Your SlideShare is downloading. ×
0
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Spark-summit-2013 Matei Zaharia
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×
Saving this for later? Get the SlideShare app to save on your phone or tablet. Read anywhere, anytime – even offline.
Text the download link to your phone
Standard text messaging rates apply

Spark-summit-2013 Matei Zaharia

436

Published on

Spark Summit 2013 …

Spark Summit 2013
The State of Spark, and Where We’re Going Next
Matei Zaharia (CTO, Databricks; Assistant Professor, MIT)

Summary of recent growth spark project

Published in: Data & Analytics, Technology
0 Comments
2 Likes
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total Views
436
On Slideshare
0
From Embeds
0
Number of Embeds
2
Actions
Shares
0
Downloads
16
Comments
0
Likes
2
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide

Transcript

  • 1. The State of Spark And Where We’re Going Next Matei Zaharia
  • 2. Community Growth
  • 3. Project History Spark started as research project in 2009 Open sourced in 2010 » 1st version was 1600 LOC, could run Wikipedia demo Growing community since Entered Apache Incubator in June 2013
  • 4. Development Community With over 100 developers and 25 companies, one of the most active communities in big data Comparison: Storm (48), Giraph (52), Drill (18), Tez (12) Past 6 months: more active devs than Hadoop MapReduce!
  • 5. Development Community Healthy across the whole ecosystem
  • 6. Release Growth Spark 0.6: - Java API, Maven, standalone mode - 17 contributors Sept ‘13 Feb ‘13 Oct ‘12 Spark 0.7: - Python API, Spark Streaming - 31 contributors Spark 0.8: - YARN, MLlib, monitoring UI - 67 contributors
  • 7. YARN support (Yahoo!) Columnar compression in Shark (Yahoo!) Fair scheduling (Intel) Metrics reporting (Intel, Quantifind) New RDD operators (Bizo, ClearStory) Scala 2.10 support (Imaginea) Some Community Contributions
  • 8. Conferences 0 50 100 150 200 250 300 350 400 450 500 AMP Camp 1 (Aug 2012) AMP Camp 2 (Aug 2013) Spark Summit (Nov 2013) Attendees
  • 9. Projects Built on Spark Spark Spark Streaming (real-time) GraphX (graph) … Shark" (SQL) MLbase (machine learning) BlinkDB
  • 10. What’s Next?
  • 11. Our View While big data tools have advanced a lot, they are still far too difficult to tune and use Goal: design big data systems that are as powerful & seamless as those for small data
  • 12. Current Priorities Standard libraries Deployment Out-of-the-box usability Enterprise use +
  • 13. Standard Libraries While writing K-means in 30 lines is great, it’s even better to call it from a library! Spark’s MLlib and GraphX will be standard libraries supported by core developers » MLlib in Spark 0.8 with 7 algorithms » GraphX coming soon » Both operate directly on RDDs
  • 14. Standard Libraries val  rdd:  RDD[Array[Double]]  =  ...   val  model  =  KMeans.train(rdd,  k  =  10)       val  graph  =  Graph(vertexRDD,  edgeRDD)   val  ranks  =  PageRank.run(graph,  iters  =  10)  
  • 15. Standard Libraries Beyond these libraries, Databricks is investing heavily in higher-level projects Spark Streaming:" easier 24/7 operation and optimizations coming in 0.9 Shark:" calling Spark libs (e.g. MLlib), optimizer, Hive 0.11 & 0.12 Goal: a complete and interoperable stack
  • 16. Deployment Want Spark to easily run anywhere Spark 0.8: much improved YARN, EC2 support Spark 0.8.1: support for YARN 2.2 SIMR: launch Spark in MapReduce clusters as a Hadoop job (no installation needed!) » For experimenting; see talk by Ahir
  • 17. Monitoring and metrics (0.8) Better support for large # of tasks (0.8.1) High availability for standalone mode (0.8.1) External hashing & sorting (0.9) Ease of Use Long-term: remove need to tune beyond defaults
  • 18. Next Releases Spark 0.8.1 (this month) » YARN 2.2, standalone mode HA, optimized shuffle, broadcast & result fetching Spark 0.9 (Jan 2014) » Scala 2.10 support, configuration system, Spark Streaming improvements
  • 19. What Makes Spark Unique?
  • 20. Big Data Systems Today MapReduce Pregel Dremel GraphLab Storm Giraph Drill Tez Impala S4 … Specialized systems (iterative, interactive and" streaming apps) General batch" processing
  • 21. Spark’s Approach Instead of specializing, generalize MapReduce" to support new apps in same engine Two changes (general task DAG & data sharing) are enough to express previous models! Unification has big benefits » For the engine » For users Spark Streaming GraphX … Shark MLbase
  • 22. Code Size 0 20000 40000 60000 80000 100000 120000 140000 Hadoop MapReduce Storm (Streaming) Impala (SQL) Giraph (Graph) Spark non-test, non-example source lines
  • 23. Code Size 0 20000 40000 60000 80000 100000 120000 140000 Hadoop MapReduce Storm (Streaming) Impala (SQL) Giraph (Graph) Spark non-test, non-example source lines Streaming
  • 24. Code Size 0 20000 40000 60000 80000 100000 120000 140000 Hadoop MapReduce Storm (Streaming) Impala (SQL) Giraph (Graph) Spark non-test, non-example source lines Streaming Shark* * also calls into Hive
  • 25. Code Size 0 20000 40000 60000 80000 100000 120000 140000 Hadoop MapReduce Storm (Streaming) Impala (SQL) Giraph (Graph) Spark non-test, non-example source lines Streaming GraphX Shark* * also calls into Hive
  • 26. Performance Hive Impala(disk) Impala(mem) Shark(disk) Shark(mem) 0 5 10 15 20 25 30 35 40 45 ResponseTime(s) SQL Hadoop Giraph GraphX 0 5 10 15 20 25 30 ResponseTime(min) Graph Storm Spark 0 5 10 15 20 25 30 35 Throughput(MB/s/node) Streaming
  • 27. What it Means for Users Separate frameworks: … HDFS read HDFS write ETL HDFS read HDFS write train HDFS read HDFS write query HDFS HDFS read ETL train query Spark: Interactive" analysis
  • 28. Combining Processing Types 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”,  _  +  _)   From Scala:
  • 29. Combining Processing Types GENERATE  KMeans(tweet_locations)  AS  TABLE  tweet_clusters       //  Scala  table  generating  function  (TGF):   object  KMeans  {      @Schema(spec  =  “x  double,  y  double,  cluster  int”)      def  apply(points:  RDD[(Double,  Double)])  =  {          ...      }   }     From SQL (in Shark 0.8.1):
  • 30. Conclusion Next challenge in big data will be complex and low-latency applications Spark offers a unified engine to tackle and combine these apps Best strength is the community: enjoy Spark Summit!
  • 31. Contributors Aaron Davidson Alexander Pivovarov Ali Ghodsi Ameet Talwalkar Andre Shumacher Andrew Ash Andrew Psaltis Andrew Xia Andrey Kouznetsov Andy Feng Andy Konwinski Ankur Dave Antonio Lupher Benjamin Hindman Bill Zhao Charles Reiss Chris Mattmann Christoph Grothaus Christopher Nguyen Chu Tong Cliff Engle Cody Koeninger David McCauley Denny Britz Dmitriy Lyubimov Edison Tung Eric Zhang Erik van Oosten Ethan Jewett Evan Chan Evan Sparks Ewen Cheslack-Postava Fabrizio Milo Fernand Pajot Frank Dai Gavin Li Ginger Smith Giovanni Delussu Grace Huang Haitao Yao Haoyuan Li Harold Lim Harvey Feng Henry Milner Henry Saputra Hiral Patel Holden Karau Ian Buss Imran Rashid Ismael Juma James Phillpotts Jason Dai Jerry Shao Jey Kottalam Joseph E. Gonzalez Josh Rosen Justin Ma Kalpit Shah Karen Feng Karthik Tunga Kay Ousterhout Kody Koeniger Konstantin Boudnik Lee Moon Soo Lian Cheng Liang-Chi Hsieh Marc Mercer Marek Kolodziej Mark Hamstra Matei Zaharia Matthew Taylor Michael Heuer Mike Potts Mikhail Bautin Mingfei Shi Mosharaf Chowdhury Mridul Muralidharan Nathan Howell Neal Wiggins Nick Pentreath Olivier Grisel Patrick Wendell Paul Cavallaro Paul Ruan Peter Sankauskas Pierre Borckmans Prabeesh K. Prashant Sharma Ram Sriharsha Ravi Pandya Ray Racine Reynold Xin Richard Benkovsky Richard McKinley Rohit Rai Roman Tkalenko Ryan LeCompte S. Kumar Sean McNamara Shane Huang Shivaram Venkataraman Stephen Haberman Tathagata Das Thomas Dudziak Thomas Graves Timothy Hunter Tyson Hamilton Vadim Chekan Wu Zeming Xinghao Pan

×