Sunteți pe pagina 1din 32

Oracle

Database In-Memory
Whats New and Whats Coming

Juan Loaiza
Senior Vice President
Oracle Database Systems
October 27, 2015
#DBIM12c
Copyright 2015, Oracle and/or its aliates. All rights reserved. |

Safe Harbor Statement


The following is intended to outline our general product direcVon. It is intended for
informaVon purposes only, and may not be incorporated into any contract. It is not a
commitment to deliver any material, code, or funcVonality, and should not be relied upon
in making purchasing decisions. The development, release, and Vming of any features or
funcVonality described for Oracles products remains at the sole discreVon of Oracle.

Copyright 2015, Oracle and/or its aliates. All rights reserved. |

What is Database
In-Memory

Copyright 2015, Oracle and/or its aliates. All rights reserved. |

Oracle Database In-Memory Goals


Real-Time Analy-cs
100X
100X

Accelerate Mixed
Workload
TransacVons

Enable Real-Time
Business Decisions

Risk-Free

Trivial to Implement

Proven Scale-Out,
Availability, Security

No ApplicaVon Changes
Not Limited by Memory

AnalyVcs

Run analyVcs on
OperaVonal Systems

Copyright 2015, Oracle and/or its aliates. All rights reserved. |

Row Format Databases vs. Column Format Databases


SALES

Columns
Stored
ConVguously

Transac-ons run faster on row format


Example: Query or Insert a sales order
Fast processing few rows, many columns

Query

SALES
Query

Rows Stored
ConVguously

Analy-cs run faster on column format


Example : Report on sales totals by region
Fast accessing few columns, many rows

Un-l Now Must Choose One Format and Suer Tradeos


Copyright 2015, Oracle and/or its aliates. All rights reserved. |

Breakthrough: Dual Format Database


Normal
Buer Cache

New In-Memory
Format

SALES

SALES

Row
Format

Column
Format

SALES

BOTH row and column


formats for same table
Simultaneously acVve and
transacVonally consistent
AnalyVcs & reporVng use new
in-memory Column format
OLTP uses proven row format

Copyright 2015, Oracle and/or its aliates. All rights reserved. |

Oracle In-Memory Columnar Technology


Pure In-Memory Columnar

Pure in-memory column format


Not persistent, and no logging
Quick to change data: fast OLTP

Enabled at table or parVVon


Only acVve data in-memory

SALES


2x to 20x compression typical
Available on all hardware
pladorms

Copyright 2015, Oracle and/or its aliates. All rights reserved. |

Orders of Magnitude Faster AnalyVc Data Scans


Memory

REGION

Example:
Find sales in
California region

CPU

Scans use super fast SIMD


vector instructions

CA

Vector Register

Load
mulVple
region
values

CA
CA

Each CPU core scans local


in-memory columns

Originally designed for


graphics & science

Vector
Compare
all values
an 1 cycle

Billions of rows/sec scan


rate per CPU core

CA

> 100x Faster

Row format is millions/sec

Copyright 2015, Oracle and/or its aliates. All rights reserved. |

Improvements to All Aspects of AnalyVc Query

CPU
Load
mulVple
region
values

Vector Register

Data Scans

Joins

In-Memory Aggrega-on

HASH JOIN
CA
CA
CA

Vector
Compare
all values
an 1 cycle

CA

Speed of memory
Scan and Filter only
the needed Columns
Vector InstrucVons

Table A

Table B

Convert Star Joins into 10X


Faster Column Scans
Search large table for values
that match small table

Create In-Memory
Report Outline that is
Populated during Fast Scan
Runs Reports Instantly

Copyright 2015, Oracle and/or its aliates. All rights reserved. |

10

Database In-Memory Accelerates Mixed Workloads


Complex OLTP is Slowed by AnalyVc Indexes
1 3
OLTP
Table
Indexes

Column Store Replaces AnalyVc Indexes

10 20
AnalyVc
Indexes
REPLACE

InserVng one row into a table requires updaVng Fast analyVcs on any columns
10-20 analyVc indexes: Slow!
Column Store not persistent so update cost
is much lower
Copyright 2015, Oracle and/or its aliates. All rights reserved. |

11

Database In-Memory Scales to Any Size


Scale-Out

Scale-Up

Combine with Flash and Disk


Holest
Data

DRAM

Scale-Out Across Servers to


Grow Memory and CPUs
In-Memory Queries
Parallelized Across Servers

Scale-Up on large SMPs


NUMA OpVmized

AcVve
Data

Cold Data

PCI FLASH

DISK

Easily place data on most


cost eecVve Ver
Simultaneously Achieve:
Speed of DRAM
I/Os of Flash
Cost of Disk

Copyright 2015, Oracle and/or its aliates. All rights reserved. |

12

Database In-Memory: Industrial Strength Availability


Data Guard & GoldenGate

Pure In-Memory format does not


change Oracles storage format,
logging, backup, recovery, etc.

All Oracles proven availability


technologies work transparently

Protection from all failures

RAC

ASM

RMAN

Node, site, corruption, human error, etc.


Copyright 2015, Oracle and/or its aliates. All rights reserved. |

13

Database In-Memory: Unique Fault Tolerance


Similar to storage mirroring
Duplicate in-memory
columns on another node
Enabled per table/partition
E.g. only recent data
Application transparent

Downtime eliminated by
using duplicate after failure

Only Available on Engineered Systems


Copyright 2015, Oracle and/or its aliates. All rights reserved. |

14

Database In-Memory: Trivial to Implement


Easy to Deploy

100% Compa-ble

Full Func-onality

No data migraVon

No applicaVon
changes

No SQL restricVons

Easy to Use

No complex setup
1. Set column store
size

2. Declare In-Memory tables

Copyright 2015, Oracle and/or its aliates. All rights reserved. |

15

Fastest Customer AdopVon Of All Time


Oracle Database In-Memory and Mul-tenant Op-ons

Copyright 2015, Oracle and/or its aliates. All rights reserved. |

16

Oracle In-Memory 2X Faster than HANA on SAPs Benchmark


Naviga-on Steps/Hour on SAP BW-EML

300K

Iden-cal
Servers

148K
SAP HANA
36 Cores, 1 Server

ORACLE
36 Cores, 1 Server

BW-EML is SAPs benchmark for Business Warehouse Workloads


SAP has not cerVed or endorsed the Oracle result
Copyright 2015, Oracle and/or its aliates. All rights reserved. |

17

Customer In-Memory Sessions here at Openworld

Monday 2:45pm
Room 301

Wednesday 1:45pm
Room 103

Monday 4:00pm
Room 310

Tuesday 4:00pm
Room 307

Wednesday 1:45pm Wednesday 1:45pm


Room 103
Room 103

Copyright 2015, Oracle and/or its aliates. All rights reserved. |

Wednesday 3:00pm
Room 254

Wednesday 1:45pm
Room 103

18

Whats Coming

Copyright 2015, Oracle and/or its aliates. All rights reserved. |

19

SPARC M7 Sooware in Silicon


TradiVonal DB algorithms too complex for chips
Big Change: In-memory algorithms are much simpler

SoYware
in Silicon

5 years ago Oracle iniVated a revoluVonary project


Build fastest ever microprocessor
Most processing cores (32) and concurrent threads (256)
Fastest Memory Bandwidth (160 GB/sec)

Add In-Memory DB operaVons directly on chip


Only high-volume CPU with naVve SQL opVmizaVons
Copyright 2015, Oracle and/or its aliates. All rights reserved. |

20

In-Memory Algorithms NaVvely Implemented in Silicon


SQL in Silicon

DB AcceleraVon

Capacity in Silicon

SPARC M7
SoYware
in Silicon

Decompression Engines

Silicon Secured Memory


Fine-Grained Memory
ProtecVon

Database SoYware
Support Shipping
Since Mid-Year


Copyright 2015, Oracle and/or its aliates. All rights reserved. |

21

SQL in Silicon: Database In-Memory AcceleraVon Engines


SIMD Vectors instrucVons are fast, but were designed for
graphics, not database

SPARC M7
Core

Core

Core

Core

Shared Cache
DB
Accel

DB
Accel

DB
Accel

DB
Accel

New SPARC M7 chip has 32 opVmized database


acceleraVon engines (DAX) built on chip
Independently process streams of columns
E.g. nd all values that match California
Up to 170 Billion rows per second!
Like adding 32 addiVonal specialized cores to chip
Using less than 1% of chip space

Copyright 2015, Oracle and/or its aliates. All rights reserved. |

22

Capacity in Silicon: Decompression Engines


Compression is key to puung more data in-memory
Decompression is far more import for databases than compression
Data is loaded once, queried many Vmes

Bit palern decompression in normal cores is slow


64 CPU cores needed to decompress at full memory speed

Doubles Memory
Capacity

SPARC M7 adds 32 opVmized decompress engines


Run bit-palern decompress at memory speed

Copyright 2015, Oracle and/or its aliates. All rights reserved. |

23

Silicon Secured Memory: Fine Grained Memory ProtecVon


Database In-memory places terabytes of data in memory
More vulnerable to corrupVon by bugs/alacks than storage

SPARC M7 locks memory as it is allocated so only the owner


can access it
Hidden color bits added to pointers (key), and content (lock)
Pointer color (key) must match content color or program is aborted
Hardware support eliminates performance impact

Helps prevent access o end of structure, stale pointer access,


malicious alacks, etc. plus improves developer producVvity

STOP

Memory
Pointers
Copyright 2015, Oracle and/or its aliates. All rights reserved. |

Memory
Content
24

When and How


Should I Use
In-Memory

Copyright 2015, Oracle and/or its aliates. All rights reserved. |

25

Geung The Most From In-Memory


Understand Where it Helps

Fast cars speed up travel, not meeVngs


In-Memory speeds up analyVc data access, not:
Network round trips, logon/logo
Parsing, PL/SQL, complex funcVons
Data processing (as opposed to access)
Complex joins or aggregaVons where not much data is ltered before processing

Load and select once Staging tables, ETL, temp tables

Know your bo\leneck!


Copyright 2015, Oracle and/or its aliates. All rights reserved. |

26

Geung The Most From In-Memory


The Driver Ma\ers

Avoid stop and go trac


Process data in sets of rows in the Database
Not one row at a Vme in the applicaVon

Plan ahead, take shortest route


Help the opVmizer help you: Gather representaVve staVsVcs using DBMS_STATS

Use all your cylinders


Enable parallel execuVon
In-Memory removes storage bollenecks allowing parallelism to increase

Copyright 2015, Oracle and/or its aliates. All rights reserved. |

27

In-Memory Use Cases

OLTP
Real-Vme reporVng directly on
OLTP source data
Removes need for separate ODS
Speeds data extracVon

Data Warehouse
Staging/ETL/Temp not a candidate
Write once, read once
All or a subset of FoundaVon Layer
For Vme sensiVve analyVcs
PotenVal to replace Access Layer
Copyright 2015, Oracle and/or its aliates. All rights reserved. |

28

Where can I get


more informaVon

Copyright 2015, Oracle and/or its aliates. All rights reserved. |

29

AddiVonal
Resources
Join the Conversa-on
hlps://twiler.com/db_inmemory
hlps://blogs.oracle.com/in-memory/
hlps://www.facebook.com/OracleDatabase
hlp://www.oracle.com/goto/dbim.html

Related White Papers


Oracle Database In-Memory White Paper
Oracle Database In-Memory AggregaVon Paper
When to use Oracle Database In-Memory
Oracle Database In-Memory Advisor

Related Videos
In-Memory YouTube Channel
Managing Oracle Database In-Memory
Database In-Memory and Oracle MulVtenant
Industry Experts Discuss Oracle Database In-Memory
Sooware on Silicon
Any Addi-onal Ques-ons
Oracle Database In-Memory Blog

Copyright 2015, Oracle and/or its aliates. All rights reserved. |

30

Q & A

If you have more ques-ons later, feel free to ask



Copyright 2015, Oracle and/or its aliates. All rights reserved. |

31

S-ar putea să vă placă și