i ][ j ] += matA[ i ][ k ] * matBT[ j ][ k ];
- 21. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Results Of Matrix Multiply Challenge Cache-Friendly Matrix Multiply 21 Naïve Matrix Multiply perf stat –event L1-dcache-load-misses,L1-dcache-prefetch-misses,LLC-load-misses, LLC-prefetch-misses,cache-misses,stalled-cycles-frontend -96% -93% -93% -70% -53% % Change -63% +8543%?
- 22. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Thread and Context Switch Sympathy Problem Atomically Increment 2 Counters (each at different increments) by 1000’s of Simultaneous Threads 22 Possible Solutions Synchronized Immutable Synchronized Mutable AtomicReference CAS Volatile??? Context Switches are Expen$ive!! aka “LLC”
- 23. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Synchronized Immutable Counters case class Counters(left: Int, right: Int) object SynchronizedImmutableCounters { var counters = new Counters(0,0) def getCounters(): Counters = { this.synchronized { counters } } def increment(leftIncrement: Int, rightIncrement: Int): Unit = { this.synchronized { counters = new Counters(counters.left + leftIncrement, counters.right + rightIncrement) } } } 23 Locks whole outer object!!
- 24. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Synchronized Mutable Counters class MutableCounters(left: Int, right: Int) { def increment(leftIncrement: Int, rightIncrement: Int): Unit={ this.synchronized {…} } def getCountersTuple(): (Int, Int) = { this.synchronized{ (counters.left, counters.right) } } } object SynchronizedMutableCounters { val counters = new MutableCounters(0,0) … def increment(leftIncrement: Int, rightIncrement: Int): Unit = { counters.increment(leftIncrement, rightIncrement) } } 24 Locks just MutableCounters
- 25. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Lock-Free AtomicReference Counters object LockFreeAtomicReferenceCounters { val counters = new AtomicReference[Counters](new Counters(0,0)) … def increment(leftIncrement: Int, rightIncrement: Int) : Long = { var originalCounters: Counters = null var updatedCounters: Counters = null do { originalCounters = getCounters() updatedCounters = new Counters(originalCounters.left+ leftIncrement, originalCounters.right+ rightIncrement) } // Retry lock-free, optimistic compareAndSet() until AtomicRef updates while !(counters.compareAndSet(originalCounters, updatedCounters)) } } 25 Lock Free!!
- 26. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Lock-Free AtomicLong Counters object LockFreeAtomicLongCounters { // a single Long (64-bit) will maintain 2 separate Ints (32-bits each) val counters = new AtomicLong() … def increment(leftIncrement: Int, rightIncrement: Int): Unit = { var originalCounters = 0L var updatedCounters = 0L do { originalCounters = counters.get() … // Store two 32-bit Int into one 64-bit Long // Use >>> 32 and << 32 to set and retrieve each Int from the Long } // Retry lock-free, optimistic compareAndSet() until AtomicLong updates while !(counters.compareAndSet(originalCounters, updatedCounters)) } } 26 Q: Why not @volatile long? A: The JVM does not guarantee atomic updates of 64-bit longs and doubles Lock Free!!
- 27. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc Demo! Performance Impact of Thread Blocking & Context Switching 27
- 28. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Results of 2 Counter Synchronization Immutable Case Class 28 Lock-Free AtomicLong -64% -46% -17% -33% perf stat –event context-switches,L1-dcache-load-misses,L1-dcache-prefetch-misses, LLC-load-misses, LLC-prefetch-misses,cache-misses,stalled-cycles-frontend % Change -31% -32% -33% -27% case class Counters(left: Int, right: Int) ... this.synchronized { counters = new Counters(counters.left + leftIncrement, counters.right + rightIncrement) } val counters = new AtomicLong() … do { … } while !(counters.compareAndSet(originalCounters, updatedCounters))
- 29. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Profile Visualizations: Flame Graphs 29 Example: Spark Word Count Java Stack Traces are Good! JDK 1.8+ (-XX:-Inline -XX:+PreserveFramePointer) Plateaus are Bad! I/O stalls, Heavy CPU serialization, etc
- 30. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Project Tungsten: CPU and Memory Create Custom Data Structures & Algorithms Operate on serialized and compressed ByteArrays! Minimize Garbage Collection Reuse ByteArrays In-place updates for aggregations Maximize CPU Cache Effectiveness 8-byte alignment AlphaSort-based Key-Prefix Utilize Catalyst Dynamic Code Generation Dynamic optimizations using entire query plan Developer implements genCode() to create Scala source code (String) Project Janino compiles source code into JVM bytecode 30
- 31. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Why is CPU the Bottleneck? CPU for serialization, hashing, & compression Spark 1.2 updates saturated Network, Disk I/O 10x increase in I/O throughput relative to CPU More partition, pruning, and pushdown support Newer columnar file formats help reduce I/O 31
- 32. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Custom Data Structs & Algos: Aggs UnsafeFixedWidthAggregationMap Uses BytesToBytesMap internally In-place updates of serialized aggregation No object creation on hot-path TungstenAggregate & TungstenAggregationIterator Operates directly on serialized, binary UnsafeRow 2 steps to avoid single-key OOMs Hash-based (grouping) agg spills to disk if needed Sort-based agg performs external merge sort on spills 32
- 33. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Custom Data Structs & Algos: Sorting o.a.s.util.collection.unsafe.sort. UnsafeSortDataFormat UnsafeExternalSorter UnsafeShuffleWriter UnsafeInMemorySorter RecordPointerAndKeyPrefix 33 Ptr Key-Prefix 2x CPU Cache-line Friendly! SortDataFormat<RecordPointerAndKeyPrefix, Long[ ]> Note: Mixing multiple subclasses of SortDataFormat simultaneously will prevent JIT inlining. Supports merging compressed records (if compression CODEC supports it, ie. LZF) In-place external sorting of spilled BytesToBytes data AlphaSort-based, 8-byte aligned sort key In-place sorting of BytesToBytesMap data
- 34. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Code Generation Problem Boxing creates excessive objects Expression tree evaluations are costly JVM can’t inline polymorphic impls Lack of polymorphism == poor code design Solution Code generation enables inlining Rewrite and optimize code using overall plan, 8-byte align Defer source code generation to each operator, UDF, UDAF Use Janino to compile generated source code into bytecode (More IDE friendly than Scala quasiquotes) 34
- 35. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Autoscaling Spark Workers (Spark 1.5+) Scaling up is easy J SparkContext.addExecutors() until max is reached Scaling down is hard L SparkContext.removeExecutors() Lose RDD cache inside Executor JVM Must rebuild active RDD partitions in another Executor JVM Uses External Shuffle Service from Spark 1.1-1.2 If Executor JVM dies/restarts, shuffle keeps shufflin’! 35
- 36. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark “Hidden” Spark Submit REST API http://arturmkrtchyan.com/apache-spark-hidden-rest-api Submit Spark Job curl -X POST http://127.0.0.1:6066/v1/submissions/create --header "Content-Type:application/json;charset=UTF-8" --data ’{"action" : "CreateSubmissionRequest”, "mainClass" : "org.apache.spark.examples.SparkPi”, "sparkProperties" : { "spark.jars" : "file:/spark/lib/spark-examples-1.5.1.jar", "spark.app.name" : "SparkPi",… }}’ Get Spark Job Status curl http://127.0.0.1:6066/v1/submissions/status/<job-id-from-submit-request> Kill Spark Job curl -X POST http://127.0.0.1:6066/v1/submissions/kill/<job-id-from-submit-request> 36 (the snitch)
- 37. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Outline Spark Streaming and Spark ML Kafka, Cassandra, ElasticSearch, Redis, Docker Spark Core Tuning and Profiling Spark SQL Tuning and Customizing 37
- 38. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Parquet Columnar File Format Based on Google Dremel paper ~2010 Collaboration with Twitter and Cloudera Columnar storage format for fast columnar aggs Supports evolving schema Supports pushdowns Support nested partitions Tight compression Min/max heuristics enable file and chunk skipping 38 Min/Max Heuristics For Chunk Skipping
- 39. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Partitions Partition Based on Data Access Patterns /genders.parquet/gender=M/… /gender=F/… <-- Use Case: Access Users by Gender /gender=U/… Dynamic Partition Creation (Write) Dynamically create partitions on write based on column (ie. Gender) SQL: INSERT TABLE genders PARTITION (gender) SELECT … DF: gendersDF.write.format("parquet").partitionBy("gender") .save("/genders.parquet") Partition Discovery (Read) Dynamically infer partitions on read based on paths (ie. /gender=F/…) SQL: SELECT id FROM genders WHERE gender=F DF: gendersDF.read.format("parquet").load("/genders.parquet/").select($"id"). .where("gender=F") 39
- 40. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Pruning Partition Pruning Filter out rows by partition SELECT id, gender FROM genders WHERE gender = ‘F’ Column Pruning Filter out columns by column filter Extremely useful for columnar storage formats (ie. Parquet) Skip entire blocks of columns SELECT id, gender FROM genders 40
- 41. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Pushdowns aka. Predicate or Filter Pushdowns Predicate returns true or false for given function Filters rows deep into the data source Reduces number of rows returned Data Source must implement PrunedFilteredScan def buildScan(requiredColumns: Array[String], filters: Array[Filter]): RDD[Row] 41
- 42. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc Demo! Compare Performance of File Formats, Partitions, and Pushdowns 42
- 43. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Filter Pushdown and Collapsing 43 Filter pushdown (Parquet) Filter collapsing (next best thing)
- 44. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Join Between Partitioned & Unpartitioned 44
- 45. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Join Between Partitioned & Partitioned 45
- 46. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark 46 Cartesian Join vs. Inner Join
- 47. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark 47 Broadcast Join vs. Normal Shuffle Join
- 48. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Visualizing the Query Plan 48 Effectiveness of Filter Cost-based Join Optimization Similar to MapReduce Map-side Join & DistributedCache Peak Memory for Joins and Aggs UnsafeFixedWidthAggregationMap getPeakMemoryUsedBytes()
- 49. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Data Source API Relations (o.a.s.sql.sources.interfaces.scala) BaseRelation (abstract class): Provides schema of data TableScan (impl): Read all data from source PrunedFilteredScan (impl): Column pruning & predicate pushdowns InsertableRelation (impl): Insert/overwrite data based on SaveMode RelationProvider (trait/interface): Handle options, BaseRelation factory Filters (o.a.s.sql.sources.filters.scala) Filter (abstract class): Handles all filters supported by this source EqualTo (impl) GreaterThan (impl) StringStartsWith (impl) 49
- 50. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Native Spark SQL Data Sources 50
- 51. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark JSON Data Source DataFrame val ratingsDF = sqlContext.read.format("json") .load("file:/root/pipeline/datasets/dating/ratings.json.bz2") -- or – val ratingsDF = sqlContext.read.json ("file:/root/pipeline/datasets/dating/ratings.json.bz2") SQL Code CREATE TABLE genders USING json OPTIONS (path "file:/root/pipeline/datasets/dating/genders.json.bz2") 51 json() convenience method
- 52. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Parquet Data Source Configuration spark.sql.parquet.filterPushdown=true spark.sql.parquet.mergeSchema=false (unless your schema is evolving) spark.sql.parquet.cacheMetadata=true (requires sqlContext.refreshTable()) spark.sql.parquet.compression.codec=[uncompressed,snappy,gzip,lzo] DataFrames val gendersDF = sqlContext.read.format("parquet") .load("file:/root/pipeline/datasets/dating/genders.parquet") gendersDF.write.format("parquet").partitionBy("gender") .save("file:/root/pipeline/datasets/dating/genders.parquet") SQL CREATE TABLE genders USING parquet OPTIONS (path "file:/root/pipeline/datasets/dating/genders.parquet") 52
- 53. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark ElasticSearch Data Source Github https://github.com/elastic/elasticsearch-hadoop Maven org.elasticsearch:elasticsearch-spark_2.10:2.1.0 Code val esConfig = Map("pushdown" -> "true", "es.nodes" -> "<hostname>", "es.port" -> "<port>") df.write.format("org.elasticsearch.spark.sql”).mode(SaveMode.Overwrite) .options(esConfig).save("<index>/<document-type>") 53
- 54. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Cassandra Data Source (DataStax) Github https://github.com/datastax/spark-cassandra-connector Maven com.datastax.spark:spark-cassandra-connector_2.10:1.5.0-M1 Code ratingsDF.write .format("org.apache.spark.sql.cassandra") .mode(SaveMode.Append) .options(Map("keyspace"->"<keyspace>", "table"->"<table>")).save(…) 54
- 55. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Mythical New Cassandra Data Source By-pass Cassandra CQL “front door” (which is optimized for transactional data) Bulk read and write directly against SSTables (Similar to 5 year old NetflixOSS project Aegisthus) Promotes Cassandra to first-class analytics option Replicated analytics cluster not needed 55
- 56. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark DB2 and BigSQL Data Sources (IBM) Coming Soon! 56
- 57. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Creating a Custom Data Source ① Study existing implementations o.a.s.sql.execution.datasources.jdbc.JDBCRelation ② Extend base traits & implement required methods o.a.s.sql.sources.{BaseRelation,PrunedFilterScan} Spark JDBC (o.a.s.sql.execution.datasources.jdbc) class JDBCRelation extends BaseRelation with PrunedFilteredScan with InsertableRelation DataStax Cassandra (o.a.s.sql.cassandra) class CassandraSourceRelation extends BaseRelation with PrunedFilteredScan with InsertableRelation! 57
- 58. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc Demo! Create a Custom Data Source 58
- 59. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Publishing Custom Data Sources 59 spark-packages.org
- 60. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc IBM | spark.tc Spark SQL UDF Code Generation 100+ UDFs now generating code More to come in Spark 1.6+ Details in SPARK-8159, SPARK-9571 Every UDF must use Expressions and implement Expression.genCode() to participate in the fun Lambdas (RDD or Dataset API) and sqlContext.udf.registerFunction() are not enough!!
- 61. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Creating a Custom UDF with Code Gen ① Study existing implementations o.a.s.sql.catalyst.expressions.Substring ② Extend and implement base trait o.a.s.sql.catalyst.expressions.Expression.genCode ③ Don’t forget about Python! python.pyspark.sql.functions.py 61
- 62. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc Demo! Creating a Custom UDF participating in Code Generation 62
- 63. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc Spark 1.6 Improvements 63
- 64. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Adaptive Query Execution Adapt query execution using data from previous stages Dynamically choose spark.sql.shuffle.partitions (default 200) 64 Broadcast Join (popular keys) Shuffle Join (not-so-popular keys) Hybrid Join
- 65. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Adaptive Memory Management Spark <1.6 Manual configure between 2 memory regions Spark execution engine (shuffles, joins, sorts, aggs) spark.shuffle.memoryFraction RDD Data Cache spark.storage.memoryFraction Spark 1.6+ Unified memory regions Dynamically expand/contract memory regions Supports minimum for RDD storage (LRU Cache) 65
- 66. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Metrics Shows exact memory usage per operator & per node Helps debugging and identifying skew 66
- 67. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Spark SQL API Datasets type safe API (similar to RDDs) utilizing Tungsten val ds = sqlContext.read.text("ratings.csv").as[String] val df = ds.flatMap(_.split(",")).filter(_ != "").toDF() // RDD API, convert to DF val agg = df.groupBy($"rating").agg(count("*") as "ct”).orderBy($"ct" desc) Typed Aggregators used alongside UDFs and UDAFs val simpleSum = new Aggregator[Int, Int, Int] with Serializable { def zero: Int = 0 def reduce(b: Int, a: Int) = b + a def merge(b1: Int, b2: Int) = b1 + b2 def finish(b: Int) = b }.toColumn val sum = Seq(1,2,3,4).toDS().select(simpleSum) Query files directly without registerTempTable() %sql SELECT * FROM json.`/datasets/movielens/ml-latest/movies.json` 67
- 68. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Spark Streaming State Management New trackStateByKey() Store deltas, compaction later No longer scan every key Sessionization TTL Integrated A/B Testing Shows failure output in Admin UI 68
- 69. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Thank You!!! Chris Fregly IBM Spark Tech Center (http://spark.tc) San Francisco, California, USA advancedspark.com Sign up for the Meetup and Book Clone, Contribute, Commit on Github Run All Demos in Docker 2100+ Docker Downloads!! Find me on LinkedIn, Twitter, Github, Email, Fax 69
- 70. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc What’s Next? 70 After Dark 1.6+
- 71. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark What’s Next? Autoscaling Docker/Spark Workers Completely Docker-based Docker Compose, Google Kubernetes Lots of Demos and Examples More Zeppelin & IPython/Jupyter notebooks More advanced analytics use cases More Performance Profiling Work closely with Netflix & Databricks Identify & fix performance bottlenecks 71
- 72. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark Upcoming Meetups and Conferences London Spark Meetup (Oct 12th) Scotland Data Science Meetup (Oct 13th) Dublin Spark Meetup (Oct 15th) Barcelona Spark Meetup (Oct 20th) Madrid Big Data Meetup (Oct 22nd) Paris Spark Meetup (Oct 26th) Amsterdam Spark Summit (Oct 27th) Brussels Spark Meetup (Oct 30th) Zurich Big Data Meetup (Nov 2nd) Geneva Spark Meetup (Nov 5th) San Francisco Datapalooza (Nov 10th) San Francisco Advanced Spark (Nov 12th) 72 Oslo Big Data Hadoop Meetup (Nov 19th) Helsinki Spark Meetup (Nov 20th) Stockholm Spark Meetup (Nov 23rd) Copenhagen Spark Meetup (Nov 25th) Budapest Spark Meetup (Nov 26th) Istanbul Spark Meetup (Nov 28th) Singapore Pre-Strata Spark Meetup (Dec 1st) Sydney Spark Meetup (Dec 8th) Melbourne Spark Meetup (Dec 9th) San Francisco Advanced Spark (Dec 10th) San Francisco Univ of San Francisco (Dec 11th) Toronto Spark Meetup (Dec 14th)
- 73. Power of data. Simplicity of design. Speed of innovation. IBM Spark spark.tc Power of data. Simplicity of design. Speed of innovation. IBM Spark advancedspark.com @cfregly