Introduction to Google BigQuery
Download as PPTX, PDF1 like1,003 views
This document provides an introduction to Google BigQuery, a cloud-based data warehouse that allows users to interactively query and analyze massive datasets. It begins with background on big data and technologies like Hadoop, Hive, and Spark. It then explains the differences between row-based and column-based data stores, with BigQuery using a columnar approach. The rest of the document demonstrates BigQuery through an example query on public datasets and provides pricing and resource information.
Ad
More Related Content
What's hot (20)
Similar to Introduction to Google BigQuery (20)
Ad
More from Csaba Toth (17)
Ad
Recently uploaded (20)
Introduction to Google BigQuery
- 1. Cloud Csaba Toth Presented By: Introduction to Google BigQuery
- 2. Our sponsors
- 3. Disclaimer
- 5. Goal • Being able to issue queries • Preferably in an SQL dialect • Over Big Data • As small response time as possible • Preferably interactive web interface (thus no need to install anything)
- 6. Agenda • Big Data • Brief look at Hadoop, HIVE and Spark • Row based data store vs. Column data store • Google BigQuery • Demo
- 7. Big Data Wikipedia: “collection of data sets so large and complex that it becomes difficult to process using on-hand database management tools or traditional data processing applications” Examples: (Wikibon - A Comprehensive List of Big Data Statistics) • 100 Terabytes of data is uploaded to Facebook every day • Facebook Stores, Processes, and Analyzes more than 30 Petabytes of user generated data • Twitter generates 12 Terabytes of data every day • LinkedIn processes and mines Petabytes of user data to power the "People You May Know" feature • YouTube users upload 48 hours of new video content every minute of the day • Decoding of the human genome used to take 10 years. Now it can be done in 7 days
- 8. Little Hadoop history “The Google File System” - October 2003 • http://labs.google.com/papers/gfs.html – describes a scalable, distributed, fault-tolerant file system tailored for data-intensive applications, running on inexpensive commodity hardware, delivers high aggregate performance “MapReduce: Simplified Data Processing on Large Clusters” - April 2004 • http://queue.acm.org/detail.cfm?id=988408 – describes a programming model and an implementation for processing large data sets.
- 9. Hadoop • Hadoop is an open-source software framework that supports data- intensive distributed applications • A Hadoop cluster is composed of a single master node and multiple worker nodes
- 10. Hadoop Has two main services: 1. Storing large amounts of data: HDFS – Hadoop Distributed File System 2. Processing large amounts of data: implementing the MapReduce programming model
- 11. HDFS Name node Metadata Store Data node Data node Data node Node 1 Node 2 Block A Block B Block A Block B Node 3 Block A Block B
- 12. Job / task management Name node Heart beat signals and communication Jobtracker Data node Data node Data node Task- tracker Task- tracker Map 1 Reduce 1 Map 2 Reduce 2 Task- tracker Map 3 Reduce 3 Map-Reduce
- 13. Hadoop vs. RDBMS Hadoop / MapReduce RDBMS Size of data Petabytes Gigabytes Integrity of data Low High (referential, typed) Data schema Dynamic Static Access method Batch Interactive and Batch Scaling Linear Nonlinear (worse than linear) Data structure Unstructured Structured Normalization of data Not Required Required Query Response Time Has latency (due to batch processing) Can be near immediate
- 14. Apache Hive Log Data RDBMS Data Integration LayerFlume Sqoop Storage Layer (HDFS): row and columnar data, file data Computing Layer (MapReduce) Advanced Query Engines (Hive, Pig) Data Mining (Pegasus, Mahout) Index, Searches (Lucene) DB drivers (Hive driver) GUI (web interface, RESTful API, JavaScript) System management Distribution coordination (Zookeeper) JDBC ODBC JS
- 15. Apache Hive UI
- 16. Apache Hive UI
- 17. Beyond Apache Hive Goals: decrease latency • YARN: the “next generation Hadoop”, improves performance in many respects (resource management and allocation, …) • Hadoop distribution specific solution: e.g. Cloudera Impala, MPP SQL Query engine, based on Hadoop
- 18. Apache Spark • Cluster computing framework with multi- stage in-memory primitives • Open Source, originates from Berkeley • In contrast to Hadoop’s two-stage disk- based MapReduce paradigm, multi-stage in-memory primitives can provide up to 100x performance increase • Requires YARN and HDFS
- 19. Spark and Hadoop
- 20. Spark and Hadoop
- 21. Storing data: row stores • Traditional RDBMS and often the document stores are row oriented too • The engine stores and retrieves rows from disk (unless indexes help) • Row is a collection of column cell values together • Rows are materialized on disk
- 22. Row stores Row cells are stored together on disk id scientist death_by movie_name 1 Reinhardt Maximillian The Black Hole 2 Tyrell Roy Batty Blade Runner 3 Hammond Dinosaur Jurassic Park 4 Soong Lore Star Trek: TNG 5 Morbius His mind Forbidden Planet 6 Dyson Skynet Terminator 2: Judgment Day
- 23. Row stores • Not so great for wide rows • If only a small subset of columns queried, reading the entire row wastes IO • (Indexing strategies can help but I don’t have time to cover them)
- 24. Row stores Bad case scenario: • select sum(bigint_column) from table • Million rows in table • Average row length is 1 KiB The select reads one bigint column (8 bytes) • Entire row must be read • Reads ~1 GiB data for ~8MiB of column data
- 25. Column stores • Data is organized by columns instead of rows • Non material world: often not materialized during storage, exists only in memory • Each row still has some sort of “row id”
- 26. Column stores • A row is a collection of column values that are associated with one another • Associated: every row has some type of “row id“ • Can still produce row output (assembling a row maybe complex though – under the hood)
- 27. Column stores Stores each COLUMN on disk id 1 2 3 4 5 6 title Mrs. Doubtfire The Big Lebowski The Fly Steel Magnolias The Birdcage Erin Brokovitch actor Robin Williams Jeff Bridges Jeff Goldblum Dolly Parton Nathan Lane Julia Roberts genre Comedy Comedy Horror Drama Comedy Drama row id = 1 row id = 6 Natural order may be unusual Each column has a file or segment on disk
- 28. Column stores • Column compression can be way more efficient than row based compression (sometimes 10:1 to 30:1 ratio) • Compression: RLE, Integer packing, dictionaries and lookup, other… • Reduces both storage and IO (thus response time)
- 29. Column stores Best case scenario: • select sum(bigint_column) from table • Million rows in table • Average row length is 1 KiB The select reads one bigint column (8 bytes) • Only single column read from disk • Reads ~8 MiB of column data, even less with compression
- 30. Column stores Bad case scenario: select * from long_wide_table where order_line_id = 34653875; • Accessing all columns doesn’t save anything, could be even more expensive than row store • Not ideal for tables with few columns
- 31. Column stores Updating and deleting rows is expensive • Some column stores are append only • Others just strongly discourage writes • Some split storage into row and column areas
- 32. Column / Row stores • RDBMS provide ACID capabilities • Row stores mainly use tree style indexes • B-tree derivative index structure provides very fast binary search as long as it fits into memory • Very large datasets end up unmanageably big indexes • Column stores: bitmap indexing Very expensive to update
- 33. BigQuery history “Dremel: Interactive Analysis of Web-Scale Datasets” – 2010, describes a column store / retrieval system • https://static.googleusercontent.com/media/research.goo gle.com/en//pubs/archive/36632.pdf Presentation with illustration about principles used in Dremel, from Google • http://www.cs.berkeley.edu/~istoica/classes/cs294/11/not es/12-sameer-dremel.pdf
- 34. BigQuery • A service that enables interactive analysis of massively large datasets • Based on Dremel, a scalable, interactive ad hoc query system for analysis of read- only nested data • Working in conjunction with Google Storage • Has a RESTful web service interface
- 35. BigQuery • You can issue SQL queries over big data • Interactive web interface • As small response time as possible • Auto scales under the hood
- 36. BigQuery SaaS (/ PaaS) Interfacing: • REST API • Web console • Command line tools • Language libraries Insert only
- 37. Demo! Wikipedia public dataset Natalities public dataset Names (uploaded) Google Genomics https://cloud.google.com/genomics/ https://cloud.google.com/genomics/v1/public-data https://cloud.google.com/bigquery/web-ui-quickstart https://cloud.google.com/bigquery/query-reference
- 38. Future thoughts How to visualize data • Possibly using Google Charts • BigQuery alongside Google Maps Playing with genomics data – requires some bio-informatics knowledge
- 40. Our sponsors
- 41. Resources • Slides: http://www.slideshare.net/tothc • Contact: http://www.meetup.com/CCalJUG/ • Csaba Toth: Introduction to Hadoop and MapReduce - http://www.slideshare.net/tothc/introduction-to-hadoop- and-map-reduce • Justin Swanhart: Introduction to column stores - http://www.slideshare.net/MySQLGeek/intro-to-column- stores • Jan Steemann: Column-oriented databases - http://www.slideshare.net/arangodb/introduction-to- column-oriented-databases
- 42. Resources • https://anonymousbi.wordpress.com/2012/11/02/hadoo p-beginners-tutorial-on-ubuntu/ • https://www.capgemini.com/blog/capping-it- off/2012/01/what-is-hadoop • http://blog.iquestgroup.com/en/hadoop/#.Vgg2w2sRMeI • https://www.cloudera.com/content/cloudera/en/docume ntation/core/latest/PDF/cloudera-impala.pdf • https://www.keithrozario.com/2012/07/google-bigquery- wikipedia-dataset-malaysia-singapore.html • https://cloud.google.com/bigquery/web-ui-quickstart • https://cloud.google.com/bigquery/query-reference
- 43. Resources • https://github.com/googlegenomics/getting-started-bigquery • https://github.com/googlegenomics/bigquery-examples • https://github.com/googlegenomics/readthedocs
- 44. Pricing https://cloud.google.com/bigquery/pricing Storage: $0.020 per GB/mo. Queries: $5 per TB processed Streaming inserts: $0.01 per 200 MiB (1 KiB rows) Columns are compressed but price is based on the uncompressed size
- 45. Tips • Storage dominates the costs • Plus you need to restrict queries, use as few columns as possible • BigQuery scans the full columns which are involved in the query • Do not select * • LIMIT the result set • Result caching