Oracle 12.2 - My Favorite Top Five New or Improved Features: Janis Griffin Senior DBA / Performance Evangelist

Oracle 12.
2 – My Favorite Top Five

New or Improved Features
Janis Griffin
Senior DBA / Performance Evangelist
Who Am I
• Senior DBA / Performance Evangelist for SolarWinds

• Janis.Griffin@solarwindscom
• Twitter® - @DoBoutAnything
• Current – 25+ Years in Oracle®, SQL Server®, ASE, MySQL®
• DBA and Developer
• Specialize in Performance Tuning
• Review Database Performance for Customers and Prospects
• Common Question – How do I tune it?
© 2018 SolarWinds Worldwide, LLC. All rights reserved.

Agenda
• In-Memory Option
• How it works
• Virtual Columns, join groups
• SQL, Dictionary, and Optimizer Enhancements
• New Functions and Syntax
• Approximate and First n Rows
• SQLPlus Improvements
• PDB Improvements
• Hot Clone
• Memory Management
• New Partitioning Features
• Sharding Capabilities
• The ‘Horizontal’ Database
Memory – In The Beginning…
• 9i - SGA_MAX_SIZE
• Memory could grow/shrink
• DB_CACHE_SIZE
• No more DB_BLOCK_BUFFERS
• 10g - Automatic Shared Memory Management (ASMM)
• SGA_TARGET
• Set minimum values for
• Data Cache, Large Pool
• Shared Pool, Java Pool
• 11g - MEMORY_TARGET (AMM)
• Sets SGA + PGA size together
• Can still control minimum sizes - e.g., PGA_AGGREGATE_LIMIT (new in 12c)
• 12c - INMEMORY_SIZE
• Set the size of IN-MEMORY Column Store
12c In-Memory Option
• Designed for mixed workloads

• Can combine OLAP with OLTP queries
• No more Data Warehouse (maybe)
• Dual-format architecture doesn’t double memory requirements
• Approximately less than a 20% additional memory overhead
• No additional storage costs or synchronization issues
• No changes to application code
• Optimizer is fully aware of the column format
• Routes analytic queries to the column format
• OLTP operations to the row format
• Only objects with the INMEMORY attribute
• Can be in the IN-Memory Column Store (IM)
• Works on tablespaces, tables, partitions, sub-partitions or materialized views
12c In-Memory Option
Transactions in Analytics in In-Memory

Buffer Cache Column Store
http://www.oracle.com/technetwork/database/in-memory/overview/twp-oracle-database-in-memory-2245633.pdf

IM – Why Use it?
• Access only the column data needed

• Can prioritize when IM is populated
• Partially load some columns at startup
• Load other columns when accessed
ALTER TABLE customers INMEMORY PRIORITY CRITICAL;
• In compressed format
• Saves space
• Can scan or filter compressed data
• Only decompressed when used in the result set
CREATE TABLE test (c1 NUMBER, c2 NUMBER, c3 CLOB)
INMEMORY MEMCOMPRESS FOR QUERY NO INMEMORY(c3),
INMEMORY MEMCOMPRESS FOR CAPACITY HIGH (c2);
• DBMS_COMPRESSION (advisor)
• Now supports IM compression
IM – Why So Fast
• IM is made up of multiple In-Memory Compression Units (IMCUs)

• Background worker processes allocate their own IMCUs
• Then populates their subset of database blocks
• Works with the IMCO (In-Memory Coordinator) background process Best for queries that scan
• Uses SIMD Vector processing large amounts of data or
compute values on the fly
• Definition: Single Instruction processing Multiple Data values
• Allows a set of column values to be evaluated simultaneously in one CPU instruction
• Look at V$STATNAME
• ‘IM scan CUs columns accessed’, 'IM scan segments minmax eligible', 'IM scan CUs pruned');
• New in 12.2 – IM FastStart
• Allows for checkpoint of IMCUs to FastStart area on disk
• Takes up storage space but less CPU intensive on startup
BEGIN
dbms_inmemory_admin.faststart_enable('FS_TBS');
END;
IM – Why So Fast
• IM Storage Indexes
• Automatically created and maintained on each column in the column store
• Allow pruning to occur based on filter predicates in the query
• Keeps track of min/max values for each column in an IMCU
• Helps to avoid scans of IM column store
• New in 12.2 - In-Memory Expressions
• Materialize commonly used expressions
• Prevents re-computation for every row
• Are derived values that take advantage of IM
• Can be done in one of 2 ways
• Expression Statistics Store(ESS) automatically stores top 20 popular expressions
• Manage with DBMS_INMEMORY_ADMIN package
• Set with INMEMORY_EXPRESSIONS_USAGE parameter
• In-Memory Virtual Columns
• Needs INMEMORY_VIRTUAL_COLUMNS parameter set to ENABLED or MANUAL
IM - How To Use It
• IM column store is disabled by default (requires re-start)

• Set INMEMORY_SIZE parameter to enable
• Needs to be set to 100mb or higher
• New in 12.2 - Can increase on the fly at 128mb or higher
• Part of SGA
• May need to increase SGA_TARGET / MEMORY_TARGET

IM - Enable It

IM – How To Use It
• Objects must have the INMEMORY attribute to use the IM column store
• Can be tablespace, table, (sub)partition or materialized view
• To view objects in column store – see V$IM_SEGMENTS, V$IM_COLUMN_LEVEL, any V$IM% views

IM Performance
• No IM IM
•

IM Virtual Columns Example – New 12.2 Feature
alter system set inmemory_virtual_columns = ENABLE;

IM Joins
• In-Memory Joins
• Bloom Filters (10g)
• Transforms join into a filter if there are filtering predicates
• Can be used in column format via SIMD vector processing
• New in 12.2 – Join Groups
• Can be used when there are no filtering predicates
• Allow join columns from multiple tables to share a single compression dictionary
• Hash joins don’t have to decompress the data first
• View join groups in USER_JOINGROUPS
• View shared dictionary in v$IM_SEGDICT
CREATE INMEMORY JOIN GROUP jgroup_name(order_item(ol_i_id),item(i_id));

IM Join Group Example – Cont.
SELECT ol_o_id,ol_number, ol_delivery_d, i_name, i_price

FROM order_item, item
WHERE ol_i_id = i_id and i_price < 2.00;
• Join Groups share the same compression dictionary

• This allows for joins to occur on compressed values
• Doesn’t need to decompress and hash the data
• Join columns above are order_item.ol_i_id and item.i_id
• Tables need to be reloaded after the Join Group creation

• To create the common compression dictionary, called Global Dictionary
• Ensure the Global Dictionary exists by checking V$IM_SEGDICT_GD
SELECT o.object_name table_name,

c.column_name column_name,
gd.head_address "GD Address"
FROM user_objects o,
user_tab_columns c,
v$im_segdict gd
WHERE gd.objn = o.object_id
AND o.object_name = c.table_name
AND gd.column_number = c.column_id;

IM Join Group Example – New 12.2 Feature
Using Bloom Filter / Join Group?
No IM

IM Join Group Example
• Finding join group usage

• Need to use SQL_MONITOR - dbms_sqltune.report_sql_monitor_xml or OEM
• Query needs ‘SELECT /*+ MONITOR */’ hint
• Columnar encodings leveraged = 1
Columnar Encodings
Leveraged

New SQL
• 12.1
• FETCH FIRST n ROWS ONLY
• Retrieves first rows without scanning everything
• Faster
• OFFSET n ROWS FETCH FIRST n ROWS ONLY
• Skip some number of rows

New SQL
• 12.2 Approximate Query Processing

• Used for approximate ‘count distinct’ values and adds percentile aggregation
• Allows for faster processing of large data sets
• Not exact but usually within 95%+ range
• Three new parameters – alter system/session
• approx_for_aggregation Default=FALSE
• Can be overridden by the next 2 parameters
• If true, sets approx_for_percentile=ALL
• approx_for_count_distinct Default=FALSE
• Overrides exact COUNT DISTINCT clause
• approx_for_percentile Default=NONE
• Overrides MEDIAN clause (PERCENTILE_CONT)
• Values can be PERCENTILE_CONT, PERCENTILE_DISC, and ALL
• Can be used without any changes to existing code
• Replaces exact functions with SQL functions (next page) that return approximate results

New SQL functions in 12.2
• Approximate query functions

• APPROX_COUNT_DISTINCT (Introduced in 12.1)
• APPROX_COUNT_DISTINCT_DETAIL
• APPROX_COUNT_DISTINCT_AGG
• TO_APPROX_COUNT_DISTINCT
• APPROX_MEDIAN
• APPROX_PERCENTILE
• APPROX_PERCENTILE_DETAIL
• APPROX_PERCENTILE_AGG
• TO_APPROX_PERCENTILE
• Also in 12.2, support for Materialized Views and Query Rewrite
• VALIDATE_CONVERSION
• Determines if a given input value can be converted to a requested data type
• Simplifies coding
Approximate SQL Example
95% accurate

Approximate SQL Example With Changing Code
Why is it exact?
Need to
set both

Approximate Percentile Example With Changing Code
99.68% accurate

12.2 Dictionary Improvements and Other New Stuff
• Object identifiers have increased from 30 bytes to 128 bytes

• Such as users, roles, tables, columns, indexes, constraints, etc…
• Changed in dictionary views - instead of VARCHAR2(30), now VARCHAR2(128)
• May affect your reporting
• Exceptions:
• Disk groups, pluggable databases (PDBs), rollback segments, and tablespaces still 30 bytes
• Names of databases are still limited to 8 bytes
• SQLPlus Improvements
• HISTORY
• SET HIST[ORY] {ON | OFF | n}
• SHOW HISTORY
• RUN n
• SET ROWPREFETCH {1 | n} / SET LOBPREFETCH {0 | n}
• SET STATEMENTC[ACHE] {0 | n}

12.2 Optimizer Changes
• New init.ora parameters

• OPTIMIZER_ADAPTIVE_PLANS (Default = TRUE)
• Adaptive joins
• Bitmap pruning
• Parallel distribution method
• OPTIMIZER_ADAPTIVE_STATISTICS (Default = False)

• SQL Plan Directives (SPDs) for query optimization
• Statistics feedback (for joins only)
• Adaptive dynamic sampling for parallel queries
• Performance feedback
• Obsolete
• OPTIMIZER_ADAPTIVE_FEATURES

12.2 Optimizer Changes
• SQL Plan Directives

• Control of Auto Creation of Column Group Statistics
• New DBMS_STATS preference
• AUTO_STAT_EXTENSIONS (DEFAULT=OFF)
EXEC DBMS_STATS.SET_GLOBAL_PREFS('AUTO_STAT_EXTENSIONS','ON')
• Can view in DBA_STAT_EXTENSIONS

select owner,
table_name,
extension,
extension_name
from dba_stat_extensions
where creator = 'SYSTEM'
order by owner,table_name,extension_name;
Multitenant Changes in 12.2
• Can create a hot clone of PDB

• No downtime for PDB being cloned
• Uses redo then undo for all committed / uncommitted transactions
• Local undo required
• Undo can now exist in PDB
• Archive logging must be enabled
• Can create a refreshable PDB (for VLDBs)
• Built on top of hot clone
• Golden master for snapshot clones
• Automatically or manually refreshed from redo
• Opened in read-only mode Can help maintain SLAs for
performance and availability
• Can relocate PDBs to different CDBs online
• Built on top of Refreshable PDB
• Great way to load-balance
Multitenant Changes in 12.2
• Can have 4096 PDBs per CDB

• Instead of original 252
• Can flashback a single PDB
• Doesn’t have to use local undo but easier than shared undo
• Can use SCN, restore point, clean restore point or guarantee restore point
• View SCNs in V$FLASHBACK_DATABASE_LOG
• https://oracle-base.com/articles/12c/multitenant-flashback-pdb-12cr2
FLASHBACK PLUGGABLE DATABASE pdb1 TO TIMESTAMP some_date;
• Data Guard now supports individual PDB-level failover
• Used to be all PDBs or none
• Now can set parameter ENABLED_PDBS_ON_STANDBY on standby database
• Examples – ‘*’, ‘PDB?’, ‘PDB*’, ‘-PDB1’

New Memory Settings At PDB Level
• If using Database Smart Flash Cache

• Can set CDB resource plan to allocate memory between PDBs
• MEMORY_LIMIT – percentage that limits PDB memory usage
• MEMORY_MIN – percentage guaranteed for PDB memory usage
• http://docs.oracle.com/cd/E80920_01/SAGUG/exadata-storage-server-iorm.htm#SAGUG20421
• Can set INMEMORY_SIZE at PDB

• Can be over-subscribed

12.2 Partitioning Features
• Read-only partitions
• Useful for old partitions
• Use READ ONLY
• READ WRITE (default)
• Can use CTAS or ALTER

• Convert Non-partitioned table

• To a partitioned table ONLINE
UPDATE INDEXES clause is

optional. Indexes with SYS
names will be generated if not
used

• Automatic List Partitions Multi-column List Partitions

• Requires creation of first partition Can only have one default partition
• No default partitions

• Filtered Partitions for Maintenance Operations

• Helps with partition pruning and data cleanup
• MOVE PARTITION
• MERGE PARTITION
• SPLIT PARTITION
• Can do ONLINE
• Other New Features

• ‘For exchange with’ clause
• Partitioned external tables
• 12.2-partition-create-tables

12.2 Sharding Feature
• Sharding horizontally partitions data across independent databases

• A ‘share nothing’ architecture – CPU, memory, disk
• Each database is known as a shard
• All shards together make up a logical database
• Called sharded database (SDB)
• Intended for custom OLTP applications that:
• Have a well-defined data model
• Have a data distribution strategy
• (consistent hash, range, list, or composite)
• Primarily access data using a sharding key https://docs.oracle.com/database/122/ADMIN/sharding-overview
• e.g., customer_id, account_no, or country_id

12.2 Sharding Feature
• Benefits of Sharding
• Linear scalability for performance and capacity
• Eliminates single points of failure
• Nothing is shared
• Geographical distribution of data
• Data is closer to consumers (e.g. Facebook)
• Satisfies regulatory requirements
• Simplifies Cloud Deployment
• Easy to do rolling upgrades
• Examples of Companies
• Online payment systems
• Regulatory requirements
• Airline ticketing systems
• Social media companies http://www.oracle.com/technetwork/database/availability/oraclesharding-whitepaper

Other SDB Components
• Shard Catalog
• An Oracle database that centralizes management of a SDB
• Supports automated shard deployment
• Handles multi-shard queries
• Holds master copy of duplicated table
• Shard Directors
• Network listeners which route connections based on a sharding key
• Connection pools can act as shard directors for pooled connections
• Global service
• An extension of database services for the SDB
• Management Interfaces
• GDSCTL (command-line utility) and OEM

SDB Architecture
To install, download
Oracle Database 12c Release 2 :
linuxx64_12201_database.zip
Oracle Database 12c Release 2 Global Service Manager :
linuxx64_12201_gsm.zip
https://docs.oracle.com/database/122/ADMIN/sharding-overview
Summary
• Oracle 12.2 provides a lot of new features and enhancements

• This presentation only touches on a few
• In-Memory virtual columns
• Helps speed up analytical queries
• New approximate functions
• Great if you don’t need exact answers
• PDB hot cloning
• Online partitioning
• Sharded database (SDB) or the ‘Horizontal database’
• Good news – it’s been around in the Cloud for quite sometime
• Became available for on-premise in March 2017
• Hopefully many of the bugs have been worked out
• Try it out and tell me which new feature you like best
• Email me at janis.griffin@solarwinds.com
Resolve Performance Issues quickly—Free Trial
• Try Database Performance Analyzer FREE for 14 days

• Improve root cause of slow performance
o Quickly identify root cause of issues that impact end-user
response time
o See historical trends over days, months, and years
o Understand impact of VMware® performance
o Agentless architecture with no dependence on Oracle Packs,
installs in minutes
www.solarwinds.com/dpa-download/

Thank You!!!
The SolarWinds, SolarWinds & Design, Orion, and THWACK trademarks are the exclusive
property of SolarWinds Worldwide, LLC or its affiliates, are registered with the U.S.
Patent and Trademark Office, and may be registered or pending registration in other
countries. All other SolarWinds trademarks, service marks, and logos may be common
law marks or are registered or pending registration. All other trademarks mentioned
herein are used for identification purposes only and are trademarks of (and may be
registered trademarks) of their respective companies.

Oracle 12.2 - My Favorite Top Five New or Improved Features: Janis Griffin Senior DBA / Performance Evangelist

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Oracle 12.2 - My Favorite Top Five New or Improved Features: Janis Griffin Senior DBA / Performance Evangelist

Uploaded by

Copyright:

Available Formats

Oracle 12.

2 – My Favorite Top Five

• Senior DBA / Performance Evangelist for SolarWinds

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

• Designed for mixed workloads

Transactions in Analytics in In-Memory

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

• Access only the column data needed

• IM is made up of multiple In-Memory Compression Units (IMCUs)

• IM column store is disabled by default (requires re-start)

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

alter system set inmemory_virtual_columns = ENABLE;

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

CREATE INMEMORY JOIN GROUP jgroup_name(order_item(ol_i_id),item(i_id));

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

SELECT ol_o_id,ol_number, ol_delivery_d, i_name, i_price

• Join Groups share the same compression dictionary

• Tables need to be reloaded after the Join Group creation

SELECT o.object_name table_name,

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

Using Bloom Filter / Join Group?

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

• Finding join group usage

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

• 12.2 Approximate Query Processing

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

• Approximate query functions

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

• Object identifiers have increased from 30 bytes to 128 bytes

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

• New init.ora parameters

• OPTIMIZER_ADAPTIVE_STATISTICS (Default = False)

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

• SQL Plan Directives

• Can view in DBA_STAT_EXTENSIONS

• Can create a hot clone of PDB

• Can have 4096 PDBs per CDB

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

• If using Database Smart Flash Cache

• Can set INMEMORY_SIZE at PDB

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

• Convert Non-partitioned table

UPDATE INDEXES clause is

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

• Automatic List Partitions Multi-column List Partitions

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

• Filtered Partitions for Maintenance Operations

• Other New Features

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

• Sharding horizontally partitions data across independent databases

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

© 2018 SolarWinds Worldwide, LLC. All rights reserved.

• Oracle 12.2 provides a lot of new features and enhancements

• Try Database Performance Analyzer FREE for 14 days

© 2018 SolarWinds Worldwide, LLC. All rights reserved.