PostgreSQL 13 is Coming - Find Out What's New!

New and Improved
Features in PostgreSQL 13
Rushabh Lathia
Hosted by: Violet Seah

This session is being recorded.
The slides and recording will be available after the session.
Please submit questions via GotoWebinar question box– all questions will be
answered after the presentation.
We will be sharing info about EDB and Postgres later
Welcome – Housekeeping Items

© Copyright EnterpriseDB Corporation, 2020. All rights reserved.3
Agenda
●PostgreSQL 13 - New Features
●Postgres Advanced Server partitioning
●Postgres Advanced Server 13 - New Features
●Q & A session

PostgreSQL v13
Key New
Features

Speed: A key part of Postgres’ evolution
January 2020: 50% TPS improvement in four years
https://www.enterprisedb.com/postgres-tutorials/benchmarking-postgresql-aws-m5metal-instance

• Select list of new and enhanced capabilities
• Vacuum
• Security and consistency
• Partition-wise join
• For a complete list, please check out the release notes
https://www.postgresql.org/docs/release/13.0/
Speed is one part of the equation
Key new capabilities in PostgreSQL

• Performance for parallel vacuum of indexes
• Vacuum performance after executing 50 million in-
place updates - 4X faster in multi process benchmark
• Auto Vacuum for append-only transactions
• Get few giant vacuum
• Enables fast Index Only Scan
• Important for IOT tables
Vacuum Improvements
Parallel vacuum of indexes and vacuum for append-only tables
https://www.enterprisedb.com/postgres-tutorials/what-parallel-vacuum-postgresql-13

• libpq channel binding
• Stop man in the middle attacks
• Only for SCRAM authentication
• New capability for pg_catcheck
• tool for diagnosing system catalog corruption
• Find it https://github.com/EnterpriseDB/pg_catcheck
• New capability: check if the initial file is available for every relation (table)
• Address ‘could not open file issue’
Security and Consistency
libpq channel binding and improvements of pg_catcheck

• New option --select-from-relations.
• This option won't tell you the reason why you are getting such errors.
• Works for table, TOAST table, and materialized view in the database.
• This option still doesn't help with indexes, or relation segments other than the first one.
• >>> The relation is accessible <<< “Could not open file”.
• Supports EDB Postgres Advanced Server and PostgreSQL.
• Example:
New capability for pg_catchack
Find “could not open file xxx”
rushabh@rushabh:pg_catcheck$ ./pg_catcheck edb --select-from-relations
notice: unable to query relation "public"."emp": ERROR: could not open
file "base/16198/16394": Permission denied
notice: unable to query relation "public"."jobhist": ERROR: could not
open file "base/16198/16405": No such file or directory
progress: done (2 inconsistencies, 0 warnings, 0 errors)

• PG 11, introduced partition wise join.
• SET enable_partitionwise_join. Default is off.
• Join should be on partition key and both partitions should have same bounds for partitions
• PG 13, enhance the same by removing restriction of having same bounds for partitions.
• For more details about the Declarative Partition Enhancements, the optimizations, and
implementation journey, in community, watch my colleague Amit Langote’s talk “Table
Partitioning in Postgres, How Far We’ve Come” at Postgres Vision 2020
Partition Wise Join
Significant Enhancement of existing feature

Logical replication for partitioned tables
• PG13, one can create a PUBLICATION on partitioned table.
• Publication parameter publish_via_partition_root GUC
• If set to TRUE, then it will look like all the changes coming from partition root.
• If set to FALSE, then changes will look like they are coming from individual partitions.
• With this you can replicate from non-partitioned to partitioned table or partition to non-partition,
or different structure and different partition bounds.

• It’s part of pg_backbackup
• Default is ON.
• Generate the manifests file, with list of backfile, checksum, etc.
• New tool
• pg_verifybackup - read the manifests file and verify the backup.
Backup manifests

PostgreSQL and EDB Postgres
Advanced Server
Superset of PostgreSQL
All PostgreSQL features are available in
EDB Postgres Advanced Server
Managed fork, continuously
synchronized with PostgreSQL
© Copyright EnterpriseDB Corporation, 2020 All rights reserved.
When would you pick EDB
Postgres Advanced Server?
Native PL/SQL compatibility, key
Oracle packages, pragma
autonomous transaction, query
hints, etc.
Resource Manager manages CPU and
I/O resources
Session/System Wait Diagnostics
EDB Loader for fast bulk loads
Enhanced security features
○ Separate audit log
○ Native data redaction
○ Password policy management

Deep dive into Postgres Advanced Server Partitioning
●EDB Partitioning History
●Automatic Partition
●Interval Partition
●Automatic Hash Partition

• PostgreSQL introduced declarative partitioning in v10
• Getting new features for partitioning with every new release.
• Performance improvements.
• EPAS had Partitioning feature since 9.4, which was inheritance based with Redwood Syntex.
• From v10, EPAS moved it’s inheritance based partitioning implementation to leverage PostgreSQL
declarative partitioning.
• EPAS adding some cool partitioning features to make a life easy.
• Redwood compatible syntax for partitioning.
• Exchange partition.
• Split Partition.
• etc..
EDB Postgres Partitioning

• Easy way to manage the partition scheme.
• Don’t need to worry about partition creation at runtime.
• Easy LOAD or COPY data from old non-partition to partition table.
• Below are the type of automatic partitioning:
• INTERVAL Partition (RANGE).
• AUTOMATIC Partition (LIST).
• Provide PARTITION and SUB-PARTITION number (HASH).
Partition Automation
Automatically create new partitions when needed w.o. locking

• INTERVAL partition is an extension to RANGE partition.
• Need to provide an INTERVAL expression for the partition.
• Create a new partition automatically, if given tuple doesn’t fit to the existing partitions.
• INTERVAL clause will decide the range size for a new partition.
• Can also ALTER the existing partition to convert into INTERVAL partition.
Interval Partition

Interval Partition
Example (1 of 2)
edb@61349=#CREATE TABLE orders (id int, country_code varchar(5), order_date DATE )
PARTITION BY RANGE (order_date) INTERVAL (NUMTOYMINTERVAL(1,'MONTH'))
(PARTITION P1 values LESS THAN (TO_DATE('01-MAY-2020','DD-MON-YYYY')));
CREATE TABLE
edb@61349=#INSERT INTO orders VALUES (1, 'IND', '24-FEB-2020');
INSERT 0 1
edb@61349=#SELECT tableoid::regclass, * FROM orders;
tableoid | id | country_code | order_date
-----------+----+--------------+--------------------
orders_p1 | 1 | IND | 24-FEB-20 00:00:00
(1 row)

Interval Partition
Example (2 of 2)
edb@61349=#INSERT INTO orders VALUES (1, 'IND', '23-JUN-2020');
INSERT 0 1
tableoid | id | country_code | order_date
---------------------+----+--------------+--------------------
orders_p1 | 1 | IND | 24-FEB-20 00:00:00
orders_sys613490102 | 1 | IND | 23-JUN-20 00:00:00
(2 rows)
Other syntax options:
ALTER TABLE orders SET INTERVAL();
ALTER TABLE orders SET INTERVAL(NUMTOYMINTERVAL(1,'MONTH'));

• Automatic list partitioning creates a partition for any new distinct value of the list partitioning
key.
• An AUTOMATIC partition is an extension to list partition where system automatically create the
partition if the new tuple doesn't fit into existing partitions.
• We can also enable automatic list partitioning on the existing partition table using the ALTER
TABLE command.
• ALTER TABLE <tablename> SET [MANUAL|AUTOMATIC]
Automatic Partition
Automatically create a new partition for a LIST partition

Automatic Partition
edb@61349=#CREATE TABLE orders(id int, country_code varchar(5))PARTITION BY
LIST (country_code) AUTOMATIC
(PARTITION p1 values ('IND'), PARTITION p2 values ('USA'));
CREATE TABLE
edb@61349=#INSERT INTO orders VALUES (1, 'IND');
INSERT 0 1
edb@61349=#INSERT INTO orders VALUES (2, 'USA');
INSERT 0 1
tableoid | id | country_code
-----------+----+--------------
orders_p1 | 1 | IND
orders_p2 | 2 | USA
(2 rows)

Automatic Partition
edb@129883=#INSERT INTO orders VALUES ( 1 , 'UK');
INSERT 0 1
tableoid | id | country_code
----------------------+----+--------------
orders_p1 | 1 | IND
orders_sys1298830103 | 1 | UK
orders_p2 | 2 | USA
(3 rows)
Other syntax options:
ALTER TABLE orders SET MANUAL;
ALTER TABLE orders SET AUTOMATIC;

Automatic Hash Partitioning
Part 1
● "PARTITIONS <num> STORE IN clause"
● allows us to automatically add a specified number of HASH partitions during CREATE TABLE command.
● The STORE IN clause is used to specify the tablespaces across which the partitions should be distributed.
Example:
edb@72970=#CREATE TABLE hash_tab (
col1 NUMBER,
col2 NUMBER)
PARTITION BY HASH (col1, col2)
PARTITIONS 2 STORE IN (tablespace1, tablespace2);
edb@72970=#d hash_tab
Partitioned table "public.hash_tab"
Column | Type | Collation | Nullable | Default
--------+---------+-----------+----------+---------
col1 | numeric | | |
col2 | numeric | | |
Partition key: HASH (col1, col2)
Number of partitions: 2 (Use d+ to list them.)

Part 2
● The SUBPARTITIONS <num> creates a predefined template to auto create a specified number of subpartitions for each partition.
CREATE TABLE ORDERS (id int, country_code varchar(5), order_date DATE)
PARTITION BY LIST (country_code) AUTOMATIC
SUBPARTITION BY HASH(order_date) SUBPARTITIONS 3
( PARTITION p1 VALUES ('USA', 'IND') );
edb@89398=#SELECT table_name, partition_name, subpartition_name FROM user_tab_subpartitions WHERE
table_name = ‘ORDERS’;
table_name | partition_name | subpartition_name
------------+----------------+-------------------
ORDERS | P1 | SYS0101
(3 rows)

Part 2
edb@89398=#INSERT INTO orders VALUES ( 2, 'UK', sysdate );
INSERT 0 1
edb@89398=#SELECT table_name, partition_name, subpartition_name FROM user_tab_subpartitions
WHERE table_name = ‘ORDERS’;
table_name | partition_name | subpartition_name
------------+----------------+-------------------
ORDERS | SYS893980105 | SYS893980106
ORDERS | SYS893980105 | SYS893980107
ORDERS | SYS893980105 | SYS893980108
(6 rows)

Part 3
● The subpartitions number can be altered using following command and the new number will be used to set up subpartitions in subsequent
partitions created
ALTER TABLE tbl1 SET SUBPARTITION TEMPLATE 100;
-- The following will reset the subpartition count to 1
ALTER TABLE tbl1 SET SUBPARTITION TEMPLATE ( );
● The template is used in all partition commands like ADD PARTITION, SPLIT PARTITION where a new partition is created. This template can
be overridden by explicit SUBPARTITION description or SUBPARTITIONS <num>
ALTER TABLE ORDERS ADD PARTITION p2 VALUES (‘AUS’) SUBPARTITIONS 10;
d orders_p2
...
Partition of: orders FOR VALUES IN (‘AUS’)
Partition key: HASH (col2)
Number of partitions: 10 (Use d+ to list them.)
-- Similar also works when do the SPLIT PARTITION
ALTER TABLE tbl1 SPLIT PARTITION p1 VALUES (10, 20) INTO (PARTITION p1_a, PARTITION p1_b);

• Frequently requested feature.
• edbldr used to abort the whole operation if any record insertion failed due to unique constraint
violation.
• Fix this by using speculative insertion to insert rows.
• It happens only when 'handle_conflicts' is true and indexes are present.
• Similar to how it is done for "ON CONFLICT ... DO NOTHING". Except, the record goes to the BAD
file.
EDB*LOADER
edb*loader duplicate row aborts

EPAS v13 Features
● “USING INDEX” clause in the CREATE TABLE and ALTER TABLE.
● STATS_MODE aggregate function
● MEDIAN: add smallint, int, bigint, and numeric variants
● Added support for DEFINE_COLUMN_LONG, COLUMN_VALUE_LONG
and LAST_ERROR_POSITION additional function/procedure in
DBMS_SQL Package.
● Support for "utl_http.end_of_body" exception
● Support for function to_timestamp_tz() in EPAS.
● PARALLEL/NOPARALLEL option for CREATE TABLE and INDEX.
● Create index syntax contains column name and number i.e.
(col_name,1).
● Log the number of rows processed for bulk execution.
● Consistency across CSV and XML audit logs.
● Edbldr to support all connection parameters like psql and other clients.
● Support for function/procedure forward declaration inside package
body.
● Add support for AES192 & AES256 in DBMS_CRYPTO package.
● Enhance Redwood compatible view.
● Allow creating a compound trigger having WHEN clause with NEW/OLD
variables and STATEMENT level triggering events.
● Fix split partition behavior to be more like redwood.
● Enhancement around edbldr and multi-insert.
● More enhancement in EDB-SPL language.
● Support FM format in to_number() function.
● SYSDATE to behave more like Oracle (STABLE).

• TPS improvements of over 50% in 4 years
• Parallel vacuum of indexes: 4 X faster
• New and enhanced capabilities
• Vacuum for append only tables
• Security and consistency
• Data Loading
• Partitioning
Summary
Postgres is getting faster and more capable

Energy
Build capabilities and
momentum to push
PostgreSQL further
Expertise Education
EDB supercharges PostgreSQL
We focus on 3 things:
If you’re looking to do more and go faster, plug in.
Deliver decades of
experience into every
PostgreSQL deployment
Enable teams to
understand the full
potential of PostgreSQL

Core team Major contributors Contributors
EDB Open Source Leadership
Named EDB open source committers and contributors
Akshay Joshi Amul Sul Ashesh Vashi Ashutosh Sharma Jeevan Chalke
Dilip Kumar Jeevan Ladhe Mithun Cy Rushabh Lathia Amit Khandekar
Amit Langote Devrim Gündüz
Robert Haas
Bruce Momjian
Dave Page
Designates PostgreSQL committers

• Enterprise PostgreSQL innovations
• 4,000+ global customers
• Recognized by Gartner Magic Quadrant for 7 years in a row
• One of the only sub-$1bn revenue companies
• PostgreSQL community leadership
2019
Challengers Leaders
Niche Players Visionaries
Abilitytoexecute
Completeness of vision
1986
The Design
of PostgreSQL
1996
Birth of
PostgreSQL
2004
EDB
is founded
2020
TodayMaterialized
Views
Parallel
Query
JIT
Compilation
Heap Only
Tuples (HOT)
Serializable
Parallel Query
We’re database fanatics who care
deeply about PostgreSQL
Expertise

© Copyright EnterpriseDB Corporation, 2020. All rights reserved.35 © Copyright EnterpriseDB Corporation, 2020. All rights reserved.35
IT Director
Chief
Architect
DevOps
Manager
Developer
Database
Administrator
Enabling teams to understand the full potential of PostgreSQL
Education

Market Success Globally

Market Success in Asia Pacific & Japan

Contact EDB in APJ
Other resources
Thank You
Australia & New Zealand:
E: sales-anz@enterprisedb.com
S E Asia & Hong Kong:
E: sales-sea-hk@enterprisedb.com
Japan:
E:jp@enterprisedb.com
S Korea:
E: sales-kr@enterprisedb.com
India, Sri Lanka & Bangladesh:
E: sales-india@enterprisedb.com
Postgres
Pulse
EDB Youtube Channel

PostgreSQL 13 is Coming - Find Out What's New!

More Related Content

What's hot (20)

Similar to PostgreSQL 13 is Coming - Find Out What's New! (20)

More from EDB (20)

Recently uploaded (20)

PostgreSQL 13 is Coming - Find Out What's New!