Postgres的未来

下载 0

无羡

发布于

2276

人观看

#信息技术

Ken将分享关于Postgres未来的见解——什么是新的，什么是下一个，以及我们所追求的。他将涉及：
PostgreSQL 12及更高版本的路线图 社区更新
*Postgres在市场上的新用途

展开查看详情

2 . We are the world leader Postgres-based software and services PROVEN COMMITTED GLOBAL • Recognized RDBMS leader • Founded in 2004 • Customer global base > 4000 by Gartner • Largest PostgreSQL contributor— • 300+ Employees world-wide • 2013-2018 Member of 40% of core team • Offices in 16 countries Gartner Magic Quadrant CONFIDENTIAL © Copyright EnterpriseDB Corporation, 2019. All rights reserved.

4 . HIGHLIGHTS OF POSTGRES 12 • Table Access Methods • Storage API • btree Index Improvements • In some cases, btree indexes are up to 40% smaller than in previous releases. • Partitioning Improvements • Foreign keys that reference partitioned tables • Faster run-time pruning • Partition pruning now covers a few more cases • ATTACH PARTITION w/ShareUpdateExclusiveLock • Use Append Rather than MergeAppend for scanning ordered partitions 4 CONFIDENTIAL © Copyright EnterpriseDB Corporation, 2019. All rights reserved.

5 . HIGHLIGHTS OF POSTGRES 12 (MORE) • Removal of recovery.conf • Generated columns (stored, not virtual) • CTE inlining (with manual control) • REINDEX CONCURRENTLY • Unified logging system for command-line programs, including colors. • SQL/JSON: jsonpath • Extended statistics: Multivariate MCV lists • Collations with nondeterministic comparison • GSSAPI encryption support • SERIALIZABLE for parallel query 5 CONFIDENTIAL © Copyright EnterpriseDB Corporation, 2019. All rights reserved.

6 . FUTURE TECHNOLOGY TRENDS Three cool areas of innovation Storage Storage zHeap Engine engine API Data Partitioning Replication Distribution & Sharding Deployment Cloud Containers Options 6 CONFIDENTIAL © Copyright EnterpriseDB Corporation, 2019. All rights reserved.

7 .PLUGGABLE STORAGE To all the MySQL folks… Yeah, I know… • Allow PostgreSQL to support pluggable storage formats • Allows innovation – major changes to the heap are impossible because everyone relies on it. Can’t go backwards for any use case! • Allows for user choice – if there are multiple storage formats available, pick the one that is best for your use case. CONFIDENTIAL © Copyright EnterpriseDB Corporation, 2019. All rights reserved.

8 .PLUGGABLE STORAGE: EXAMPLES • Columnar storage • Most queries don’t need all columns • Write-once read-many (WORM) • No support UPDATE, DELETE, or SELECT FOR UPDATE/SHARE • Index-organized storage • One index is more important than all of the others • In-memory storage • No need to spill to disk • zHeap… CONFIDENTIAL © Copyright EnterpriseDB Corporation, 2019. All rights reserved.

9 . ZHEAP: HIGH-LEVEL BENEFITS • Better Bloat Control • Perform updates “in place” to avoid creating bloat (when possible) • Reuse space right after COMMIT or ABORT – little or no need for VACUUM • Fewer Writes (eliminate write amplification) • Eliminate hint-bits, freezing and anything else that could dirty a page (other than an update) • Allowing in-place updates when index column is updated by providing delete- marking in index • Indexes are not touched if the indexed columns are not changed • Smaller in Size • Narrower tuple headers – most transactional info not stored on the page itself • Eliminate most alignment padding 9 CONFIDENTIAL © Copyright EnterpriseDB Corporation, 2019. All rights reserved.

10 . PARTITIONING IS MORE THAN JUST SHARDING EVOLVING TO SUPPORT MORE AND MORE USE CASES Systems of Record (SoR) Systems of Analysis (SoA) ⚫ Partitioning 54% % Users ⚫ MPP Data warehouses ⚫ Rapidly Advancing in Core ⚫ Greenplum, RedShift… $35B+ Systems of True Total Addressable Engagement (SoE) General Market Purpose “Corner Cases” that require ⚫ Sharding specialized technologies, DBMS e.g. MPP, NoSQL, RAC ⚫ Citus Extension 17% 63% % Users % Users % Percentage of Postgres users who run each data management workload 10 CONFIDENTIAL © Copyright EnterpriseDB Corporation, 2019. All rights reserved.

11 . LOGICAL REPLICATION ENABLES MANY NEW USE CASES Key Capabilities DB DB DB DB DB Server 1 Server 2 Server 3 Server 4 Server 5 ⚫ Table-base, logical replication EPRS1 EPRS2 EPRS3 EPRS EPRS EPRS EPRS EPRS EPRS ⚫ Row level filtering EPRS EPRS EPRS EPRS EPRS EPRS EPRS EPRS EPRS EPRS EPRS EPRS Producers Producers Consumers Consumers Producers Producers Consumers Consumers Producers Producers Consumers Consumers Producers Consumers Producers Consumers Producers Consumers ⚫ Multi-master (MMR) replication for Postgres Kafka Broker Kafka Broker Kafka Broker Leader LeaderTopics LeaderTopics Topics Leader LeaderTopics LeaderTopics Topics Leader LeaderTopics LeaderTopics Topics ⚫ Real time integration & Kafka/ Kafka/ Zookeeper Zookeeper migration from Oracle Follower Leader Topics LeaderTopics Topics Replication Follower Leader Topics LeaderTopics Topics Replication Follower Leader Topics LeaderTopics Topics and MS SQL Server ⚫ Leverages Kafka & Zookeeper Zookeeper Zookeeper Zookeeper for HA & Horizontal Scalability EPRS Controller EPRS Controller EPRS Controller ⚫ Multi-region and geographically dispersed Rest API Rest API databases EPRS Client CLI 11 CONFIDENTIAL CONFIDENTIAL © Copyright © Copyright EnterpriseDB EnterpriseDB Corporation, Corporation, 2019. All 2019. rights Allreserved. rights reserved.

14 . BUT NOT ONLY CLOUD? What is Timeline to Run Databases in Cloud? 35% We asked a 1,000 30% downloaders from 25% postgresql.org…. 20% 15% 33% 29% 10% 17% 5% 12% 9% 0% Never/No More than In 6 to 12 Within 6 Currently plans to 12 months months months running deploy DBs out DBs in the in the cloud cloud 14 CONFIDENTIAL © Copyright EnterpriseDB Corporation, 2019. All rights reserved.

18 . CONTINUALLY IMPROVING PERFORMANCE Global mobile ad network Online Brokerage Firm Global consumer financial Global stock trade • Largest database is • 1 billion writes a day services provider underwriter 14TB • 3,000 transactions • Example application • Largest database • 1.2 billion transactions a per second database is 2TB is 8 TB day, 55K transaction per • 800 concurrent • 200K SELECT statements • 6 to 10 million second users per second transactions per • 400 concurrent users • 25K WRITE transactions day • Analyzes 240TB of data per second per day 18 CONFIDENTIAL © Copyright EnterpriseDB Corporation, 2019. All rights reserved.

0点赞

0收藏

0下载