Oracle Database In-Memory & Multitenant - TROUG Days'16 Istanbul

April 28-29, 2016
Oracle Database
In-Memory & Multitenant
Mahir M. Quluzade

2 Oracle Database In-Memory - Mahir M. Quluzade © 2016 http://www.mahir-quluzade.com
Mahir M. Quluzade
 Oracle ACE
 AZEROUG (Azerbaijan Oracle User Group)
 TROUG (Turkish Oracle User Group)
 OCE 12c, 11g
 Author of OTech Magazine and OTN Articles
 Central Bank - Oracle DBA
 Oracle Blog : http//www.mahir-quluzade.com

What is
Oracle Multitenant?

Oracle Database Architecture
 Oracle Instance
– SGA & PGA
– Background Processes
 Oracle Database
– Control Files
– Data Files
– Redo Logs (Online and Archived)
– Parameter and Password file
Pre-12c Oracle Database

Oracle Database Architecture
Requires memory, processes and database files
System Resources

New Multitenant Architecture
Memory and processes required at multitenant container level only
System Resources

Multitenant Architecture
Components of a Multitenant Container Database (CDB)
Pluggable Databases (PDBs)
PDBs
Root
CDB
CREATE DATABASE …
ENABLE PLUGGABLE DATABASE
Create CDB
CREATE PLUGGABLE DATABASE …
Create PDB

Multitenant Architecture – Dynamics
 PDBs share common SGA and
background processes
 Foreground sessions see only
the PDB they connect to

What is
Oracle Database In-Memory?

Oracle Database In-Memory Goals

Breakthrough: Dual Format Database

What is
changed in Oracle Instance?

What is changed in Oracle Instance?
 SGA - System Global Area
– Buffer Cache
– Shared Pool
– Redo Buffer
– Large Pool
– Other Memory Areas
– In-Memory Area
 Background Processes
– DBWn
– LGWR
– …
– Wnnn (worker processes ora_w00n_db_name)

In-Memory Area
 Contains two sub-pools:
– IMCU pool: Stores In Memory Compression Units (IMCUs)
– SMU pool: Stores Snapshot Metadata Units
(SMUs)
 IMCUs contain column formatted data
 SMUs contain metadata and transactional
information
 In-Memory area is static area
– 1MB Pool – Populated Columns Data
– 64KB Pool – Objects Metadata

How to install
and configure Oracle Database In-Memory?

 Oracle Database In-Memory
automatically installed as part of
Oracle Database 12c Release 1
(12.1.0.2)
 But not enabled by default
Installing: Oracle Database In-Memory

In-Memory Area
 In-Memory control parameters starting “INMEMORY_” prefix
 Additional for optimizer control parameter OPTIMIZER_INMEMORY_AWARE
Core Initialization Parameters

Starting in Oracle Database 12c Release 2 (12.2), can
increase INMEMORY_SIZE size dynamically by using
ALTER SYSTEM. If free memory is available in the
SGA, then the database allocates the memory to the IM
column store.
Note: You cannot use ALTER SYSTEM to reduce
INMEMORY_SIZE.
Enable In-Memory Area
SQL> alter system set memory_max_target=1324M scope=spfile;
SQL> alter system set inmemory_size=300M scope=spfile;
SQL> shutdown immediate;
SQL> startup
ORACLE instance started.
Total System Global Area 1392508928 bytes
Fixed Size 2924304 bytes
Variable Size 1040187632 bytes
Database Buffers 16777216 bytes
Redo Buffers 13852672 bytes
In-Memory Area 318767104 bytes
Database mounted.
Database opened.

How to configure In-Memory in Multitenant
Container Database?

Parameters in Multitenant Container Database
 V$SYSTEM_PARAMETERS.ISPDB_MODIFIABLE

Parameters in Multitenant Container Database

INMEMORY attributes in DDLs

Populating In-Memory Colum Store
 SQL> CREATE TABLESAPCE users DATAFILE … INMEMORY;
 SQL> ALTER TABLESPACE users INMEMORY;
– All new tables and materialized views in this tablespace will be enabled for IM
column store by default.
INMEMORY attribute for Tablespaces

 SQL> CREATE TABLE orders (ID NUMBER, …) INMEMORY;
 SQL> ALTER TABLE orders INMEMORY;
 Can exclude one or more not important columns for aggregation
– SQL> ALTER TABLE orders INMEMORY NO INMEMORY(createuser_id);
 Can disable IM Column Store for one or more partition
– SQL> ALTER TABLE orders MODIFY PARTITION orders_p2000 NO
INMEMORY;
INMEMORY attribute for Tables

 SQL> CREATE MATERIALIZED VIEW sales_mv INMEMORY
SELECT * FROM SALES;
 SQL> SELECT table_name, inmemory, inmemory_priority,
inmemory_distribute, inmemory_compression
FROM user_tables
WHERE table_name = ‘SALES_MV';
TABLE_NAME INMEMORY INMEMORY_PRI INMEMORY_DISTRI INMEMORY_COMPRESS
---------- -------- ------------ --------------- -------------------
SALE_MV ENABLED NONE AUTO FOR QUERY LOW
INMEMORY attribute for Materialized Views

 CRITICAL
– Object is populated immediately after the database is opened
 HIGH
 MEDIUM
 LOW
 NONE (default)
– Object only populated after they are scanned for the first time, if space available.
Controlled by the PRIORITY sub-clause of the INMEMORY clause
SQL> ALTER TABLE customers INMEMORY PRIORITY HIGH;

In-Memory Compression Levels
 NO MEMCOMPRESS
… Data is populated without any compression
 MEMCOMPRESSION FOR DML
… Minimal compression optimized for DML performance
 MEMCOMPRESSION FOR QUERY LOW
… Optimized for query performance (default)
 MEMCOMPRESSION FOR QUERY HIGH
… Optimized for query performance as well as space saving
 MEMCOMPRESSION FOR CAPACITY LOW
… Balanced with a greater bias towards as space saving
 MEMCOMPRESSION FOR CAPACITY HIGH
… Optimized for space saving
Controlled by the MEMCOMPRESS sub-clause of INMEMORY clause

SQL> ALTER TABLE customers INMEMORY
2 MEMCOMPRESSION NO COMPRESSION;
SQL> ALTER TABLE customers INMEMORY
2 MEMCOMPRESSION FOR CAPACITY LOW;

 Use DBMS_COMPRESSION
 Applies MEMCOMPRESS to
sample set of data from a table
 Returns estimated compression
ratio

In-Memory Column Store for RAC
 Each node in RAC have IM Column Store
– Same size recommended
– Object by default distributed across all nodes
 DISTRIBUTE BY ROWID RANGE
 DISTRIBUTE BY PARTITION
 DISTRIBUTE BY SUBPARTITION
– Consists of 1 IMCU objects distribute only one node
 Populate same object to IM Column Store on
every nodes
– Supported only Engineered Systems – EXADATA
– Using DUPLICATE clause
SQL> ALTER TABLE orders INMEMORY
2 DISTRIBUTE BY PARTITION;
SQL> ALTER TABLE orders INMEMORY
2 DUPLICATE ALL;

How Does
Database In-Memory Improve Performance?

Querying the In-Memory Column Store
 Compressed Columnar Storage
 SIMD Vector Processing
 In-Memory Storage Indexes
 In-Memory Joins and Aggregations

Compressed Columnar Storage
 Compressed Column Units
allows an analytic query
scan only data within the
required column

 The column format used in the IM column store has
been specifically designed to maximize the number of
column entries that can be loaded into the vector
registers on the CPU and evaluated in a single CPU
instruction
 Each CPU core scans local in-memory columns
 Scans use super fast SIMD vector instructions
– Originally designed for graphics & science
 Billions of rows/sec scan rate per CPU core
– Row format is millions/sec
 SIMD vector processing allows a set of column values
to be evaluated together in a single CPU instruction
SIMD Vector Processing

In-Memory Storage Indexes
 Each column is the made up of
multiple column units
 Min / max value is recorded for
each column unit in a storage
index
 Storage index provides partition
pruning like performance for ALL
queries

In-Memory Joins
 Bloom filters enable joins to be
converted into fast column scans
 Tried and true technology originally
released in 10g
 Same technique used to offload joins
on Exadata

In-Memory Joins
Example
SQL> SELECT *
2 FROM customer c,
3 orders o
4 WHERE c.id = o.customer_id

In-Memory Aggregations
 Without need to support objects such as indexes,
summary tables and materialized views
 New SQL Execution Operations
– KEY VECTOR and VECTOR GROUP BY
 Indicate plan is a vector transformation plan
 Vector transformation plan designed for
– Optimize queries involving aggregation and joins large
fact and multiple small dimension tables
– working with Star Queries

In-Memory Aggregations
 Vector Transformation Plan supports
– Aggregation operators :
 SUM, AVERAGE, MIN, MAX, STD and VARIANCE
 COUNT DISTINCT operations not supported by VECTOR GROUP BY
– The grouping syntax used in query is GROUP BY
 GROUPING SETS, GROUP BY ROLLUP not currently supported
– All joins must be equijoins
 Multiple columns of dimension table join the fact table with same join type
and same data type (for best performance, not be wrapped in a fuction )

Vector Transformation Plan
SQL> SELECT t.calendar_year,
2 p.prod_category,
3 SUM(quantity_sold)
4 FROM times t,
5 products p,
6 sales f
7 WHERE t.time_id = f.time_id
8 AND p.prod_id = f.prod_id
9 GROUP BY t.calendar_year,
10 p.prod_category;
Example

What’s
coming with Oracle Database 12.2?

What is coming with Oracle Database 12.2 ?
 Starting in Oracle Database 12c Release 2 (12.2), can increase INMEMORY_SIZE size
dynamically by using ALTER SYSTEM.
 In-Memory Expressions (IM expressions)
– IM expressions populate the results of frequently evaluated query expressions into the In-Memory
Column Store. A candidate for an IM expression might be (monthly_sales*12)/52.
 In-Memory virtual columns (IM virtual columns)
– IM virtual columns enable the IM column store to materialized some or all virtual columns on a table.
 Active Data Guard will Support In-Memory
In-Memory Enhancements

Additional Information
• Oracle Database 12c Release 12.1.0.2
Documentation
• Oracle Database In-Memory White Paper
• Oracle Database In-Memory Aggregation Paper
• When to use Oracle Database In-Memory

Your Questions
http://www.mahir-quluzade.com
http://www.youtube.com/mahoora
https://twitter.com/marzade
https://www.linkedin.com/in/mahirquluzade
https://www.facebook.com/mahir.quluzade
Contacts

Oracle Database In-Memory & Multitenant - TROUG Days'16 Istanbul

More Related Content

What's hot (20)

Viewers also liked (13)

Similar to Oracle Database In-Memory & Multitenant - TROUG Days'16 Istanbul (20)

Recently uploaded (20)

Oracle Database In-Memory & Multitenant - TROUG Days'16 Istanbul

Editor's Notes