Elephant Roads: PostgreSQL Patches and Variants
0 likes1,306 views
This document summarizes the history of PostgreSQL forks and variants over 17 years. It describes how Michael Stonebraker's POSTGRES database led to forks like Illustra and variants like PostgreSQL. It outlines four types of PostgreSQL variants and provides many examples of expired and active forks for various purposes like clustering, data warehousing, and exotic features. It encourages innovating with new PostgreSQL forks.
Elephant Roads: PostgreSQL Patches and Variants
- 1. Elephant Roads 17 Years of Postgres Forks and Variants Josh Berkus October 2009 JDCon West
- 2. NULL = ?
- 4. E.F. Codd Relational Theory The First Day (1970)
- 5. Michael Stonebraker INteractive Graphics REtrieval System
- 6. Michael Stonebraker INGRES The Second Day (1974)
- 7. Michael Stonebraker POSTGRES The Third Day (1986)
- 8. Stonebraker Marches On ... ● Fourth Day: Cohera ● Fifth Day: Streambase ● Sixth Day: Vertica ● Seventh Day: SciDB ● Eighth Day: VoltDB ● …?
- 9. Illustra: The First Fork ● Stonebraker and a team from UC Berkeley forked POSTGRES in 1992 ● added SQL support ● added new object-relational features ● Started a new company called Miró ● Miró became Montage ● Montage became Illustra
- 10. Informix Swallows Illustra ● 1997: Informix buys Illustra ● Illustra features added to Informix OnLine ● and DataBlades ran Illustra ● 2000: Informix Universal Server 9 ● 2001: IBM Eats Informix
- 11. PostgreSQL: The Second Fork ● 1995: a rag-tag band of POSTGRES users and students decided to save the DBMS by taking it off-campus. ● It became: Postgres95 ● 1996: they put it on a public CVS server ● ported it to SQL ● It became: PostgreSQL ● you know the rest from here ...
- 12. Postgres Begat Many More Forks ● Because of the ● Because our code is license clean and easy to ● Because there's a modify history of forks ● Because Postgres is modular and easy to break up or add to ● Because the community is OK with forks and variants
- 13. Why Change PostgreSQL? ● To experiment with new DB technology ● To commercialize it ● To bundle it with useful tools ● To specialize it for specific tasks ● To add features the community doesn't want ● or that aren't ready for OSS yet ● Because you can!
- 14. But Aren't Forks Bad? No.
- 15. Forks Are Good ● Open source means freedom to fork ● If nobody forks a project, then it's not widely used or actively developed. ● Most forks and their owners contribute to core PostgreSQL ● money, code, ideas ● Some forks develop code to be integrated into main Postgres ● best way to try out really challenging ideas
- 16. 4 Types of Postgres Variants ● Forks: incompatible or proprietary major changes to the core code. ● Patches: compatible, open source major changes to the core code. ● Add-Ons: major middleware or plugins which greatly enhance or change Postgres's functionality ● Redistributions: repackaging Postgres under a different name and/or license
- 17. Expired Forks
- 18. Great Bridge PostgreSQL ● What: PostgreSQL plus some advance patches and support. ● Type: Redistribution ● Licensing: BSD ● Versions: Forked from 7.0, merged to 7.1 ● R.I.P.: Great Bridge LLC died in 2001 ● Contributed: lots of code in 1999-2001 ● most of Core worked for Great Bridge
- 19. Red Hat Database ● What: PostgreSQL redistribution with management tools by Red Hat Inc. ● Type: Re-distribution ● Licensing: BSD & GPL ● Versions: 7.2 to 7.3 ● R.I.P.: Abandoned by Red Hat in 2003 ● Contributed: Tom Lane's salary ● plus visual query analyzer to pgAdmin
- 20. NuSphere UltraSQL ● What: Native Windows Port of PostgreSQL ● Type: Fork ● Licensing: Proprietary ● Versions: 7.2 and 7.3 ● R.I.P.: Lost relevance with release of PostgreSQL 8.0 ● still technically available ● Contributed: code to PostgreSQL 8.0
- 21. PowerGres & PowerGres Plus ● What: Native Windows Port of PostgreSQL by SRA Inc., with optional custom storage ● Type: Fork ● Licensing: Proprietary ● Versions: 7.3, maybe 7.4 ● R.I.P.: Lost relevance with release of PostgreSQL 8.0, HA version still around ● Contributed: code to PostgreSQL 8.0
- 22. Pervasive Postgres ● What: Regular PostgreSQL with support. ● Type: Re-distribution ● Licensing: BSD ● Versions: 8.0 ● R.I.P.: Pervasive dropped Postgres in 2006 ● Contributed: money to the community, mainly for events
- 23. Bizgres ● What: PostgreSQL plus BI and large database features. ● Type: Fork ● Licensing: BSD ● Versions: 8.1, 8.2 ● R.I.P.: Project died out in 2008 ● Greenplum stopped merging code ● Contributed: code, money (Greenplum's)
- 24. Core Forks & Redistributions
- 25. Mammoth PostgreSQL ● What: PostgreSQL with Mammoth Replicator ● Owner: CommandPrompt Inc. ● Type: Fork ● Licensing: open source ● Compatibility: High, merges with main code ● Status: current production & development ● Contributes: money & code ● Used to be a proprietary fork
- 26. Postgres Plus ● What: PostgreSQL with open source add-ons ● Owner: EnterpriseDB ● Type: Redistribution ● Licensing: open source (varies) ● Compatibility: High, merges with main code ● Status: current production & development ● Contributes: money & code
- 27. Postgres Plus Advanced Server ● What: PostgreSQL with Oracle Compatibility and proprietary tools ● Owner: EnterpriseDB ● Type: Fork ● Licensing: Proprietary ● Compatibility: High, merges with main code ● Status: current production & development ● Contributes: money & code ● Formerly EnterpriseDB Advanced Server
- 28. Fujitsu Supported PostgreSQL ● What: PostgreSQL with custom storage engine ● Owner: Fujitsu Australia ● Type: Fork ● Licensing: Proprietary ● Compatibility: High with 8.2 ● Status: current production ● Contributes: money, occasionally
- 29. PostgreSQL for Solaris ● What: PostgreSQL packaged for Solaris with tools and support ● Owner: Sun Microsystems ● Type: Redistribution ● Licensing: open source (varies) ● Compatibility: High, 8.3 ● Status: current production ● Contributes: code ● Likely to be terminated when Oracle deal completes
- 30. Exotic Features
- 31. PostGIS ● What: PostgreSQL plus Geographic support ● #1 open source geo database ● Type: Add-On ● Licensing: BSD & GPL ● Compatibility: High, 8.4 ● Status: current production & development ● Contributes: code, users ● Used to be a package under GPL, now an add-on
- 32. Fork: Truviso ● 2000: TelegraphCQ Project started at UC Berkeley ● “Continuous Query” processing for PostgreSQL ● 2006: UCB Profs. launch startup Amalgamated Insight ● commercialize TelegraphCQ ● re-merged with current PostgreSQL code ● 2008: Amalgamated Insight renamed Truviso ● first commercial releases
- 33. Truviso ● What: PostgreSQL plus Streaming Database ● good for high-volume data streams where you don't care about old data ● Type: Fork ● Licensing: Proprietary ● Compatibility: High ● Version: Forked from 7.1, merged to 8.4 ● Status: current production & development ● Contributes: money & code
- 34. SEPostgres ● What: PostgreSQL integrated with SELinux label-based access control ● Type: Patch ● Owner: SEPostgres project / NEC ● Licensing: BSD ● Compatibility: High, 8.4 ● Status: production, working on merge with core ● Contributes: code
- 35. Clustering Forks & Add-Ons
- 36. Postgres-R ● What: PostgreSQL with Group Communication support for clustering. ● Owner: Postgres-R project ● Type: Patch ● Licensing: open source (BSD) ● Compatibility: High, merges ● Status: alpha, current development ● Contributes: code
- 37. pgCluster (and CyberCluster) ● What: PostgreSQL with Statement Replication and controller support for clustering. ● Owner: pgCluster project, Cybertech.AT ● Type: Fork ● Licensing: open source (BSD) ● Compatibility: High, 8.0 ● Status: beta, development halted ● Contributes: not anymore
- 38. pgCluster-II ● What: PostgreSQL with shared memory & storage support for clustering ala RAC ● Owner: pgCluster project, NTT ● Type: Fork ● Licensing: open source (BSD) ● Compatibility: High, 8.3 ● Status: alpha ● Contributes: code
- 39. PostgresForest ● What: PostgreSQL with middleware-based statement replication clustering ● Owner: NTT (I think) ● Type: Add-On ● Licensing: open source ● Compatibility: Not sure ● Status: beta, development halted ● Contributes: code
- 40. pgPool-II ● What: statement-replication and partitioning middleware for PostgreSQL ● Owner: pgPool project, SRA ● Type: Add-On ● Licensing: open source ● Compatibility: High, 8.4 ● Status: beta, active development ● Contributes: code, money (SRA)
- 41. Continuent ● What: statement-replication, mangagement and partitioning middleware for PostgreSQL ● Owner: Continuent corporation ● Type: Add-On ● Licensing: part open source, part proprietary ● Compatibility: High, 8.4 ● Status: old version production, Tungsten RSN ● Contributes: code, money
- 42. GridSQL ● What: statement-replication, mangagement and partitioning middleware for PostgreSQL ● Owner: EnterpriseDB ● Type: Add-On ● Licensing: GPL ● Compatibility: High, 8.4 ● Status: production, active development ● Contributes: code, money ● Used to be “ExtenDB”, integrated into Postgres Plus
- 43. Data Warehousing / BI
- 44. Netezza ● What: Some PostgreSQL code (30-50%) plus proprietary code and hardware package ● Type: Fork ● Owner: Netezza Co. ● Licensing: Proprietary ● Compatibility: Low ● Forked from 7.2, no longer compatible ● Status: current production ● Contributes: nothing
- 45. Greenplum ● What: Mostly PostgreSQL with proprietary MPP plus BI and VLDB tools ● great for huge data mining ● Owner: Greenplum ● Type: Fork ● Licensing: Proprietary ● Compatibility: Medium, 8.2 ● Status: production, active development ● Contributes: code, money ● used to be “Bizgres MPP”
- 46. Vertica ● What: Proprietary code with a few PostgreSQL libraries for multi-TB column store database ● great for OLAP ● Owner: Vertica Co. ● Type: Fork (sort of) ● Licensing: Proprietary ● Compatibility: Low ● Status: production, active development ● Stonebraker project (no contributions)
- 47. Paraccel ● What: Part-Postgres, part-proprietary in- memory clustered pseudo-column store. ● made for “online data analysis” ● Owner: Paraccel Co. ● Type: Fork ● Licensing: Proprietary ● Compatibility: Medium ● Status: production, active development ● Contributions: none to date
- 48. AsterDB ● What: Patches and proprietary middleware for map/reduce queries on many Postgres servers. ● Owner: Aster Data Co. ● including former Continuent developer ● Type: Patch, Add-On ● Licensing: Proprietary ● Compatibility: High, 8.4 ● Status: production, active development ● Contributions: code, money
- 49. Everest ● What: PostgreSQL with proprietary storage, MPP and column store ● scales to 10's of petabytes ● Owner: Yahoo! ● Type: Fork ● Licensing: Not available to public ● Compatibility: N/A ● Status: production ● Contributions: threw a nice party for us
- 50. HadoopDB ● What: Middleware for running map/reduce queries over many Postgres instances. ● brand-new ● Owner: Yale University ● Type: Add-On ● Licensing: unclear ● Compatibility: High, 8.4 ● Status: alpha, active development ● Contributions: none yet
- 51. List of Variants ● Exotic ● Clustering ● BI/DW ● Postgres-R ● Truviso ● Netezza ● pgCluster ● PostGIS ● pgCluster-II ● Greenplum ● PostgresForest ● SEPostgres ● pgPool-II ● Vertica Continuent Core Variants Paraccel ● ● ● ● GridSQL ● AsterDB ● Mammoth ● HadoopDB ● Postgres Plus ● Everest ● P.P. Advanced Server ● Fujistu Supported PG ● PostgreSQL for Solaris
- 52. So, what are you waiting for? Go make your own fork!
- 53. Contact ● Josh Berkus: josh@pgexperts.com ● blog: it.toolbox.com/blogs/database-soup ● www.pgexperts.com/presentations.html ● PostgreSQL: www.postgresql.org ● PostgreSQL Experts Inc.: www.pgexperts.com This talk is copyright 2009 Josh Berkus, and is licensed under the Creative Commons Attribution License. Most images are the property of their respective copyright holders. All images are used here for purposes of satire or reference to trademarked products.